ExternalVendorCode/corda
1
0
Fork 0
mirror of https://github.com/corda/corda.git synced 2025-01-01 02:36:44 +00:00
Code Issues Actions 2 Packages Projects Releases Wiki Activity
9abba8fe24
corda/makefile

2326 lines
72 KiB
Makefile
Raw Normal View History

revert accidental makefile change
2013-01-27 03:07:18 +00:00
MAKEFLAGS = -s
copy makefile, stdc++.cpp, and type-generator.cpp from mess project and began customizing them for this project
2007-06-03 23:16:07 +00:00
prepend copyright notice and license to all source files; add license.txt and readme.txt
2008-02-19 18:06:52 +00:00
name = avian
Have Makefile pull version from gradle.properties
2014-07-23 22:36:19 +00:00
version := $(shell grep version gradle.properties | cut -d'=' -f2)
don't include embedded boot classpath in libvm.a, only the executable
2008-01-19 00:54:36 +00:00
allow building with Java 7
2015-08-06 23:23:14 +00:00
java-version := $(shell $(JAVA_HOME)/bin/java -version 2>&1 \
| head -n 1 \
| sed 's/.*version "1.\([^.]*\).*/\1/')
enable Linux PowerPC build
2011-02-26 19:45:22 +00:00
build-arch := $(shell uname -m \
| sed 's/^i.86$$/i386/' \
add support for QNX on x86_32 All but one test is passing. The failure is due to the fact that QNX doesn't (in general) support calling fork(2) from a multithreaded process. Thus, we'll need to use spawn instead of fork/exec on QNX, which I'll attempt in a later commit. http://www.qnx.com/developers/docs/6.4.1/neutrino/getting_started/s1_procs.html
2012-07-18 00:21:17 +00:00
| sed 's/^x86pc$$/i386/' \
add support for FreeBSD on x86_64
2012-08-02 16:49:32 +00:00
| sed 's/amd64/x86_64/' \
add process=interpret support for Linux/ARM64 This makes all the tests pass for the platform=linux arch=arm64 process=interpret build. Next step: process=compile support.
2014-11-26 01:36:34 +00:00
| sed 's/^arm.*$$/arm/' \
| sed 's/aarch64/arm64/')
hello, world
2007-06-29 16:42:39 +00:00
speed up build (especially on Windows) by avoiding redundant shell command execution in makefile
2009-09-05 21:21:45 +00:00
build-platform := \
Merge branch 'master' into powerpc Conflicts: makefile src/assembler.h src/compile.cpp src/compiler.cpp src/compiler.h src/finder.cpp
2008-11-11 15:20:49 +00:00
$(shell uname -s | tr [:upper:] [:lower:] \
rename platform=darwin -> platform={ios,macosx}, fix ios simulator build
2014-06-22 18:30:52 +00:00
| sed \
add process=interpret support for Linux/ARM64 This makes all the tests pass for the platform=linux arch=arm64 process=interpret build. Next step: process=compile support.
2014-11-26 01:36:34 +00:00
-e 's/^mingw32.*$$/mingw32/' \
rename platform=darwin -> platform={ios,macosx}, fix ios simulator build
2014-06-22 18:30:52 +00:00
-e 's/^cygwin.*$$/cygwin/' \
-e 's/^darwin.*$$/macosx/')
superficial tweaks
2007-09-21 14:16:43 +00:00
refactor build system to support cross-compiling
2007-10-22 20:56:27 +00:00
arch = $(build-arch)
directly emit codeimage as a object (binaryToObject is statically linked in), as a stepping stone to including extra symbols in said codeimage
2012-05-01 19:16:32 +00:00
target-arch = $(arch)
rename platform=darwin -> platform={ios,macosx}, fix ios simulator build
2014-06-22 18:30:52 +00:00
fix cygwin/msys bootimage build
2009-06-01 18:12:29 +00:00
bootimage-platform = \
$(subst cygwin,windows,$(subst mingw32,windows,$(build-platform)))
rename platform=darwin -> platform={ios,macosx}, fix ios simulator build
2014-06-22 18:30:52 +00:00
fix cygwin/msys bootimage build
2009-06-01 18:12:29 +00:00
platform = $(bootimage-platform)
Various fixes (mostly making things more strict) to get the code to compile on Mac OS X
2007-09-19 16:22:19 +00:00
allow codegen targets (Architectures and Assemblers) to co-exist The primary motivation behind this is to allow all the different Assemblers to be built at once, on a single machine. This should dramatically reduce the time required to make sure that a particular change doesn't break the build for one of the not-so-common architectures (arm, powerpc) Simply pass "codegen-targets=all" to make to compile all src/codegen/<arch>/assembler.cpp. Note that while these architectures are built, they will not be fully- functional. Certain stuff is assumed to be the same across the entire build (such as TargetBytesPerWord), but this isn't the case anymore.
2013-02-12 15:15:30 +00:00
codegen-targets = native
make the default compilation mode 'fast'
2008-01-12 00:35:27 +00:00
mode = fast
misc makefile tweaks
2008-01-03 19:49:42 +00:00
process = compile
implement primitive testing framework and provide for GC stress testing
2007-07-16 01:03:02 +00:00
refine build directory naming system to avoid needing to 'make clean' upon changing build parameters
2009-03-14 21:16:05 +00:00
ifneq ($(process),compile)
options := -$(process)
endif
ifneq ($(mode),fast)
options := $(options)-$(mode)
endif
add optional LZMA support for compressing embedded JARs, boot images, and shared objects
2012-06-02 15:06:22 +00:00
ifneq ($(lzma),)
options := $(options)-lzma
endif
refine build directory naming system to avoid needing to 'make clean' upon changing build parameters
2009-03-14 21:16:05 +00:00
ifeq ($(bootimage),true)
options := $(options)-bootimage
add bootimage-test option to makefile This option specifies that the test classes should be AOT-compiled along with the class library, which allows us to test that everything works in AOT-compiled form as well as JIT-compiled form. This is primarily motivated by the need to test d906db6 (support for AOT-compilation of Java 8 lambda expressions). Note that I had to tweak Misc because it tested something that couldn't be done in an AOT build without a lot of extra work, and SystemClassLoader.getPackage because it was returning null when the requested package could not be populated with JAR manifest metadata. Technically, we probably *should* return null for packages that don't exist at all (in the sense that no classes have been loaded from such a package), but tracking that kind of thing seems like more trouble than it's worth unless someone complains about it.
2015-10-19 21:10:53 +00:00
ifeq ($(bootimage-test),true)
# this option indicates that we should AOT-compile the test
# classes as well as the class library
options := $(options)-test
endif
refine build directory naming system to avoid needing to 'make clean' upon changing build parameters
2009-03-14 21:16:05 +00:00
endif
support tail calls and continuations as build options
2009-05-26 05:27:10 +00:00
ifeq ($(tails),true)
options := $(options)-tails
endif
ifeq ($(continuations),true)
options := $(options)-continuations
endif
allow codegen targets (Architectures and Assemblers) to co-exist The primary motivation behind this is to allow all the different Assemblers to be built at once, on a single machine. This should dramatically reduce the time required to make sure that a particular change doesn't break the build for one of the not-so-common architectures (arm, powerpc) Simply pass "codegen-targets=all" to make to compile all src/codegen/<arch>/assembler.cpp. Note that while these architectures are built, they will not be fully- functional. Certain stuff is assumed to be the same across the entire build (such as TargetBytesPerWord), but this isn't the case anymore.
2013-02-12 15:15:30 +00:00
ifeq ($(codegen-targets),all)
options := $(options)-all
endif
preliminary support for using OpenJDK's class library Whereas the GNU Classpath port used the strategy of patching Classpath with core classes from Avian so as to minimize changes to the VM, this port uses the opposite strategy: abstract and isolate classpath-specific features in the VM similar to how we abstract away platform-specific features in system.h. This allows us to use an unmodified copy of OpenJDK's class library, including its core classes and augmented by a few VM-specific classes in the "avian" package.
2010-09-10 21:05:29 +00:00
allow stress and stress-major as valid mode= settings in the makefile (note: all tests fail in both modes)
2014-06-30 17:33:51 +00:00
ifeq ($(filter debug debug-fast fast stress stress-major small,$(mode)),)
x := $(error "'$(mode)' is not a valid mode (choose one of: debug debug-fast fast stress stress-major small)")
add sanity checks for makefile parameters
2014-05-07 19:01:07 +00:00
endif
ifeq ($(filter compile interpret,$(process)),)
x := $(error "'$(process)' is not a valid process (choose one of: compile interpret)")
endif
add process=interpret support for Linux/ARM64 This makes all the tests pass for the platform=linux arch=arm64 process=interpret build. Next step: process=compile support.
2014-11-26 01:36:34 +00:00
ifeq ($(filter x86_64 i386 arm arm64,$(arch)),)
x := $(error "'$(arch)' is not a supported architecture (choose one of: x86_64 i386 arm arm64)")
add sanity checks for makefile parameters
2014-05-07 19:01:07 +00:00
endif
rename platform=darwin -> platform={ios,macosx}, fix ios simulator build
2014-06-22 18:30:52 +00:00
ifeq ($(platform),darwin)
fix malformed variable reference in makefile
2014-10-28 20:49:50 +00:00
x := $(error "please use 'platform=macosx' or 'platform=ios' instead of 'platform=$(platform)'")
rename platform=darwin -> platform={ios,macosx}, fix ios simulator build
2014-06-22 18:30:52 +00:00
endif
ifneq ($(ios),)
x := $(error "please use 'platform=ios' instead of 'ios=true'")
endif
ifeq ($(filter linux windows macosx ios freebsd,$(platform)),)
x := $(error "'$(platform)' is not a supported platform (choose one of: linux windows macosx ios freebsd)")
endif
ifeq ($(platform),macosx)
fix #373 (makefile syntax on mac)
2014-11-27 03:05:46 +00:00
ifneq ($(filter arm arm64,$(arch)),)
fix process=interpret build for 'arch=arm64 platform=ios'
2014-12-09 21:44:28 +00:00
x := $(error "please use ('arch=arm' or 'arch=arm64') 'platform=ios' to build for ios-arm")
rename platform=darwin -> platform={ios,macosx}, fix ios simulator build
2014-06-22 18:30:52 +00:00
endif
endif
ifeq ($(platform),ios)
add support for iOS/x86_64
2015-03-16 17:32:35 +00:00
ifeq ($(filter i386 x86_64 arm arm64,$(arch)),)
x := $(error "please specify 'arch=i386', 'arch=x86_64', 'arch=arm', or 'arch=arm64' with 'platform=ios'")
rename platform=darwin -> platform={ios,macosx}, fix ios simulator build
2014-06-22 18:30:52 +00:00
endif
add sanity checks for makefile parameters
2014-05-07 19:01:07 +00:00
endif
Makefile changes, added additional compiler flags
2013-01-30 05:31:02 +00:00
aot-only = false
rework OpenJDK build to derive classpath and library path from environment We now consult the JAVA_HOME environment variable to determine where to find the system library JARs and SOs. Ultimately, we'll want to support self-contained build, but this allows Avian to behave like a conventional libjvm.so.
2010-09-20 23:31:23 +00:00
root := $(shell (cd .. && pwd))
build = build/$(platform)-$(arch)$(options)
Makefile fix
2013-01-28 17:15:29 +00:00
host-build-root = $(build)/host
rework OpenJDK build to derive classpath and library path from environment We now consult the JAVA_HOME environment variable to determine where to find the system library JARs and SOs. Ultimately, we'll want to support self-contained build, but this allows Avian to behave like a conventional libjvm.so.
2010-09-20 23:31:23 +00:00
classpath-build = $(build)/classpath
test-build = $(build)/test
src = src
classpath-src = classpath
test = test
add unittest framework
2013-02-15 03:54:20 +00:00
unittest = unittest
Updates to .gitignore to get rid of any possibility of checking in binaries in lib/distrib directories, and generalized method of setting location of win32 and win64 libs.
2012-02-20 23:25:37 +00:00
win32 ?= $(root)/win32
win64 ?= $(root)/win64
Windows Phone 8 / Windows RT initial support Conflicts: makefile
2013-01-28 15:20:52 +00:00
winrt ?= $(root)/winrt
wp8 ?= $(root)/wp8
rework OpenJDK build to derive classpath and library path from environment We now consult the JAVA_HOME environment variable to determine where to find the system library JARs and SOs. Ultimately, we'll want to support self-contained build, but this allows Avian to behave like a conventional libjvm.so.
2010-09-20 23:31:23 +00:00
preliminary support for using OpenJDK's class library Whereas the GNU Classpath port used the strategy of patching Classpath with core classes from Avian so as to minimize changes to the VM, this port uses the opposite strategy: abstract and isolate classpath-specific features in the VM similar to how we abstract away platform-specific features in system.h. This allows us to use an unmodified copy of OpenJDK's class library, including its core classes and augmented by a few VM-specific classes in the "avian" package.
2010-09-10 21:05:29 +00:00
classpath = avian
add bootimage-test option to makefile This option specifies that the test classes should be AOT-compiled along with the class library, which allows us to test that everything works in AOT-compiled form as well as JIT-compiled form. This is primarily motivated by the need to test d906db6 (support for AOT-compilation of Java 8 lambda expressions). Note that I had to tweak Misc because it tested something that couldn't be done in an AOT build without a lot of extra work, and SystemClassLoader.getPackage because it was returning null when the requested package could not be populated with JAR manifest metadata. Technically, we probably *should* return null for packages that don't exist at all (in the sense that no classes have been loaded from such a package), but tracking that kind of thing seems like more trouble than it's worth unless someone complains about it.
2015-10-19 21:10:53 +00:00
bootimage-classpath = $(classpath-build)
ifeq ($(bootimage-test),true)
bootimage-classpath = $(classpath-build):$(test-build)
endif
use absolute path for test-executable in makefile This is necessary to avoid crashes on OS X due to CFBundleGetMainBundle and friends not behaving well otherwise.
2011-10-01 00:40:43 +00:00
test-executable = $(shell pwd)/$(executable)
rework OpenJDK build to derive classpath and library path from environment We now consult the JAVA_HOME environment variable to determine where to find the system library JARs and SOs. Ultimately, we'll want to support self-contained build, but this allows Avian to behave like a conventional libjvm.so.
2010-09-20 23:31:23 +00:00
boot-classpath = $(classpath-build)
add support for accessing embedded JARs as if they were directories This allows OpenJDK to access time zone data which is normally found under java.home, but which we must embed in the executable itself to create a self-contained build. The VM intercepts various file operations, looking for paths which start with a prefix specified by the avian.embed.prefix property and redirecting those operations to an embedded JAR. For example, if avian.embed.prefix is "/avian-embedded", and code calls File.exists() with a path of "/avian-embedded/javahomeJar/foo.txt", the VM looks for a function named javahomeJar via dlsym, calls the function to find the memory region containing the embeded JAR, and finally consults the JAR to see if the file "foo.txt" exists.
2010-11-05 19:18:28 +00:00
embed-prefix = /avian-embedded
preliminary support for using OpenJDK's class library Whereas the GNU Classpath port used the strategy of patching Classpath with core classes from Avian so as to minimize changes to the VM, this port uses the opposite strategy: abstract and isolate classpath-specific features in the VM similar to how we abstract away platform-specific features in system.h. This allows us to use an unmodified copy of OpenJDK's class library, including its core classes and augmented by a few VM-specific classes in the "avian" package.
2010-09-10 21:05:29 +00:00
fix native Windows GCC 3.4 OpenJDK build
2010-11-13 01:29:12 +00:00
native-path = echo
rename platform=darwin -> platform={ios,macosx}, fix ios simulator build
2014-06-22 18:30:52 +00:00
platform-kernel = $(subst macosx,darwin,$(subst ios,darwin,$1))
build-kernel = $(call platform-kernel,$(build-platform))
kernel = $(call platform-kernel,$(platform))
fix native Windows GCC 3.4 OpenJDK build
2010-11-13 01:29:12 +00:00
ifeq ($(build-platform),cygwin)
native-path = cygpath -m
endif
add support for LZMA on Windows
2012-06-02 21:43:42 +00:00
windows-path = echo
fix native Windows GCC 3.4 OpenJDK build
2010-11-13 01:29:12 +00:00
path-separator = :
ifneq (,$(filter mingw32 cygwin,$(build-platform)))
path-separator = ;
endif
fix remote-test functionality for windows
2014-02-22 22:35:12 +00:00
target-path-separator = :
ifeq ($(platform),windows)
target-path-separator = ;
endif
fix OS X openjdk and openjdk-src builds
2010-11-16 03:28:53 +00:00
library-path-variable = LD_LIBRARY_PATH
rename platform=darwin -> platform={ios,macosx}, fix ios simulator build
2014-06-22 18:30:52 +00:00
ifeq ($(build-kernel),darwin)
fix OS X openjdk and openjdk-src builds
2010-11-16 03:28:53 +00:00
library-path-variable = DYLD_LIBRARY_PATH
endif
put test library in same directory as avian executable This is a workaround for the fact that there's no precise equivalent to LD_LIBRARY_PATH available on Windows.
2012-08-02 18:56:15 +00:00
library-path = $(library-path-variable)=$(build)
add JNI test to test suite
2012-08-01 16:02:55 +00:00
fix openjdk-src build failures on osx 10.9
2013-12-07 23:38:00 +00:00
fix openjdk-src bootimage build The main change here is to use a lazily-populated vector to associate runtime data with classes instead of referencing them directly from the class which requires updating immutable references in the heap image. The other changes employ other strategies to avoid trying to update immutable references.
2010-11-26 19:41:31 +00:00
ifneq ($(openjdk),)
fix OS X openjdk and openjdk-src builds
2010-11-16 03:28:53 +00:00
openjdk-arch = $(arch)
ifeq ($(arch),x86_64)
openjdk-arch = amd64
endif
add sanity checks for makefile parameters
2014-05-07 19:01:07 +00:00
ifneq ($(android),)
x := $(error "android and openjdk are incompatible")
endif
fix openjdk-src bootimage build The main change here is to use a lazily-populated vector to associate runtime data with classes instead of referencing them directly from the class which requires updating immutable references in the heap image. The other changes employ other strategies to avoid trying to update immutable references.
2010-11-26 19:41:31 +00:00
ifneq ($(openjdk-src),)
enable standalone OpenJDK builds As described in readme.txt, a standalone OpenJDK build embeds all libraries, classes, and other files needed at runtime in the resulting binary, eliminating dependencies on external resources.
2010-11-04 17:02:09 +00:00
include openjdk-src.mk
fix linux->darwin cross-compile There were two issues: the linux->darwin cross compiler is more stringent about unused variables, and the makefile specified flags for building ON darwin that were actually applicable whenever we are building FOR darwin.
2013-06-04 19:32:22 +00:00
options := $(options)-openjdk-src
various fixes to get Eclipse 3.6 working * add libnet.so and libnio.so to built-in libraries for openjdk-src build * implement sun.misc.Unsafe.park/unpark * implement JVM_SetClassSigners/JVM_GetClassSigners * etc.
2010-11-27 18:25:02 +00:00
classpath-objects = $(openjdk-objects) $(openjdk-local-objects)
add support for accessing embedded JARs as if they were directories This allows OpenJDK to access time zone data which is normally found under java.home, but which we must embed in the executable itself to create a self-contained build. The VM intercepts various file operations, looking for paths which start with a prefix specified by the avian.embed.prefix property and redirecting those operations to an embedded JAR. For example, if avian.embed.prefix is "/avian-embedded", and code calls File.exists() with a path of "/avian-embedded/javahomeJar/foo.txt", the VM looks for a function named javahomeJar via dlsym, calls the function to find the memory region containing the embeded JAR, and finally consults the JAR to see if the file "foo.txt" exists.
2010-11-05 19:18:28 +00:00
classpath-cflags = -DAVIAN_OPENJDK_SRC -DBOOT_JAVAHOME
enable standalone OpenJDK builds As described in readme.txt, a standalone OpenJDK build embeds all libraries, classes, and other files needed at runtime in the resulting binary, eliminating dependencies on external resources.
2010-11-04 17:02:09 +00:00
openjdk-jar-dep = $(build)/openjdk-jar.dep
classpath-jar-dep = $(openjdk-jar-dep)
add support for accessing embedded JARs as if they were directories This allows OpenJDK to access time zone data which is normally found under java.home, but which we must embed in the executable itself to create a self-contained build. The VM intercepts various file operations, looking for paths which start with a prefix specified by the avian.embed.prefix property and redirecting those operations to an embedded JAR. For example, if avian.embed.prefix is "/avian-embedded", and code calls File.exists() with a path of "/avian-embedded/javahomeJar/foo.txt", the VM looks for a function named javahomeJar via dlsym, calls the function to find the memory region containing the embeded JAR, and finally consults the JAR to see if the file "foo.txt" exists.
2010-11-05 19:18:28 +00:00
javahome = $(embed-prefix)/javahomeJar
add support for openjdk-src builds with OpenJDK 8 I've tested this on Linux with both OpenJDK 7 and 8. Other platforms will probably require further changes.
2014-04-23 23:09:52 +00:00
javahome-files = lib/currency.data lib/security/java.security \
only include security policy jars if they exist
2011-04-10 21:51:17 +00:00
lib/security/java.policy lib/security/cacerts
add support for openjdk-src builds with OpenJDK 8 I've tested this on Linux with both OpenJDK 7 and 8. Other platforms will probably require further changes.
2014-04-23 23:09:52 +00:00
ifneq (,$(wildcard $(openjdk)/jre/lib/zi))
javahome-files += lib/zi
endif
ifneq (,$(wildcard $(openjdk)/jre/lib/tzdb.dat))
javahome-files += lib/tzdb.dat
endif
only include security policy jars if they exist
2011-04-10 21:51:17 +00:00
local-policy = lib/security/local_policy.jar
add support for openjdk-src builds with OpenJDK 8 I've tested this on Linux with both OpenJDK 7 and 8. Other platforms will probably require further changes.
2014-04-23 23:09:52 +00:00
ifneq (,$(wildcard $(openjdk)/jre/$(local-policy)))
only include security policy jars if they exist
2011-04-10 21:51:17 +00:00
javahome-files += $(local-policy)
endif
export-policy = lib/security/US_export_policy.jar
add support for openjdk-src builds with OpenJDK 8 I've tested this on Linux with both OpenJDK 7 and 8. Other platforms will probably require further changes.
2014-04-23 23:09:52 +00:00
ifneq (,$(wildcard $(openjdk)/jre/$(export-policy)))
only include security policy jars if they exist
2011-04-10 21:51:17 +00:00
javahome-files += $(export-policy)
endif
fix Windows openjdk (non-openjdk-src) build
2010-11-07 17:08:04 +00:00
ifeq ($(platform),windows)
javahome-files += lib/tzmappings
endif
add support for accessing embedded JARs as if they were directories This allows OpenJDK to access time zone data which is normally found under java.home, but which we must embed in the executable itself to create a self-contained build. The VM intercepts various file operations, looking for paths which start with a prefix specified by the avian.embed.prefix property and redirecting those operations to an embedded JAR. For example, if avian.embed.prefix is "/avian-embedded", and code calls File.exists() with a path of "/avian-embedded/javahomeJar/foo.txt", the VM looks for a function named javahomeJar via dlsym, calls the function to find the memory region containing the embeded JAR, and finally consults the JAR to see if the file "foo.txt" exists.
2010-11-05 19:18:28 +00:00
javahome-object = $(build)/javahome-jar.o
various fixes for embedded resource loading in OpenJDK build
2010-12-06 00:40:50 +00:00
boot-javahome-object = $(build)/boot-javahome.o
fix openjdk-src build regression
2013-03-26 14:30:26 +00:00
stub-sources = $(src)/openjdk/stubs.cpp
stub-objects = $(call cpp-objects,$(stub-sources),$(src),$(build))
enable standalone OpenJDK builds As described in readme.txt, a standalone OpenJDK build embeds all libraries, classes, and other files needed at runtime in the resulting binary, eliminating dependencies on external resources.
2010-11-04 17:02:09 +00:00
else
only define soname-flag and version-script-flag for openjdk builds These flags are only needed to ensure that libjvm.so can be used as a drop-in replacement for OpenJDK's VM, and that only makes sense for openjdk builds (without the openjdk-src option). It didn't hurt anything to define them unconditionally, but it was misleading.
2014-04-12 01:32:16 +00:00
soname-flag = -Wl,-soname -Wl,$(so-prefix)jvm$(so-suffix)
version-script-flag = -Wl,--version-script=openjdk.ld
fix linux->darwin cross-compile There were two issues: the linux->darwin cross compiler is more stringent about unused variables, and the makefile specified flags for building ON darwin that were actually applicable whenever we are building FOR darwin.
2013-06-04 19:32:22 +00:00
options := $(options)-openjdk
use absolute path for test-executable in makefile This is necessary to avoid crashes on OS X due to CFBundleGetMainBundle and friends not behaving well otherwise.
2011-10-01 00:40:43 +00:00
test-executable = $(shell pwd)/$(executable-dynamic)
rename platform=darwin -> platform={ios,macosx}, fix ios simulator build
2014-06-22 18:30:52 +00:00
ifeq ($(build-kernel),darwin)
fix non-embedded openjdk build on OS X The JRE lib dir for OpenJDK 7 on OS X seems to be just "lib", not e.g. "lib/amd64" by default, so we use that now. Also, the default library compatibility version for libjvm.dylib is 0.0.0, but OpenJDK wants 1.0.0, so we set it explicitly.
2012-03-16 17:01:20 +00:00
library-path = \
put test library in same directory as avian executable This is a workaround for the fact that there's no precise equivalent to LD_LIBRARY_PATH available on Windows.
2012-08-02 18:56:15 +00:00
$(library-path-variable)=$(build):$(openjdk)/jre/lib
fix non-embedded openjdk build on OS X The JRE lib dir for OpenJDK 7 on OS X seems to be just "lib", not e.g. "lib/amd64" by default, so we use that now. Also, the default library compatibility version for libjvm.dylib is 0.0.0, but OpenJDK wants 1.0.0, so we set it explicitly.
2012-03-16 17:01:20 +00:00
else
library-path = \
put test library in same directory as avian executable This is a workaround for the fact that there's no precise equivalent to LD_LIBRARY_PATH available on Windows.
2012-08-02 18:56:15 +00:00
$(library-path-variable)=$(build):$(openjdk)/jre/lib/$(openjdk-arch)
fix non-embedded openjdk build on OS X The JRE lib dir for OpenJDK 7 on OS X seems to be just "lib", not e.g. "lib/amd64" by default, so we use that now. Also, the default library compatibility version for libjvm.dylib is 0.0.0, but OpenJDK wants 1.0.0, so we set it explicitly.
2012-03-16 17:01:20 +00:00
endif
fix native Windows GCC 3.4 OpenJDK build
2010-11-13 01:29:12 +00:00
javahome = "$$($(native-path) "$(openjdk)/jre")"
enable standalone OpenJDK builds As described in readme.txt, a standalone OpenJDK build embeds all libraries, classes, and other files needed at runtime in the resulting binary, eliminating dependencies on external resources.
2010-11-04 17:02:09 +00:00
endif
Makefile fix
2013-01-28 17:15:29 +00:00
classpath = openjdk
fix native Windows GCC 3.4 OpenJDK build
2010-11-13 01:29:12 +00:00
boot-classpath := "$(boot-classpath)$(path-separator)$$($(native-path) "$(openjdk)/jre/lib/rt.jar")"
add support for accessing embedded JARs as if they were directories This allows OpenJDK to access time zone data which is normally found under java.home, but which we must embed in the executable itself to create a self-contained build. The VM intercepts various file operations, looking for paths which start with a prefix specified by the avian.embed.prefix property and redirecting those operations to an embedded JAR. For example, if avian.embed.prefix is "/avian-embedded", and code calls File.exists() with a path of "/avian-embedded/javahomeJar/foo.txt", the VM looks for a function named javahomeJar via dlsym, calls the function to find the memory region containing the embeded JAR, and finally consults the JAR to see if the file "foo.txt" exists.
2010-11-05 19:18:28 +00:00
build-javahome = $(openjdk)/jre
add build option for using GNU Classpath
2009-06-03 00:55:49 +00:00
endif
refine build directory naming system to avoid needing to 'make clean' upon changing build parameters
2009-03-14 21:16:05 +00:00
first pass at Android classpath support Stuff compiles, but linking breaks spectacularly. Next step is to figure out how to build the dependencies without checking out and building the entire Android platform.
2013-02-19 16:36:19 +00:00
ifneq ($(android),)
options := $(options)-android
classpath-jar-dep = $(build)/android.dep
luni-native = $(android)/libcore/luni/src/main/native
all tests now pass for Android class library build
2013-02-23 00:23:59 +00:00
classpath-cflags = -DBOOT_JAVAHOME
Fix for https://github.com/ReadyTalk/avian/issues/216
2014-04-04 16:21:16 +00:00
android-cflags = -I$(luni-native) \
first pass at Android classpath support Stuff compiles, but linking breaks spectacularly. Next step is to figure out how to build the dependencies without checking out and building the entire Android platform.
2013-02-19 16:36:19 +00:00
-I$(android)/libnativehelper/include/nativehelper \
update to more recent version of Android class library Lots has changed since we forked Android's libcore, so merging the latest upstream code has required extensive changes to the Avian/Android port. One big change is that we now use Avian's versions of java.lang.Object, java.lang.Class, java.lang.ClassLoader, some java.lang.reflect.* classes, etc. instead of the Android versions. The main reason is that the Android versions have become very Dex/Dalvik-specific, and since Avian is based on Java class files, not dex archives, that code doesn't make sense here. This has the side benefit that we can share more native code with classpath-avian.cpp and reduce the amount of Java/C++ code duplication.
2014-08-21 19:09:42 +00:00
-I$(android)/libnativehelper \
use Android directory layout when accessing class library code Per https://github.com/ReadyTalk/avian/issues/53, Avian should build against a standard AOSP checkout, which means we should look for subprojects in the directories the repo utility would place them.
2013-04-22 23:10:23 +00:00
-I$(android)/system/core/include \
-I$(android)/external/zlib \
-I$(android)/external/icu4c/i18n \
-I$(android)/external/icu4c/common \
-I$(android)/external/expat \
-I$(android)/external/openssl/include \
fix SSL stack and Android JAR resource loading
2014-08-22 13:05:58 +00:00
-I$(android)/external/openssl \
first pass at Android classpath support Stuff compiles, but linking breaks spectacularly. Next step is to figure out how to build the dependencies without checking out and building the entire Android platform.
2013-02-19 16:36:19 +00:00
-I$(android)/libcore/include \
-I$(build)/android-src/external/fdlibm \
-I$(build)/android-src \
-fno-exceptions \
progress towards Android classpath support It now builds and links, but fails at runtime because register_libcore_icu_ICU can't find the file it wants. We'll probably need to replace register_libcore_icu_ICU with a better-behaved version.
2013-02-19 23:48:33 +00:00
-D_FILE_OFFSET_BITS=64 \
fix Android classpath build Hopefully it will stay fixed this time, since we now specify commit hashes for each dependency which are known to work.
2013-07-05 20:34:16 +00:00
-DOS_SHARED_LIB_FORMAT_STR="\"$(so-prefix)%s$(so-suffix)\"" \
-DJNI_JARJAR_PREFIX= \
fix Android classpath build for iOS This mainly involved reworking the makefile to avoid conflating Darwin/ARM builds with iOS, since we may also want to build for the iOS Simulator, which is i386. Note that I was only able to test this on the Simulator, since I don't have a real iOS device to test with. Sorry if I broke something; if so, please fix it :)
2013-07-31 22:20:32 +00:00
-D__DARWIN_UNIX03=1 \
Enabled "FIXME_STUB" macro
2014-03-03 20:41:58 +00:00
-D__PROVIDE_FIXMES \
fix SSL stack and Android JAR resource loading
2014-08-22 13:05:58 +00:00
-DSTATIC_LIB \
Fixes for avian-pack
2015-06-02 21:01:03 +00:00
-D__STDC_FORMAT_MACROS=1 \
enable use of Android class library on 64-bit systems This involved patching libcore.git to make it 64-bit safe.
2013-02-25 23:41:46 +00:00
-g3 \
a couple of tweaks to get Mac/Android tests passing
2014-09-02 16:14:04 +00:00
-Wno-shift-count-overflow
Added Class.getGenericSuperclass() and improved Member subclasses. Fixed Makefile on OS X
2014-10-08 22:32:56 +00:00
Fixed missing __BEGIN_DECLS / __END_DECLS on MinGW
2014-08-31 22:35:06 +00:00
# on Windows (in MinGW-based build) there are neither __BEGIN_DECLS nor __END_DECLS
# defines; we don't want to patch every file that uses them, so we stub them in
# using CFLAGS mechanism
Small fix for Android build on Windows
2014-10-09 13:50:09 +00:00
# Also we have off64_t defined in mingw-w64 headers, so let's tell that
Fixed missing __BEGIN_DECLS / __END_DECLS on MinGW
2014-08-31 22:35:06 +00:00
ifeq ($(platform),windows)
Small fix for Android build on Windows
2014-10-09 13:50:09 +00:00
android-cflags += "-D__BEGIN_DECLS=extern \"C\" {" "-D__END_DECLS=}" "-DHAVE_OFF64_T"
Fixed missing __BEGIN_DECLS / __END_DECLS on MinGW
2014-08-31 22:35:06 +00:00
endif
first step towards supporting Android class library on Windows A majority of the tests are passing, but more work is needed to get them all to pass.
2013-03-14 21:33:05 +00:00
fix Android classpath build Hopefully it will stay fixed this time, since we now specify commit hashes for each dependency which are known to work.
2013-07-05 20:34:16 +00:00
luni-cpps := $(shell find $(luni-native) -name '*.cpp')
update to more recent version of Android class library Lots has changed since we forked Android's libcore, so merging the latest upstream code has required extensive changes to the Avian/Android port. One big change is that we now use Avian's versions of java.lang.Object, java.lang.Class, java.lang.ClassLoader, some java.lang.reflect.* classes, etc. instead of the Android versions. The main reason is that the Android versions have become very Dex/Dalvik-specific, and since Avian is based on Java class files, not dex archives, that code doesn't make sense here. This has the side benefit that we can share more native code with classpath-avian.cpp and reduce the amount of Java/C++ code duplication.
2014-08-21 19:09:42 +00:00
libziparchive-native := $(android)/system/core/libziparchive
libziparchive-ccs := $(libziparchive-native)/zip_archive.cc
libutils-native := $(android)/system/core/libutils
libutils-cpps := $(libutils-native)/FileMap.cpp
fix Android classpath build Hopefully it will stay fixed this time, since we now specify commit hashes for each dependency which are known to work.
2013-07-05 20:34:16 +00:00
libnativehelper-native := $(android)/libnativehelper
libnativehelper-cpps := $(libnativehelper-native)/JniConstants.cpp \
$(libnativehelper-native)/toStringArray.cpp
fix SSL stack and Android JAR resource loading
2014-08-22 13:05:58 +00:00
crypto-native := $(android)/external/conscrypt/src/main/native
only define soname-flag and version-script-flag for openjdk builds These flags are only needed to ensure that libjvm.so can be used as a drop-in replacement for OpenJDK's VM, and that only makes sense for openjdk builds (without the openjdk-src option). It didn't hurt anything to define them unconditionally, but it was misleading.
2014-04-12 01:32:16 +00:00
crypto-cpps := $(crypto-native)/org_conscrypt_NativeCrypto.cpp
NativeCrypto added for Windows
2014-03-15 18:21:24 +00:00
first step towards supporting Android class library on Windows A majority of the tests are passing, but more work is needed to get them all to pass.
2013-03-14 21:33:05 +00:00
ifeq ($(platform),windows)
Fixed CFLAGS for Android classpath under MinGW
2014-09-01 20:20:38 +00:00
android-cflags += -D__STDC_CONSTANT_MACROS -DHAVE_WIN32_FILEMAP -D__STDC_FORMAT_MACROS
first step towards supporting Android class library on Windows A majority of the tests are passing, but more work is needed to get them all to pass.
2013-03-14 21:33:05 +00:00
blacklist = $(luni-native)/java_io_Console.cpp \
update to more recent version of Android class library Lots has changed since we forked Android's libcore, so merging the latest upstream code has required extensive changes to the Avian/Android port. One big change is that we now use Avian's versions of java.lang.Object, java.lang.Class, java.lang.ClassLoader, some java.lang.reflect.* classes, etc. instead of the Android versions. The main reason is that the Android versions have become very Dex/Dalvik-specific, and since Avian is based on Java class files, not dex archives, that code doesn't make sense here. This has the side benefit that we can share more native code with classpath-avian.cpp and reduce the amount of Java/C++ code duplication.
2014-08-21 19:09:42 +00:00
$(luni-native)/java_lang_ProcessManager.cpp
only define soname-flag and version-script-flag for openjdk builds These flags are only needed to ensure that libjvm.so can be used as a drop-in replacement for OpenJDK's VM, and that only makes sense for openjdk builds (without the openjdk-src option). It didn't hurt anything to define them unconditionally, but it was misleading.
2014-04-12 01:32:16 +00:00
Makefile fix for new ICU4C on Windows
2014-09-01 21:09:29 +00:00
icu-libs := $(android)/external/icu4c/lib/libsicuin.a \
$(android)/external/icu4c/lib/libsicuuc.a \
use Android directory layout when accessing class library code Per https://github.com/ReadyTalk/avian/issues/53, Avian should build against a standard AOSP checkout, which means we should look for subprojects in the directories the repo utility would place them.
2013-04-22 23:10:23 +00:00
$(android)/external/icu4c/lib/sicudt.a
Fixes for avian-pack
2015-06-02 21:01:03 +00:00
platform-lflags := -lgdi32 -lshlwapi -lwsock32 -static-libgcc -static-libstdc++ -Wl,-Bstatic -lstdc++ -lpthread -Wl,-Bdynamic
first step towards supporting Android class library on Windows A majority of the tests are passing, but more work is needed to get them all to pass.
2013-03-14 21:33:05 +00:00
else
update to more recent version of Android class library Lots has changed since we forked Android's libcore, so merging the latest upstream code has required extensive changes to the Avian/Android port. One big change is that we now use Avian's versions of java.lang.Object, java.lang.Class, java.lang.ClassLoader, some java.lang.reflect.* classes, etc. instead of the Android versions. The main reason is that the Android versions have become very Dex/Dalvik-specific, and since Avian is based on Java class files, not dex archives, that code doesn't make sense here. This has the side benefit that we can share more native code with classpath-avian.cpp and reduce the amount of Java/C++ code duplication.
2014-08-21 19:09:42 +00:00
android-cflags += -fPIC -DHAVE_SYS_UIO_H -DHAVE_POSIX_FILEMAP
blacklist =
fix JNIEnv::FindClass calls from JNI_OnLoad for all supported class libraries This also fixes the some Android build rot and updates the version of OpenSSL used.
2014-07-01 16:18:45 +00:00
use Android directory layout when accessing class library code Per https://github.com/ReadyTalk/avian/issues/53, Avian should build against a standard AOSP checkout, which means we should look for subprojects in the directories the repo utility would place them.
2013-04-22 23:10:23 +00:00
icu-libs := $(android)/external/icu4c/lib/libicui18n.a \
$(android)/external/icu4c/lib/libicuuc.a \
$(android)/external/icu4c/lib/libicudata.a
first step towards supporting Android class library on Windows A majority of the tests are passing, but more work is needed to get them all to pass.
2013-03-14 21:33:05 +00:00
endif
Fixed Android build on Windows
2014-07-29 09:01:51 +00:00
luni-cpps := $(filter-out $(blacklist),$(luni-cpps))
first step towards supporting Android class library on Windows A majority of the tests are passing, but more work is needed to get them all to pass.
2013-03-14 21:33:05 +00:00
various fixes to get avian-swt-examples working with the Android class library
2013-02-26 23:24:02 +00:00
classpath-lflags := \
first step towards supporting Android class library on Windows A majority of the tests are passing, but more work is needed to get them all to pass.
2013-03-14 21:33:05 +00:00
$(icu-libs) \
use Android directory layout when accessing class library code Per https://github.com/ReadyTalk/avian/issues/53, Avian should build against a standard AOSP checkout, which means we should look for subprojects in the directories the repo utility would place them.
2013-04-22 23:10:23 +00:00
$(android)/external/fdlibm/libfdm.a \
$(android)/external/expat/.libs/libexpat.a \
add Android class library instructions to README.md
2013-02-27 17:03:17 +00:00
$(android)/openssl-upstream/libssl.a \
$(android)/openssl-upstream/libcrypto.a \
first step towards supporting Android class library on Windows A majority of the tests are passing, but more work is needed to get them all to pass.
2013-03-14 21:33:05 +00:00
$(platform-lflags) \
progress towards Android classpath support It now builds and links, but fails at runtime because register_libcore_icu_ICU can't find the file it wants. We'll probably need to replace register_libcore_icu_ICU with a better-behaved version.
2013-02-19 23:48:33 +00:00
-lstdc++
first step towards supporting Android class library on Windows A majority of the tests are passing, but more work is needed to get them all to pass.
2013-03-14 21:33:05 +00:00
prevent openjdk-src build from trying to load the installed 'JavaRuntimeSupport' (fixes #128)
2013-12-08 01:49:23 +00:00
ifeq ($(platform),linux)
a couple of tweaks to get Mac/Android tests passing
2014-09-02 16:14:04 +00:00
android-cflags += -DHAVE_OFF64_T
prevent openjdk-src build from trying to load the installed 'JavaRuntimeSupport' (fixes #128)
2013-12-08 01:49:23 +00:00
classpath-lflags += -lrt
endif
only use librt on Linux; OS X doesn't have or need it
2013-07-07 17:35:20 +00:00
first pass at Android classpath support Stuff compiles, but linking breaks spectacularly. Next step is to figure out how to build the dependencies without checking out and building the entire Android platform.
2013-02-19 16:36:19 +00:00
classpath-objects = \
fix Android classpath build Hopefully it will stay fixed this time, since we now specify commit hashes for each dependency which are known to work.
2013-07-05 20:34:16 +00:00
$(call cpp-objects,$(luni-cpps),$(luni-native),$(build)) \
$(call cpp-objects,$(crypto-cpps),$(crypto-native),$(build)) \
update to more recent version of Android class library Lots has changed since we forked Android's libcore, so merging the latest upstream code has required extensive changes to the Avian/Android port. One big change is that we now use Avian's versions of java.lang.Object, java.lang.Class, java.lang.ClassLoader, some java.lang.reflect.* classes, etc. instead of the Android versions. The main reason is that the Android versions have become very Dex/Dalvik-specific, and since Avian is based on Java class files, not dex archives, that code doesn't make sense here. This has the side benefit that we can share more native code with classpath-avian.cpp and reduce the amount of Java/C++ code duplication.
2014-08-21 19:09:42 +00:00
$(call cpp-objects,$(libnativehelper-cpps),$(libnativehelper-native),$(build)) \
$(call cc-objects,$(libziparchive-ccs),$(libziparchive-native),$(build)) \
$(call cpp-objects,$(libutils-cpps),$(libutils-native),$(build))
first pass at Android classpath support Stuff compiles, but linking breaks spectacularly. Next step is to figure out how to build the dependencies without checking out and building the entire Android platform.
2013-02-19 16:36:19 +00:00
luni-java = $(android)/libcore/luni/src/main/java
update to more recent version of Android class library Lots has changed since we forked Android's libcore, so merging the latest upstream code has required extensive changes to the Avian/Android port. One big change is that we now use Avian's versions of java.lang.Object, java.lang.Class, java.lang.ClassLoader, some java.lang.reflect.* classes, etc. instead of the Android versions. The main reason is that the Android versions have become very Dex/Dalvik-specific, and since Avian is based on Java class files, not dex archives, that code doesn't make sense here. This has the side benefit that we can share more native code with classpath-avian.cpp and reduce the amount of Java/C++ code duplication.
2014-08-21 19:09:42 +00:00
luni-blacklist = \
$(luni-java)/libcore/reflect/AnnotationAccess.java
luni-javas := \
$(filter-out $(luni-blacklist),$(shell find $(luni-java) -name '*.java'))
Fixed .properties publishing and patching
2014-03-30 19:41:26 +00:00
luni-nonjavas := $(shell find $(luni-java) -not -type d -not -name '*.java')
luni-copied-nonjavas = $(call noop-files,$(luni-nonjavas),$(luni-java),)
update to more recent version of Android class library Lots has changed since we forked Android's libcore, so merging the latest upstream code has required extensive changes to the Avian/Android port. One big change is that we now use Avian's versions of java.lang.Object, java.lang.Class, java.lang.ClassLoader, some java.lang.reflect.* classes, etc. instead of the Android versions. The main reason is that the Android versions have become very Dex/Dalvik-specific, and since Avian is based on Java class files, not dex archives, that code doesn't make sense here. This has the side benefit that we can share more native code with classpath-avian.cpp and reduce the amount of Java/C++ code duplication.
2014-08-21 19:09:42 +00:00
fix SSL stack and Android JAR resource loading
2014-08-22 13:05:58 +00:00
crypto-java = $(android)/external/conscrypt/src/main/java
fix Android classpath build Hopefully it will stay fixed this time, since we now specify commit hashes for each dependency which are known to work.
2013-07-05 20:34:16 +00:00
crypto-javas := $(shell find $(crypto-java) -name '*.java')
fix SSL stack and Android JAR resource loading
2014-08-22 13:05:58 +00:00
crypto-platform-java = $(android)/external/conscrypt/src/platform/java
crypto-platform-javas := $(shell find $(crypto-platform-java) -name '*.java')
first pass at Android classpath support Stuff compiles, but linking breaks spectacularly. Next step is to figure out how to build the dependencies without checking out and building the entire Android platform.
2013-02-19 16:36:19 +00:00
dalvik-java = $(android)/libcore/dalvik/src/main/java
update to more recent version of Android class library Lots has changed since we forked Android's libcore, so merging the latest upstream code has required extensive changes to the Avian/Android port. One big change is that we now use Avian's versions of java.lang.Object, java.lang.Class, java.lang.ClassLoader, some java.lang.reflect.* classes, etc. instead of the Android versions. The main reason is that the Android versions have become very Dex/Dalvik-specific, and since Avian is based on Java class files, not dex archives, that code doesn't make sense here. This has the side benefit that we can share more native code with classpath-avian.cpp and reduce the amount of Java/C++ code duplication.
2014-08-21 19:09:42 +00:00
dalvik-javas := \
$(dalvik-java)/dalvik/system/DalvikLogHandler.java \
$(dalvik-java)/dalvik/system/CloseGuard.java \
$(dalvik-java)/dalvik/system/VMDebug.java \
$(dalvik-java)/dalvik/system/BlockGuard.java \
$(dalvik-java)/dalvik/system/SocketTagger.java \
$(dalvik-java)/dalvik/system/DalvikLogging.java \
libart-java = $(android)/libcore/libart/src/main/java
libart-javas := \
$(libart-java)/dalvik/system/VMRuntime.java \
$(libart-java)/dalvik/system/VMStack.java \
$(libart-java)/java/lang/Thread.java \
$(libart-java)/java/lang/ThreadGroup.java \
$(libart-java)/java/lang/Enum.java \
$(libart-java)/java/lang/String.java \
$(libart-java)/java/lang/ref/Reference.java \
$(libart-java)/java/lang/reflect/AccessibleObject.java \
first pass at Android classpath support Stuff compiles, but linking breaks spectacularly. Next step is to figure out how to build the dependencies without checking out and building the entire Android platform.
2013-02-19 16:36:19 +00:00
xml-java = $(android)/libcore/xml/src/main/java
xml-javas := $(shell find $(xml-java) -name '*.java')
Makefile improvement to support okhttp, and Bouncy Castle for Android version
2014-11-23 09:55:09 +00:00
okhttp-android-java = $(android)/external/okhttp/android/main/java
okhttp-android-javas := $(shell find $(okhttp-android-java) -name '*.java')
okhttp-java = $(android)/external/okhttp/okhttp/src/main/java
okhttp-javas := $(shell find $(okhttp-java) -name '*.java')
okio-java = $(android)/external/okhttp/okio/src/main/java
okio-javas := $(shell find $(okio-java) -name '*.java')
fix Cygwin build 80ce92a introduced a regression which caused the Cygwin build to fail, since it tried to set platform=$(build-platform) when calling make recursively, and platform=cygwin is rejected by the makefile. Instead, we need to use $(bootimage-platform), which normalizes "cygwin" and "mingw32" to "windows".
2015-01-22 16:57:15 +00:00
Makefile improvement to support okhttp, and Bouncy Castle for Android version
2014-11-23 09:55:09 +00:00
bcpkix-java = $(android)/external/bouncycastle/bcpkix/src/main/java
bcpkix-javas := $(shell find $(bcpkix-java) -name '*.java')
bcprov-java = $(android)/external/bouncycastle/bcprov/src/main/java
bcprov-javas := $(shell find $(bcprov-java) -name '*.java')
fix Cygwin build 80ce92a introduced a regression which caused the Cygwin build to fail, since it tried to set platform=$(build-platform) when calling make recursively, and platform=cygwin is rejected by the makefile. Instead, we need to use $(bootimage-platform), which normalizes "cygwin" and "mingw32" to "windows".
2015-01-22 16:57:15 +00:00
first pass at Android classpath support Stuff compiles, but linking breaks spectacularly. Next step is to figure out how to build the dependencies without checking out and building the entire Android platform.
2013-02-19 16:36:19 +00:00
android-classes = \
$(call java-classes,$(luni-javas),$(luni-java),$(build)/android) \
fix Android classpath build Hopefully it will stay fixed this time, since we now specify commit hashes for each dependency which are known to work.
2013-07-05 20:34:16 +00:00
$(call java-classes,$(crypto-javas),$(crypto-java),$(build)/android) \
fix SSL stack and Android JAR resource loading
2014-08-22 13:05:58 +00:00
$(call java-classes,$(crypto-platform-javas),$(crypto-platform-java),$(build)/android) \
first pass at Android classpath support Stuff compiles, but linking breaks spectacularly. Next step is to figure out how to build the dependencies without checking out and building the entire Android platform.
2013-02-19 16:36:19 +00:00
$(call java-classes,$(dalvik-javas),$(dalvik-java),$(build)/android) \
update to more recent version of Android class library Lots has changed since we forked Android's libcore, so merging the latest upstream code has required extensive changes to the Avian/Android port. One big change is that we now use Avian's versions of java.lang.Object, java.lang.Class, java.lang.ClassLoader, some java.lang.reflect.* classes, etc. instead of the Android versions. The main reason is that the Android versions have become very Dex/Dalvik-specific, and since Avian is based on Java class files, not dex archives, that code doesn't make sense here. This has the side benefit that we can share more native code with classpath-avian.cpp and reduce the amount of Java/C++ code duplication.
2014-08-21 19:09:42 +00:00
$(call java-classes,$(libart-javas),$(libart-java),$(build)/android) \
Makefile improvement to support okhttp, and Bouncy Castle for Android version
2014-11-23 09:55:09 +00:00
$(call java-classes,$(xml-javas),$(xml-java),$(build)/android) \
$(call java-classes,$(okhttp-javas),$(okhttp-java),$(build)/android) \
$(call java-classes,$(okhttp-android-javas),$(okhttp-android-java),$(build)/android) \
$(call java-classes,$(okio-javas),$(okio-java),$(build)/android) \
$(call java-classes,$(bcpkix-javas),$(bcpkix-java),$(build)/android) \
$(call java-classes,$(bcprov-javas),$(bcprov-java),$(build)/android)
first pass at Android classpath support Stuff compiles, but linking breaks spectacularly. Next step is to figure out how to build the dependencies without checking out and building the entire Android platform.
2013-02-19 16:36:19 +00:00
classpath = android
all tests now pass for Android class library build
2013-02-23 00:23:59 +00:00
javahome-files = tzdata
javahome-object = $(build)/javahome-jar.o
boot-javahome-object = $(build)/boot-javahome.o
build-javahome = $(android)/bionic/libc/zoneinfo
first step towards supporting Android class library on Windows A majority of the tests are passing, but more work is needed to get them all to pass.
2013-03-14 21:33:05 +00:00
stub-sources = $(src)/android/stubs.cpp
stub-objects = $(call cpp-objects,$(stub-sources),$(src),$(build))
first pass at Android classpath support Stuff compiles, but linking breaks spectacularly. Next step is to figure out how to build the dependencies without checking out and building the entire Android platform.
2013-02-19 16:36:19 +00:00
endif
enable standalone OpenJDK builds As described in readme.txt, a standalone OpenJDK build embeds all libraries, classes, and other files needed at runtime in the resulting binary, eliminating dependencies on external resources.
2010-11-04 17:02:09 +00:00
ifeq ($(classpath),avian)
preliminary support for using OpenJDK's class library Whereas the GNU Classpath port used the strategy of patching Classpath with core classes from Avian so as to minimize changes to the VM, this port uses the opposite strategy: abstract and isolate classpath-specific features in the VM similar to how we abstract away platform-specific features in system.h. This allows us to use an unmodified copy of OpenJDK's class library, including its core classes and augmented by a few VM-specific classes in the "avian" package.
2010-09-10 21:05:29 +00:00
jni-sources := $(shell find $(classpath-src) -name '*.cpp')
build everything (including classes) in options-specific directory Previously, we only built platform specific code in an options-specific directory (e.g. build/linux-x86_64-debug), but built the Java classes in the shared, top-level build directory. This minimized duplication when building with different sets of options, but now that we're supporting the option of selecting a third-party class library, we need to put the classes in the options-specific directory as well.
2010-09-17 22:03:37 +00:00
jni-objects = $(call cpp-objects,$(jni-sources),$(classpath-src),$(build))
enable standalone OpenJDK builds As described in readme.txt, a standalone OpenJDK build embeds all libraries, classes, and other files needed at runtime in the resulting binary, eliminating dependencies on external resources.
2010-11-04 17:02:09 +00:00
classpath-objects = $(jni-objects)
fix rebuild when using GNU Classpath
2009-06-04 23:54:06 +00:00
endif
intercept SIGSEGV and throw NullPointerExceptions
2007-12-31 22:40:56 +00:00
input = List
support classloading from jar files compiled into the executable
2007-10-25 22:06:05 +00:00
add use-clang option for building with LLVM Clang instead of GCC This also fixes several errors and warnings emitted by Clang.
2012-06-01 23:43:42 +00:00
ifeq ($(use-clang),true)
Bugfixes for Mac OS X, etc.
2013-10-27 14:36:32 +00:00
build-cxx = clang++ -std=c++11
add use-clang option for building with LLVM Clang instead of GCC This also fixes several errors and warnings emitted by Clang.
2012-06-01 23:43:42 +00:00
build-cc = clang
else
revert accidentally-committed build-cc/build-cxx change
2012-08-02 16:40:49 +00:00
build-cxx = g++
build-cc = gcc
add use-clang option for building with LLVM Clang instead of GCC This also fixes several errors and warnings emitted by Clang.
2012-06-01 23:43:42 +00:00
endif
more GC stress fixes
2007-07-17 01:55:49 +00:00
fixed linux build because darwin fix broke it
2010-02-06 00:15:12 +00:00
mflag =
rename platform=darwin -> platform={ios,macosx}, fix ios simulator build
2014-06-22 18:30:52 +00:00
ifneq ($(kernel),darwin)
preliminary support for using OpenJDK's class library Whereas the GNU Classpath port used the strategy of patching Classpath with core classes from Avian so as to minimize changes to the VM, this port uses the opposite strategy: abstract and isolate classpath-specific features in the VM similar to how we abstract away platform-specific features in system.h. This allows us to use an unmodified copy of OpenJDK's class library, including its core classes and augmented by a few VM-specific classes in the "avian" package.
2010-09-10 21:05:29 +00:00
ifeq ($(arch),i386)
mflag = -m32
endif
ifeq ($(arch),x86_64)
mflag = -m64
endif
Updated to always specify bitness of target. cc and cxx will always have -m32 or -m64 appended based on the arch that is being built.
2010-02-04 18:15:16 +00:00
endif
Fixed darwin build.
2010-02-05 23:40:48 +00:00
clean up binary format code Linux, FreeBSD, and QNX all use ELF, so no need to distinguish between them when generating object files. To avoid confusion, I've switch from using operating system names to using binary format names where applicable.
2012-08-02 18:15:15 +00:00
target-format = elf
fixed linux build because darwin fix broke it
2010-02-06 00:15:12 +00:00
cxx = $(build-cxx) $(mflag)
cc = $(build-cc) $(mflag)
build standalone executable (no seperate .so) on all platforms; build libvm.a too
2007-10-23 23:23:55 +00:00
ar = ar
ranlib = ranlib
Merge branch 'master' into powerpc Conflicts: makefile src/assembler.h src/compile.cpp src/compiler.cpp src/compiler.h src/finder.cpp
2008-11-11 15:20:49 +00:00
dlltool = dlltool
whitespace tweak
2008-01-04 02:17:42 +00:00
vg = nice valgrind --num-callers=32 --db-attach=yes --freelist-vol=100000000
update valgrind.supp to suppress ld-2.6.so warnings
2008-05-19 04:24:12 +00:00
vg += --leak-check=full --suppressions=valgrind.supp
various performance tweaks and bugfixes
2007-08-19 19:45:51 +00:00
db = gdb --args
specify UTF-8 encoding to javac This fixes "illegal character" errors when compiling e.g. Misc.java on systems where the default encoding is not UTF-8.
2013-02-06 17:04:55 +00:00
javac = "$(JAVA_HOME)/bin/javac" -encoding UTF-8
enable standalone OpenJDK builds As described in readme.txt, a standalone OpenJDK build embeds all libraries, classes, and other files needed at runtime in the resulting binary, eliminating dependencies on external resources.
2010-11-04 17:02:09 +00:00
javah = "$(JAVA_HOME)/bin/javah"
Merge branch 'master' into powerpc Conflicts: makefile src/assembler.h src/compile.cpp src/compiler.cpp src/compiler.h src/finder.cpp
2008-11-11 15:20:49 +00:00
jar = "$(JAVA_HOME)/bin/jar"
strip binaries if and only if optimizing
2009-08-24 22:38:00 +00:00
strip = strip
Factored JNIEXPORT into one place. Adopted the visbility attribute that should allow strip to work on Mac OS X
2007-11-01 20:00:22 +00:00
strip-all = --strip-all
copy makefile, stdc++.cpp, and type-generator.cpp from mess project and began customizing them for this project
2007-06-03 23:16:07 +00:00
refactor build system to support cross-compiling
2007-10-22 20:56:27 +00:00
rdynamic = -rdynamic
Windows Phone 8 / Windows RT initial support Conflicts: makefile
2013-01-28 15:20:52 +00:00
cflags_debug = -O0 -g3
cflags_debug_fast = -O0 -g3
cflags_stress = -O0 -g3
cflags_stress_major = -O0 -g3
ifeq ($(use-clang),true)
only enable LTO if use-lto is true When using clang, as with GCC, we shouldn't use LTO if the developer has specified use-lto=false.
2013-12-18 21:53:46 +00:00
cflags_fast = -O3 -g3
Windows Phone 8 / Windows RT initial support Conflicts: makefile
2013-01-28 15:20:52 +00:00
cflags_small = -Oz -g3
else
cflags_fast = -O3 -g3
cflags_small = -Os -g3
endif
refine build directory naming system to avoid needing to 'make clean' upon changing build parameters
2009-03-14 21:16:05 +00:00
# note that we suppress the non-virtual-dtor warning because we never
Merge branch 'master' into powerpc Conflicts: makefile src/assembler.h src/compile.cpp src/compiler.cpp src/compiler.h src/finder.cpp
2008-11-11 15:20:49 +00:00
# use the delete operator, which means we don't need virtual
# destructors:
warnings = -Wall -Wextra -Werror -Wunused-parameter -Winit-self \
-Wno-non-virtual-dtor
copy makefile, stdc++.cpp, and type-generator.cpp from mess project and began customizing them for this project
2007-06-03 23:16:07 +00:00
more progress towards cross-architecture bootimage builds This commit fixes a lot of bugs. All tests are now pass for Linux x86_64 to Linux i386 cross builds.
2011-09-01 03:18:00 +00:00
target-cflags = -DTARGET_BYTES_PER_WORD=$(pointer-size)
use c++11 enhanced-for and auto in type-generator
2014-07-12 05:15:09 +00:00
common-cflags = $(warnings) -std=c++0x -fno-rtti -fno-exceptions -I$(classpath-src) \
use -I$(src) instead of -idirafter $(src) This fixes the iOS build, where the wrong version of util.h was being used. The change to use -idirafter happened way back in 2007, and I can't remember what its intention was. We'll have to watch for regressions on other platforms.
2013-02-27 19:02:59 +00:00
"-I$(JAVA_HOME)/include" -I$(src) -I$(build) -Iinclude $(classpath-cflags) \
support building Avian as a self-contained dynamic library
2008-03-31 03:43:43 +00:00
-D__STDC_LIMIT_MACROS -D_JNI_IMPLEMENTATION_ -DAVIAN_VERSION=\"$(version)\" \
Set java.vm.info based on makefile info= and set java.vm.version based on makefile version= in order to display relevant OpenJDK -version information. Signed-off-by: Matthias Klose <doko@ubuntu.com> Signed-off-by: Xerxes Rånby <xerxes@zafena.se>
2012-09-05 12:31:30 +00:00
-DAVIAN_INFO="\"$(info)\"" \
fix native Windows GCC 3.4 OpenJDK build
2010-11-13 01:29:12 +00:00
-DUSE_ATOMIC_OPERATIONS -DAVIAN_JAVA_HOME=\"$(javahome)\" \
more progress towards cross-architecture bootimage builds This commit fixes a lot of bugs. All tests are now pass for Linux x86_64 to Linux i386 cross builds.
2011-09-01 03:18:00 +00:00
-DAVIAN_EMBED_PREFIX=\"$(embed-prefix)\" $(target-cflags)
build standalone executable (no seperate .so) on all platforms; build libvm.a too
2007-10-23 23:23:55 +00:00
Windows Phone 8 / Windows RT initial support Conflicts: makefile
2013-01-28 15:20:52 +00:00
asmflags = $(target-cflags) -I$(src)
abort on startup if Thread field offset constants don't match the real runtime values
2012-06-20 19:14:16 +00:00
only enable frame pointer on x86(_64)
2011-01-29 00:05:42 +00:00
ifneq (,$(filter i386 x86_64,$(arch)))
ifeq ($(use-frame-pointer),true)
common-cflags += -fno-omit-frame-pointer -DAVIAN_USE_FRAME_POINTER
asmflags += -DAVIAN_USE_FRAME_POINTER
endif
enable use-frame-pointer=true build Also, include Continuations, Coroutines, and DynamicWind tests in test suite for continuations=true build.
2011-01-28 04:06:01 +00:00
endif
makefile fixes
2007-10-26 14:34:54 +00:00
build-cflags = $(common-cflags) -fPIC -fvisibility=hidden \
Merge branch 'master' into powerpc Conflicts: makefile src/assembler.h src/compile.cpp src/compiler.cpp src/compiler.h src/finder.cpp
2008-11-11 15:20:49 +00:00
"-I$(JAVA_HOME)/include/linux" -I$(src) -pthread
makefile fixes
2007-10-26 14:34:54 +00:00
use c++11 enhanced-for and auto in type-generator
2014-07-12 05:15:09 +00:00
converter-cflags = -D__STDC_CONSTANT_MACROS -std=c++0x -Iinclude/ -Isrc/ \
directly emit codeimage as a object (binaryToObject is statically linked in), as a stepping stone to including extra symbols in said codeimage
2012-05-01 19:16:32 +00:00
-fno-rtti -fno-exceptions \
-DAVIAN_TARGET_ARCH=AVIAN_ARCH_UNKNOWN \
clean up binary format code Linux, FreeBSD, and QNX all use ELF, so no need to distinguish between them when generating object files. To avoid confusion, I've switch from using operating system names to using binary format names where applicable.
2012-08-02 18:15:15 +00:00
-DAVIAN_TARGET_FORMAT=AVIAN_FORMAT_UNKNOWN \
fix OS X build regressions
2012-06-03 00:56:07 +00:00
-Wall -Wextra -Werror -Wunused-parameter -Winit-self -Wno-non-virtual-dtor
fix OS X 10.4 PowerPC cross-build
2010-11-23 22:54:35 +00:00
makefile fixes
2007-10-26 14:34:54 +00:00
cflags = $(build-cflags)
various tweaks to get hello-ios working with the Android library
2013-03-04 21:24:41 +00:00
common-lflags = -lm -lz
build fixes and refactoring
2007-10-26 21:02:39 +00:00
Bugfixes for Mac OS X, etc.
2013-10-27 14:36:32 +00:00
ifeq ($(use-clang),true)
add process=interpret support for Linux/ARM64 This makes all the tests pass for the platform=linux arch=arm64 process=interpret build. Next step: process=compile support.
2014-11-26 01:36:34 +00:00
ifeq ($(build-kernel),darwin)
common-lflags += -Wl,-export_dynamic
else
ifneq ($(platform),windows)
common-lflags += -Wl,-E
else
common-lflags += -Wl,--export-all-symbols
endif
endif
Bugfixes for Mac OS X, etc.
2013-10-27 14:36:32 +00:00
endif
rework OpenJDK build to derive classpath and library path from environment We now consult the JAVA_HOME environment variable to determine where to find the system library JARs and SOs. Ultimately, we'll want to support self-contained build, but this allows Avian to behave like a conventional libjvm.so.
2010-09-20 23:31:23 +00:00
build-lflags = -lz -lpthread -ldl
Merge branch 'master' into powerpc Conflicts: makefile src/assembler.h src/compile.cpp src/compiler.cpp src/compiler.h src/finder.cpp
2008-11-11 15:20:49 +00:00
various tweaks to get hello-ios working with the Android library
2013-03-04 21:24:41 +00:00
lflags = $(common-lflags) -lpthread -ldl
makefile fixes
2007-10-26 14:34:54 +00:00
initial support for Windows OpenJDK build All the tests are passing for openjdk-src builds, but the non-src openjdk build is crashing and there's trouble loading time zone info from the embedded java.home directory.
2010-11-07 04:21:19 +00:00
build-system = posix
refactor build system to support cross-compiling
2007-10-22 20:56:27 +00:00
system = posix
sketch of windows port
2007-10-23 01:00:57 +00:00
asm = x86
build fixes and refactoring
2007-10-26 21:02:39 +00:00
pointer-size = 8
support building Avian as a self-contained dynamic library
2008-03-31 03:43:43 +00:00
so-prefix = lib
so-suffix = .so
Windows Phone 8 / Windows RT initial support Conflicts: makefile
2013-01-28 15:20:52 +00:00
static-prefix = lib
static-suffix = .a
output = -o $(1)
asm-output = -o $(1)
asm-input = -c $(1)
asm-format = S
as = $(cc)
add remote-test capability To execute tests on a remote host (for instance, because you're cross-compiling), simply do: make remote-test=true remote-test-host=<host_to_test_on> test You can set several variables to control the functionality of remote-test. See them below, along with their default values: remote-test-host = localhost # host to ssh to remote-test-port = 22 remote-test-user = ${USER} # user to execute tests as remote-test-dir = /tmp/avian-test-${USER} # dir to rsync build output to
2013-02-15 00:21:57 +00:00
ld = $(cc)
Windows Phone 8 / Windows RT initial support Conflicts: makefile
2013-01-28 15:20:52 +00:00
build-ld = $(build-cc)
generate better typed accessors from type-generator (and switch to using standard library there)
2014-06-21 04:16:33 +00:00
build-ld-cpp = $(build-cxx)
Windows Phone 8 / Windows RT initial support Conflicts: makefile
2013-01-28 15:20:52 +00:00
set remote-test variable based on the presence of remote-test-host or remote-test-port The new way to run a remote test is: make arch=<arch> remote-test-host=<host_to_test_on> test
2013-02-16 00:43:21 +00:00
default-remote-test-host = localhost
default-remote-test-port = 22
ifeq ($(remote-test-host),)
remote-test-host = $(default-remote-test-host)
else
remote-test = true
endif
ifeq ($(remote-test-port),)
remote-test-port = $(default-remote-test-port)
else
remote-test = true
endif
add remote-test capability To execute tests on a remote host (for instance, because you're cross-compiling), simply do: make remote-test=true remote-test-host=<host_to_test_on> test You can set several variables to control the functionality of remote-test. See them below, along with their default values: remote-test-host = localhost # host to ssh to remote-test-port = 22 remote-test-user = ${USER} # user to execute tests as remote-test-dir = /tmp/avian-test-${USER} # dir to rsync build output to
2013-02-15 00:21:57 +00:00
remote-test-user = ${USER}
remote-test-dir = /tmp/avian-test-${USER}
Windows Phone 8 / Windows RT initial support Conflicts: makefile
2013-01-28 15:20:52 +00:00
static = -static
support building Avian as a self-contained dynamic library
2008-03-31 03:43:43 +00:00
shared = -shared
don't specify -Wl,-rpath argument on Darwin or Windows (take 2)
2013-02-12 16:27:47 +00:00
rpath = -Wl,-rpath=\$$ORIGIN -Wl,-z,origin
don't specify -Wl,-rpath argument on Darwin or Windows It remains to be seen what form that argument should take on those operating systems; for now, it only causes trouble, so we'll just leave it out.
2013-02-12 15:22:21 +00:00
fix Cygwin/MSYS openjdk-src builds
2010-11-15 23:27:00 +00:00
openjdk-extra-cflags = -fvisibility=hidden
more progress towards cross-architecture bootimage builds This commit fixes a lot of bugs. All tests are now pass for Linux x86_64 to Linux i386 cross builds.
2011-09-01 03:18:00 +00:00
bootimage-cflags = -DTARGET_BYTES_PER_WORD=$(pointer-size)
Windows Phone 8 / Windows RT initial support Conflicts: makefile
2013-01-28 15:20:52 +00:00
bootimage-symbols = _binary_bootimage_bin_start:_binary_bootimage_bin_end
codeimage-symbols = _binary_codeimage_bin_start:_binary_codeimage_bin_end
more progress towards cross-architecture bootimage builds This commit fixes a lot of bugs. All tests are now pass for Linux x86_64 to Linux i386 cross builds.
2011-09-01 03:18:00 +00:00
fix sdk detection logic
2013-09-04 03:03:29 +00:00
developer-dir := $(shell if test -d /Developer/Platforms/$(target).platform/Developer/SDKs; then echo /Developer; \
add support for Xcode 4.3
2012-03-07 15:55:01 +00:00
else echo /Applications/Xcode.app/Contents/Developer; fi)
add process=interpret support for Linux/ARM64 This makes all the tests pass for the platform=linux arch=arm64 process=interpret build. Next step: process=compile support.
2014-11-26 01:36:34 +00:00
ifneq (,$(filter i386 arm,$(arch)))
add back pointer-size assignment in makefile for i386 (accidentally removed)
2012-05-02 20:52:19 +00:00
pointer-size = 4
endif
add process=interpret support for Linux/ARM64 This makes all the tests pass for the platform=linux arch=arm64 process=interpret build. Next step: process=compile support.
2014-11-26 01:36:34 +00:00
ifneq (,$(filter arm arm64,$(arch)))
enable ARM cross builds
2010-11-09 03:48:08 +00:00
asm = arm
directly emit codeimage as a object (binaryToObject is statically linked in), as a stepping stone to including extra symbols in said codeimage
2012-05-01 19:16:32 +00:00
rename platform=darwin -> platform={ios,macosx}, fix ios simulator build
2014-06-22 18:30:52 +00:00
ifneq ($(platform),ios)
add process=interpret support for Linux/ARM64 This makes all the tests pass for the platform=linux arch=arm64 process=interpret build. Next step: process=compile support.
2014-11-26 01:36:34 +00:00
ifneq ($(arch),arm64)
no-psabi = -Wno-psabi
cflags += -marm $(no-psabi)
rename platform=darwin -> platform={ios,macosx}, fix ios simulator build
2014-06-22 18:30:52 +00:00
add process=interpret support for Linux/ARM64 This makes all the tests pass for the platform=linux arch=arm64 process=interpret build. Next step: process=compile support.
2014-11-26 01:36:34 +00:00
# By default, assume we can't use armv7-specific instructions on
# non-iOS platforms. Ideally, we'd detect this at runtime.
armv6=true
endif
add support for Apple iOS on ARM Most tests are passing, but not all, so this still needs more work.
2011-08-11 01:21:48 +00:00
endif
ARM makefile tweaks
2010-11-09 18:28:58 +00:00
ifneq ($(arch),$(build-arch))
rename platform=darwin -> platform={ios,macosx}, fix ios simulator build
2014-06-22 18:30:52 +00:00
ifneq ($(kernel),darwin)
add process=interpret support for Linux/ARM64 This makes all the tests pass for the platform=linux arch=arm64 process=interpret build. Next step: process=compile support.
2014-11-26 01:36:34 +00:00
ifeq ($(arch),arm64)
cxx = aarch64-linux-gnu-g++
cc = aarch64-linux-gnu-gcc
ar = aarch64-linux-gnu-ar
ranlib = aarch64-linux-gnu-ranlib
strip = aarch64-linux-gnu-strip
else
cxx = arm-linux-gnueabi-g++
cc = arm-linux-gnueabi-gcc
ar = arm-linux-gnueabi-ar
ranlib = arm-linux-gnueabi-ranlib
strip = arm-linux-gnueabi-strip
endif
add support for Apple iOS on ARM Most tests are passing, but not all, so this still needs more work.
2011-08-11 01:21:48 +00:00
endif
ARM makefile tweaks
2010-11-09 18:28:58 +00:00
endif
added ARM modifications to makefile
2009-08-06 19:25:20 +00:00
endif
refactor build system to support cross-compiling
2007-10-22 20:56:27 +00:00
use armv7 memory barriers by default armv7 and later provide weaker cache coherency models than armv6 and earlier, so we cannot just implement memory barriers as no-ops. This patch uses the DMB instruction (or the equivalent OS-provided barrier function) to implement barriers. This should fix concurrency issues on newer chips such as the Apple A6 and A7. If you still need to support ARMv6 devices, you should pass "armv6=true" to make when building Avian. Ideally, the VM would detect what kind of CPU it was executing on at runtime and direct the JIT compiler accordingly, but I don't know how to do that on ARM. Patches are welcome, though!
2014-01-10 21:38:57 +00:00
ifeq ($(armv6),true)
cflags += -DAVIAN_ASSUME_ARMV6
endif
rename platform=darwin -> platform={ios,macosx}, fix ios simulator build
2014-06-22 18:30:52 +00:00
ifeq ($(platform),ios)
add __attribute__ ((used)) to EXPORT macro definition This ensures that exported functions are not stripped when using the -dead_strip option to the Apple linker.
2011-09-22 22:55:45 +00:00
cflags += -DAVIAN_IOS
only enable LTO if use-lto is true When using clang, as with GCC, we shouldn't use LTO if the developer has specified use-lto=false.
2013-12-18 21:53:46 +00:00
use-lto = false
add __attribute__ ((used)) to EXPORT macro definition This ensures that exported functions are not stripped when using the -dead_strip option to the Apple linker.
2011-09-22 22:55:45 +00:00
endif
Merge branch 'ios' of oss:/var/local/git/avian into ios Conflicts: makefile
2011-08-30 01:07:42 +00:00
rename platform=darwin -> platform={ios,macosx}, fix ios simulator build
2014-06-22 18:30:52 +00:00
ifeq ($(build-kernel),darwin)
support cross builds on OS X to non-Mac targets
2011-08-23 23:04:04 +00:00
build-cflags = $(common-cflags) -fPIC -fvisibility=hidden -I$(src)
fix jni_md.h not found 10.9 problem
2013-11-05 00:26:57 +00:00
cflags += -Wno-deprecated-declarations
support cross builds on OS X to non-Mac targets
2011-08-23 23:04:04 +00:00
build-lflags += -framework CoreFoundation
fix linux->darwin cross-compile There were two issues: the linux->darwin cross compiler is more stringent about unused variables, and the makefile specified flags for building ON darwin that were actually applicable whenever we are building FOR darwin.
2013-06-04 19:32:22 +00:00
endif
add support for QNX on x86_32 All but one test is passing. The failure is due to the fact that QNX doesn't (in general) support calling fork(2) from a multithreaded process. Thus, we'll need to use spawn instead of fork/exec on QNX, which I'll attempt in a later commit. http://www.qnx.com/developers/docs/6.4.1/neutrino/getting_started/s1_procs.html
2012-07-18 00:21:17 +00:00
ifeq ($(platform),qnx)
cflags = $(common-cflags) -fPIC -fvisibility=hidden -I$(src)
lflags = $(common-lflags) -lsocket
ifeq ($(build-platform),qnx)
build-cflags = $(common-cflags) -fPIC -fvisibility=hidden -I$(src)
build-lflags = $(common-lflags)
else
add QNX ARM support This is untested so far, since I haven't figured out how to install the ARM port of QNX on QEMU.
2012-07-18 01:38:39 +00:00
ifeq ($(arch),i386)
prefix = i486-pc-nto-qnx6.5.0-
else
prefix = arm-unknown-nto-qnx6.5.0-
endif
add support for QNX on x86_32 All but one test is passing. The failure is due to the fact that QNX doesn't (in general) support calling fork(2) from a multithreaded process. Thus, we'll need to use spawn instead of fork/exec on QNX, which I'll attempt in a later commit. http://www.qnx.com/developers/docs/6.4.1/neutrino/getting_started/s1_procs.html
2012-07-18 00:21:17 +00:00
endif
cxx = $(prefix)g++
cc = $(prefix)gcc
ar = $(prefix)ar
ranlib = $(prefix)ranlib
strip = $(prefix)strip
rdynamic = -Wl,--export-dynamic
endif
add support for FreeBSD
2012-08-02 16:36:16 +00:00
ifeq ($(platform),freebsd)
# There is no -ldl on FreeBSD
build-lflags = $(common-lflags) -lz -lpthread
lflags = $(common-lflags) -lpthread
# include/freebsd instead of include/linux
build-cflags = $(common-cflags) -fPIC -fvisibility=hidden \
"-I$(JAVA_HOME)/include/freebsd" -I$(src) -pthread
cflags = $(build-cflags)
endif
Add android platform
2013-01-23 23:55:23 +00:00
ifeq ($(platform),android)
fix linux->darwin cross-compile There were two issues: the linux->darwin cross compiler is more stringent about unused variables, and the makefile specified flags for building ON darwin that were actually applicable whenever we are building FOR darwin.
2013-06-04 19:32:22 +00:00
ifeq ($(build-platform),cygwin)
Android toolchain
2013-01-24 14:17:52 +00:00
ndk = "$$(cygpath -u "$(ANDROID_NDK)")"
else
ndk = $(ANDROID_NDK)
endif
Windows Phone 8 / Windows RT initial support Conflicts: makefile
2013-01-28 15:20:52 +00:00
ifeq ($(android-version),)
android-version = 5
endif
ifeq ($(android-toolchain),)
android-toolchain = 4.7
endif
ifeq ($(arch),arm)
android-toolchain-name = arm-linux-androideabi
android-toolchain-prefix = arm-linux-androideabi-
endif
ifeq ($(arch),i386)
android-toolchain-name = x86
android-toolchain-prefix = i686-linux-android-
endif
ifeq ($(android-arm-arch),)
android-arm-arch = armv5
endif
options := $(options)-api$(android-version)-$(android-toolchain)-$(android-arm-arch)
Android toolchain
2013-01-24 14:17:52 +00:00
build-cflags = $(common-cflags) -I$(src)
Windows Phone 8 / Windows RT initial support Conflicts: makefile
2013-01-28 15:20:52 +00:00
build-lflags = -lz -lpthread
Android toolchain
2013-01-24 14:17:52 +00:00
ifeq ($(subst cygwin,windows,$(subst mingw32,windows,$(build-platform))),windows)
toolchain-host-platform = $(subst cygwin,windows,$(subst mingw32,windows,$(build-platform)))
build-system = windows
build-cxx = i686-w64-mingw32-g++
build-cc = i686-w64-mingw32-gcc
Windows Phone 8 / Windows RT initial support Conflicts: makefile
2013-01-28 15:20:52 +00:00
sysroot = "$$(cygpath -w "$(ndk)/platforms/android-$(android-version)/arch-arm")"
Android toolchain
2013-01-24 14:17:52 +00:00
build-cflags += "-I$(JAVA_HOME)/include/win32"
else
Path fix
2013-01-24 14:39:14 +00:00
toolchain-host-platform = $(subst cygwin,windows,$(subst mingw32,windows,$(build-platform)))-*
Windows Phone 8 / Windows RT initial support Conflicts: makefile
2013-01-28 15:20:52 +00:00
sysroot = $(ndk)/platforms/android-$(android-version)/arch-arm
Android toolchain
2013-01-24 14:17:52 +00:00
build-cflags += "-I$(JAVA_HOME)/include/linux"
Windows Phone 8 / Windows RT initial support Conflicts: makefile
2013-01-28 15:20:52 +00:00
build-lflags += -ldl
Android toolchain
2013-01-24 14:17:52 +00:00
endif
Windows Phone 8 / Windows RT initial support Conflicts: makefile
2013-01-28 15:20:52 +00:00
toolchain = $(ndk)/toolchains/$(android-toolchain-name)-$(android-toolchain)/prebuilt/$(toolchain-host-platform)
cflags = "-I$(sysroot)/usr/include" "-I$(JAVA_HOME)/include/linux" $(common-cflags) "-I$(src)" -std=c++11 $(no-psabi)
lflags = "-L$(sysroot)/usr/lib" $(common-lflags) -llog
target-format = elf
Android toolchain
2013-01-24 14:17:52 +00:00
use-lto = false
specify UTF-8 encoding to javac This fixes "illegal character" errors when compiling e.g. Misc.java on systems where the default encoding is not UTF-8.
2013-02-06 17:04:55 +00:00
Windows Phone 8 / Windows RT initial support Conflicts: makefile
2013-01-28 15:20:52 +00:00
ifeq ($(arch),arm)
cflags += -marm -march=$(android-arm-arch) -ftree-vectorize -ffast-math -mfloat-abi=softfp
endif
ifeq ($(arch),i386)
endif
cxx = $(toolchain)/bin/$(android-toolchain-prefix)g++ --sysroot="$(sysroot)"
cc = $(toolchain)/bin/$(android-toolchain-prefix)gcc --sysroot="$(sysroot)"
as = $(cxx)
ar = $(toolchain)/bin/$(android-toolchain-prefix)ar
ranlib = $(toolchain)/bin/$(android-toolchain-prefix)ranlib
strip = $(toolchain)/bin/$(android-toolchain-prefix)strip
Add android platform
2013-01-23 23:55:23 +00:00
endif
add support for FreeBSD
2012-08-02 16:36:16 +00:00
rename platform=darwin -> platform={ios,macosx}, fix ios simulator build
2014-06-22 18:30:52 +00:00
ifeq ($(kernel),darwin)
clean up binary format code Linux, FreeBSD, and QNX all use ELF, so no need to distinguish between them when generating object files. To avoid confusion, I've switch from using operating system names to using binary format names where applicable.
2012-08-02 18:15:15 +00:00
target-format = macho
OSX Lion build changes Added JavaVM to include path. Also allow dynamically building with different SDK versions through the OSX_SDK_VERSION and OSX_SDK_SYSROOT environment variables (these default to the previosly hardcoded value of 10.4 and 10.4u respectively).
2011-06-02 21:15:08 +00:00
ifeq (${OSX_SDK_SYSROOT},)
rename platform=darwin -> platform={ios,macosx}, fix ios simulator build
2014-06-22 18:30:52 +00:00
OSX_SDK_SYSROOT = 10.6u
OSX Lion build changes Added JavaVM to include path. Also allow dynamically building with different SDK versions through the OSX_SDK_VERSION and OSX_SDK_SYSROOT environment variables (these default to the previosly hardcoded value of 10.4 and 10.4u respectively).
2011-06-02 21:15:08 +00:00
endif
ifeq (${OSX_SDK_VERSION},)
rename platform=darwin -> platform={ios,macosx}, fix ios simulator build
2014-06-22 18:30:52 +00:00
OSX_SDK_VERSION = 10.6
OSX Lion build changes Added JavaVM to include path. Also allow dynamically building with different SDK versions through the OSX_SDK_VERSION and OSX_SDK_SYSROOT environment variables (these default to the previosly hardcoded value of 10.4 and 10.4u respectively).
2011-06-02 21:15:08 +00:00
endif
rename platform=darwin -> platform={ios,macosx}, fix ios simulator build
2014-06-22 18:30:52 +00:00
ifneq ($(build-kernel),darwin)
enable i686-apple-darwin8 cross compile
2011-02-18 00:39:41 +00:00
cxx = i686-apple-darwin8-g++ $(mflag)
cc = i686-apple-darwin8-gcc $(mflag)
ar = i686-apple-darwin8-ar
ranlib = i686-apple-darwin8-ranlib
strip = i686-apple-darwin8-strip
OSX Lion build changes
2011-05-26 17:57:30 +00:00
sysroot = /opt/mac/SDKs/MacOSX${OSX_SDK_SYSROOT}.sdk
fix Mac OS PowerPC cross build While we can use Linux's jni.h to cross compile the i386 Mac OS build, that doesn't work for the PowerPC one. Now we use the proper /System/Library/Frameworks/JavaVM.framework/Versions/1.5.0/Headers/jni.h from the sysroot instead.
2011-02-22 22:45:46 +00:00
cflags = -I$(sysroot)/System/Library/Frameworks/JavaVM.framework/Versions/1.5.0/Headers/ \
$(common-cflags) -fPIC -fvisibility=hidden -I$(src)
enable i686-apple-darwin8 cross compile
2011-02-18 00:39:41 +00:00
endif
rename platform=darwin -> platform={ios,macosx}, fix ios simulator build
2014-06-22 18:30:52 +00:00
ifneq ($(platform),ios)
fix openjdk-src build failures on osx 10.9
2013-12-07 23:38:00 +00:00
platform-dir = $(developer-dir)/Platforms/MacOSX.platform
sdk-dir = $(platform-dir)/Developer/SDKs
mac-version := $(shell \
add process=interpret support for Linux/ARM64 This makes all the tests pass for the platform=linux arch=arm64 process=interpret build. Next step: process=compile support.
2014-11-26 01:36:34 +00:00
if test -d $(sdk-dir)/MacOSX10.9.sdk; then echo 10.9; \
fix openjdk-src build failures on osx 10.9
2013-12-07 23:38:00 +00:00
elif test -d $(sdk-dir)/MacOSX10.8.sdk; then echo 10.8; \
elif test -d $(sdk-dir)/MacOSX10.7.sdk; then echo 10.7; \
elif test -d $(sdk-dir)/MacOSX10.6.sdk; then echo 10.6; \
else echo; fi)
sysroot = $(sdk-dir)/MacOSX$(mac-version).sdk
endif
only define soname-flag and version-script-flag for openjdk builds These flags are only needed to ensure that libjvm.so can be used as a drop-in replacement for OpenJDK's VM, and that only makes sense for openjdk builds (without the openjdk-src option). It didn't hurt anything to define them unconditionally, but it was misleading.
2014-04-12 01:32:16 +00:00
soname-flag =
fix OS X build
2010-11-08 04:23:25 +00:00
version-script-flag =
Fix makefiles for building on Mavericks
2014-06-05 13:29:44 +00:00
lflags = $(common-lflags) -ldl -framework CoreFoundation -framework Foundation
fix build on old Xcode versions
2012-06-04 15:52:06 +00:00
fix endianess bug in LZMA decoding
2012-06-06 18:58:24 +00:00
ifeq (,$(shell ld -v 2>&1 | grep cctools))
fix build on old Xcode versions
2012-06-04 15:52:06 +00:00
lflags += -Wl,-compatibility_version,1.0.0
endif
rename platform=darwin -> platform={ios,macosx}, fix ios simulator build
2014-06-22 18:30:52 +00:00
ifneq ($(platform),ios)
fix OS X OpenJDK 7 build
2012-03-11 22:26:46 +00:00
lflags += -framework CoreServices -framework SystemConfiguration \
-framework Security
add support for Apple iOS on ARM Most tests are passing, but not all, so this still needs more work.
2011-08-11 01:21:48 +00:00
endif
support bootimage build on OS X
2008-12-19 02:52:03 +00:00
ifeq ($(bootimage),true)
specify segment __RWX in bootimage-lflags on Darwin to match new name used in binaryToObject
2009-10-14 01:14:03 +00:00
bootimage-lflags = -Wl,-segprot,__RWX,rwx,rwx
support bootimage build on OS X
2008-12-19 02:52:03 +00:00
endif
fix darwin build
2008-03-31 04:23:51 +00:00
rdynamic =
Got strip working for Mac OS X
2007-11-01 20:37:51 +00:00
strip-all = -S -x
fix OS X openjdk and openjdk-src builds
2010-11-16 03:28:53 +00:00
so-suffix = .dylib
support building Avian as a self-contained dynamic library
2008-03-31 03:43:43 +00:00
shared = -dynamiclib
don't specify -Wl,-rpath argument on Darwin or Windows It remains to be seen what form that argument should take on those operating systems; for now, it only causes trouble, so we'll just leave it out.
2013-02-12 15:22:21 +00:00
rpath =
specify -arch explicitly to compiler and liner on OS X On 10.6 the compiler generates x86_64 code by default, even though uname -m reports "i386", so we have to be explicit
2009-11-03 23:07:13 +00:00
rename platform=darwin -> platform={ios,macosx}, fix ios simulator build
2014-06-22 18:30:52 +00:00
ifeq ($(platform),ios)
add support for iOS/x86_64
2015-03-16 17:32:35 +00:00
ifeq (,$(filter arm arm64,$(arch)))
fix Android classpath build for iOS This mainly involved reworking the makefile to avoid conflating Darwin/ARM builds with iOS, since we may also want to build for the iOS Simulator, which is i386. Note that I was only able to test this on the Simulator, since I don't have a real iOS device to test with. Sorry if I broke something; if so, please fix it :)
2013-07-31 22:20:32 +00:00
target = iPhoneSimulator
sdk = iphonesimulator$(ios-version)
add support for iOS/x86_64
2015-03-16 17:32:35 +00:00
ifeq ($(arch),i386)
arch-flag = -arch i386
else
arch-flag = -arch x86_64
endif
fix Android classpath build for iOS This mainly involved reworking the makefile to avoid conflating Darwin/ARM builds with iOS, since we may also want to build for the iOS Simulator, which is i386. Note that I was only able to test this on the Simulator, since I don't have a real iOS device to test with. Sorry if I broke something; if so, please fix it :)
2013-07-31 22:20:32 +00:00
release = Release-iphonesimulator
else
target = iPhoneOS
sdk = iphoneos$(ios-version)
add process=interpret support for Linux/ARM64 This makes all the tests pass for the platform=linux arch=arm64 process=interpret build. Next step: process=compile support.
2014-11-26 01:36:34 +00:00
ifeq ($(arch),arm)
arch-flag = -arch armv7
else
arch-flag = -arch arm64
endif
fix Android classpath build for iOS This mainly involved reworking the makefile to avoid conflating Darwin/ARM builds with iOS, since we may also want to build for the iOS Simulator, which is i386. Note that I was only able to test this on the Simulator, since I don't have a real iOS device to test with. Sorry if I broke something; if so, please fix it :)
2013-07-31 22:20:32 +00:00
release = Release-iphoneos
endif
platform-dir = $(developer-dir)/Platforms/$(target).platform
sdk-dir = $(platform-dir)/Developer/SDKs
Use latest iOS SDK found on local machine
2012-03-26 17:12:07 +00:00
fix Android classpath build for iOS This mainly involved reworking the makefile to avoid conflating Darwin/ARM builds with iOS, since we may also want to build for the iOS Simulator, which is i386. Note that I was only able to test this on the Simulator, since I don't have a real iOS device to test with. Sorry if I broke something; if so, please fix it :)
2013-07-31 22:20:32 +00:00
ios-version := $(shell \
add sdk 8.3 support
2015-05-01 19:02:52 +00:00
if test -d $(sdk-dir)/$(target)8.3.sdk; then echo 8.3; \
elif test -d $(sdk-dir)/$(target)8.2.sdk; then echo 8.2; \
add support for iOS/x86_64
2015-03-16 17:32:35 +00:00
elif test -d $(sdk-dir)/$(target)8.1.sdk; then echo 8.1; \
add support for 8.1 SDK
2014-12-04 21:47:26 +00:00
elif test -d $(sdk-dir)/$(target)8.0.sdk; then echo 8.0; \
add detection for iOS 8 SDK
2014-10-06 15:11:56 +00:00
elif test -d $(sdk-dir)/$(target)7.1.sdk; then echo 7.1; \
add support for 7.1 ios SDK
2014-04-30 19:23:22 +00:00
elif test -d $(sdk-dir)/$(target)7.0.sdk; then echo 7.0; \
fix iOS 7 build using latest Xcode It looks like the iOS 7 SDK doesn't have GCC anymore, so we need to use clang instead. Also, thread_act.h and thread_status.h have moved, so I updated arm.h accordingly. That might break the build for older SDKs, but I don't have one available at the moment. If it does break, I'll fix it.
2013-12-18 18:10:04 +00:00
elif test -d $(sdk-dir)/$(target)6.1.sdk; then echo 6.1; \
fix Android classpath build for iOS This mainly involved reworking the makefile to avoid conflating Darwin/ARM builds with iOS, since we may also want to build for the iOS Simulator, which is i386. Note that I was only able to test this on the Simulator, since I don't have a real iOS device to test with. Sorry if I broke something; if so, please fix it :)
2013-07-31 22:20:32 +00:00
elif test -d $(sdk-dir)/$(target)6.0.sdk; then echo 6.0; \
elif test -d $(sdk-dir)/$(target)5.1.sdk; then echo 5.1; \
elif test -d $(sdk-dir)/$(target)5.0.sdk; then echo 5.0; \
elif test -d $(sdk-dir)/$(target)4.3.sdk; then echo 4.3; \
use ';' instead of ':' as path separator on windows when running tests
2014-01-08 22:00:19 +00:00
elif test -d $(sdk-dir)/$(target)4.2.sdk; then echo 4.2; \
else echo; fi)
don't fail build on missing iOS SDK unless actually building for iOS
2012-04-07 23:15:59 +00:00
ifeq ($(ios-version),)
fix Android classpath build for iOS This mainly involved reworking the makefile to avoid conflating Darwin/ARM builds with iOS, since we may also want to build for the iOS Simulator, which is i386. Note that I was only able to test this on the Simulator, since I don't have a real iOS device to test with. Sorry if I broke something; if so, please fix it :)
2013-07-31 22:20:32 +00:00
x := $(error "couldn't find SDK")
don't fail build on missing iOS SDK unless actually building for iOS
2012-04-07 23:15:59 +00:00
endif
fix Android classpath build for iOS This mainly involved reworking the makefile to avoid conflating Darwin/ARM builds with iOS, since we may also want to build for the iOS Simulator, which is i386. Note that I was only able to test this on the Simulator, since I don't have a real iOS device to test with. Sorry if I broke something; if so, please fix it :)
2013-07-31 22:20:32 +00:00
ios-bin = $(platform-dir)/Developer/usr/bin
fix iOS 7 build using latest Xcode It looks like the iOS 7 SDK doesn't have GCC anymore, so we need to use clang instead. Also, thread_act.h and thread_status.h have moved, so I updated arm.h accordingly. That might break the build for older SDKs, but I don't have one available at the moment. If it does break, I'll fix it.
2013-12-18 18:10:04 +00:00
found-gcc = $(shell if test -f $(ios-bin)/gcc; then echo true; else echo false; fi)
ifeq ($(found-gcc),false)
use-clang = true
endif
fix Android classpath build for iOS This mainly involved reworking the makefile to avoid conflating Darwin/ARM builds with iOS, since we may also want to build for the iOS Simulator, which is i386. Note that I was only able to test this on the Simulator, since I don't have a real iOS device to test with. Sorry if I broke something; if so, please fix it :)
2013-07-31 22:20:32 +00:00
ifeq ($(use-clang),true)
cxx = clang -std=c++11
cc = clang
else
cxx = $(ios-bin)/g++
cc = $(ios-bin)/gcc
endif
flags = -isysroot $(sdk-dir)/$(target)$(ios-version).sdk \
$(arch-flag)
various tweaks to get hello-ios working with the Android library
2013-03-04 21:24:41 +00:00
classpath-extra-cflags += $(flags)
add support for Apple iOS on ARM Most tests are passing, but not all, so this still needs more work.
2011-08-11 01:21:48 +00:00
cflags += $(flags)
asmflags += $(flags)
lflags += $(flags)
endif
specify -arch explicitly to compiler and liner on OS X On 10.6 the compiler generates x86_64 code by default, even though uname -m reports "i386", so we have to be explicit
2009-11-03 23:07:13 +00:00
ifeq ($(arch),i386)
rename platform=darwin -> platform={ios,macosx}, fix ios simulator build
2014-06-22 18:30:52 +00:00
ifeq ($(platform),ios)
classpath-extra-cflags += \
-arch i386 -miphoneos-version-min=$(ios-version)
cflags += -arch i386 -miphoneos-version-min=$(ios-version)
asmflags += -arch i386 -miphoneos-version-min=$(ios-version)
lflags += -arch i386 -miphoneos-version-min=$(ios-version)
else
classpath-extra-cflags += \
-arch i386 -mmacosx-version-min=${OSX_SDK_VERSION}
cflags += -arch i386 -mmacosx-version-min=${OSX_SDK_VERSION}
asmflags += -arch i386 -mmacosx-version-min=${OSX_SDK_VERSION}
lflags += -arch i386 -mmacosx-version-min=${OSX_SDK_VERSION}
endif
specify -arch explicitly to compiler and liner on OS X On 10.6 the compiler generates x86_64 code by default, even though uname -m reports "i386", so we have to be explicit
2009-11-03 23:07:13 +00:00
endif
Support 64-bit Mac builds
2009-12-01 21:24:33 +00:00
ifeq ($(arch),x86_64)
add support for iOS/x86_64
2015-03-16 17:32:35 +00:00
ifeq ($(platform),ios)
classpath-extra-cflags += \
-arch x86_64 -miphoneos-version-min=$(ios-version)
cflags += -arch x86_64 -miphoneos-version-min=$(ios-version)
asmflags += -arch x86_64 -miphoneos-version-min=$(ios-version)
lflags += -arch x86_64 -miphoneos-version-min=$(ios-version)
else
classpath-extra-cflags += -arch x86_64
cflags += -arch x86_64
asmflags += -arch x86_64
lflags += -arch x86_64
endif
Support 64-bit Mac builds
2009-12-01 21:24:33 +00:00
endif
fix jni_md.h not found 10.9 problem
2013-11-05 00:26:57 +00:00
cflags += -I$(JAVA_HOME)/include/darwin
refactor build system to support cross-compiling
2007-10-22 20:56:27 +00:00
endif
build fixes and refactoring
2007-10-26 21:02:39 +00:00
various tweaks to get hello-ios working with the Android library
2013-03-04 21:24:41 +00:00
openjdk-extra-cflags += $(classpath-extra-cflags)
use x86_64-w64-mingw32- prefix for tools like dlltool as necessary This fixes architecture mismatches when cross-compiling.
2014-04-01 21:47:23 +00:00
find-tool = $(shell if ( command -v "$(1)$(2)" >/dev/null ); then (echo "$(1)$(2)") else (echo "$(2)"); fi)
refactor build system to support cross-compiling
2007-10-22 20:56:27 +00:00
ifeq ($(platform),windows)
fix Windows x86_64 build regression due to unspecified AVIAN_TARGET_FORMAT
2013-03-15 23:47:15 +00:00
target-format = pe
clean up binary format code Linux, FreeBSD, and QNX all use ELF, so no need to distinguish between them when generating object files. To avoid confusion, I've switch from using operating system names to using binary format names where applicable.
2012-08-02 18:15:15 +00:00
Updates to .gitignore to get rid of any possibility of checking in binaries in lib/distrib directories, and generalized method of setting location of win32 and win64 libs.
2012-02-20 23:25:37 +00:00
inc = "$(win32)/include"
lib = "$(win32)/lib"
refactor build system to support cross-compiling
2007-10-22 20:56:27 +00:00
initial support for Windows OpenJDK build All the tests are passing for openjdk-src builds, but the non-src openjdk build is crashing and there's trouble loading time zone info from the embedded java.home directory.
2010-11-07 04:21:19 +00:00
embed-prefix = c:/avian-embedded
refactor build system to support cross-compiling
2007-10-22 20:56:27 +00:00
system = windows
more work on the windows port; we now have some tests passing
2007-10-23 20:05:59 +00:00
support building Avian as a self-contained dynamic library
2008-03-31 03:43:43 +00:00
so-prefix =
so-suffix = .dll
Merge branch 'master' into powerpc Conflicts: makefile src/assembler.h src/compile.cpp src/compiler.cpp src/compiler.h src/finder.cpp
2008-11-11 15:20:49 +00:00
exe-suffix = .exe
don't specify -Wl,-rpath argument on Darwin or Windows It remains to be seen what form that argument should take on those operating systems; for now, it only causes trouble, so we'll just leave it out.
2013-02-12 15:22:21 +00:00
rpath =
support building Avian as a self-contained dynamic library
2008-03-31 03:43:43 +00:00
Windows Phone 8 / Windows RT initial support Conflicts: makefile
2013-01-28 15:20:52 +00:00
lflags = -L$(lib) $(common-lflags) -lws2_32 -liphlpapi -mconsole
fix MSYS/MinGW build MinGW defines __STRICT_ANSI__ when -std=c++0x is specified, which we have to override in order to get e.g. strdup. Also, we need to specify -D_WIN32_WINNT=0x0500 to get post-Windows-2000 functions like GetModuleHandleEx. Finally, it seems that GCC 4.8.1 wants us to use %I64d instead of %lld on Windows.
2014-10-27 15:39:53 +00:00
cflags = -I$(inc) $(common-cflags) -DWINVER=0x0500 -U__STRICT_ANSI__
fix Windows cross build
2011-09-30 21:21:12 +00:00
Merge branch 'master' into powerpc Conflicts: makefile src/assembler.h src/compile.cpp src/compiler.cpp src/compiler.h src/finder.cpp
2008-11-11 15:20:49 +00:00
ifeq (,$(filter mingw32 cygwin,$(build-platform)))
fix Cygwin/MSYS openjdk-src builds
2010-11-15 23:27:00 +00:00
openjdk-extra-cflags += -I$(src)/openjdk/caseSensitive
fall back to x86_64-w64-mingw32 compiler for 32-bit Windows builds if i686-w64-mingw32 compiler not found A multilib-capable x86_64-w64-mingw32 compiler should work just fine, but since we don't know if it's mutilib or not, we try the i686-w64-mingw32 version first.
2012-07-06 17:21:53 +00:00
prefix := $(shell i686-w64-mingw32-gcc --version >/dev/null 2>&1 \
&& echo i686-w64-mingw32- || echo x86_64-w64-mingw32-)
cxx = $(prefix)g++ -m32
cc = $(prefix)gcc -m32
add process=interpret support for Linux/ARM64 This makes all the tests pass for the platform=linux arch=arm64 process=interpret build. Next step: process=compile support.
2014-11-26 01:36:34 +00:00
dlltool = $(prefix)dlltool -mi386 --as-flags=--32
fall back to x86_64-w64-mingw32 compiler for 32-bit Windows builds if i686-w64-mingw32 compiler not found A multilib-capable x86_64-w64-mingw32 compiler should work just fine, but since we don't know if it's mutilib or not, we try the i686-w64-mingw32 version first.
2012-07-06 17:21:53 +00:00
ar = $(prefix)ar
ranlib = $(prefix)ranlib
strip = $(prefix)strip --strip-all
Merge branch 'master' into powerpc Conflicts: makefile src/assembler.h src/compile.cpp src/compiler.cpp src/compiler.h src/finder.cpp
2008-11-11 15:20:49 +00:00
else
fix native Windows GCC 3.4 build
2010-11-12 23:53:16 +00:00
build-system = windows
statically link libstdc++ to type-generator on windows (for real, this time!)
2014-07-17 18:22:24 +00:00
static-on-windows = -static
fix Cygwin build
2009-02-14 00:03:46 +00:00
common-cflags += "-I$(JAVA_HOME)/include/win32"
fix MSYS/MinGW build MinGW defines __STRICT_ANSI__ when -std=c++0x is specified, which we have to override in order to get e.g. strdup. Also, we need to specify -D_WIN32_WINNT=0x0500 to get post-Windows-2000 functions like GetModuleHandleEx. Finally, it seems that GCC 4.8.1 wants us to use %I64d instead of %lld on Windows.
2014-10-27 15:39:53 +00:00
build-cflags = $(common-cflags) -I$(src) -I$(inc) -mthreads \
-D_WIN32_WINNT=0x0500
fix Cygwin/MSYS openjdk-src builds
2010-11-15 23:27:00 +00:00
openjdk-extra-cflags =
fix MSYS OpenJDK build
2010-11-13 02:04:17 +00:00
build-lflags = -L$(lib) $(common-lflags)
Merge branch 'master' into powerpc Conflicts: makefile src/assembler.h src/compile.cpp src/compiler.cpp src/compiler.h src/finder.cpp
2008-11-11 15:20:49 +00:00
ifeq ($(build-platform),cygwin)
fix Cygwin 1.7.x build Recent versions of Cygwin's GCC no longer support the -mno-cygwin flag (compilation works, but linking fails). However, Cygwin now includes mingw64-i686-gcc, mingw64-x86_64-gcc, and related packages, so we use those compilers instead. This has the added benefit that we can build native x86_64 binaries on Cygwin now. This will break the build for old Cygwin installs, so it will be necessary to upgrade Cygwin and install the aforementioned packages to build Avian going forward.
2011-07-12 22:58:51 +00:00
build-cxx = i686-w64-mingw32-g++
build-cc = i686-w64-mingw32-gcc
dlltool = i686-w64-mingw32-dlltool
ar = i686-w64-mingw32-ar
ranlib = i686-w64-mingw32-ranlib
strip = i686-w64-mingw32-strip
Merge branch 'master' into powerpc Conflicts: makefile src/assembler.h src/compile.cpp src/compiler.cpp src/compiler.h src/finder.cpp
2008-11-11 15:20:49 +00:00
endif
endif
added windows-x86_64 build target
2009-06-11 15:36:53 +00:00
ifeq ($(arch),x86_64)
fix Cygwin 1.7.x build Recent versions of Cygwin's GCC no longer support the -mno-cygwin flag (compilation works, but linking fails). However, Cygwin now includes mingw64-i686-gcc, mingw64-x86_64-gcc, and related packages, so we use those compilers instead. This has the added benefit that we can build native x86_64 binaries on Cygwin now. This will break the build for old Cygwin installs, so it will be necessary to upgrade Cygwin and install the aforementioned packages to build Avian going forward.
2011-07-12 22:58:51 +00:00
ifeq ($(build-platform),cygwin)
build-cxx = x86_64-w64-mingw32-g++
build-cc = x86_64-w64-mingw32-gcc
endif
Specify -m flag on windows x86_64 gcc/g++ compilers.
2010-05-13 23:01:55 +00:00
cxx = x86_64-w64-mingw32-g++ $(mflag)
cc = x86_64-w64-mingw32-gcc $(mflag)
use x86_64-w64-mingw32- prefix for tools like dlltool as necessary This fixes architecture mismatches when cross-compiling.
2014-04-01 21:47:23 +00:00
dlltool = $(call find-tool,x86_64-w64-mingw32-,dlltool)
ar = $(call find-tool,x86_64-w64-mingw32-,ar)
ranlib = $(call find-tool,x86_64-w64-mingw32-,ranlib)
strip = $(call find-tool,x86_64-w64-mingw32-,strip)
Updates to .gitignore to get rid of any possibility of checking in binaries in lib/distrib directories, and generalized method of setting location of win32 and win64 libs.
2012-02-20 23:25:37 +00:00
inc = "$(win64)/include"
lib = "$(win64)/lib"
various Windows OpenJDK 7 fixes
2012-03-18 00:47:22 +00:00
else
shared += -Wl,--add-stdcall-alias
added windows-x86_64 build target
2009-06-11 15:36:53 +00:00
endif
fix Windows cross-compile build
2012-12-05 00:37:02 +00:00
Allow output of exceptions to debugger. Generate WinMD file
2013-01-30 08:42:05 +00:00
embed = $(build-embed)/embed$(exe-suffix)
embed-loader = $(build-embed-loader)/embed-loader$(exe-suffix)
Added new embed utility for Windows
2012-11-13 08:11:49 +00:00
embed-loader-o = $(build-embed)/embed-loader.o
refactor build system to support cross-compiling
2007-10-22 20:56:27 +00:00
endif
set minimum value for initial nextGen2 capacity; profile support
2007-08-18 22:42:11 +00:00
Windows Phone 8 / Windows RT initial support Conflicts: makefile
2013-01-28 15:20:52 +00:00
ifeq ($(platform),wp8)
ifeq ($(shell uname -s | grep -i -c WOW64),1)
programFiles = Program Files (x86)
fix OS X build regressions
2012-06-03 00:56:07 +00:00
else
Windows Phone 8 / Windows RT initial support Conflicts: makefile
2013-01-28 15:20:52 +00:00
programFiles = Program Files
fix OS X build regressions
2012-06-03 00:56:07 +00:00
endif
Windows Phone 8 / Windows RT initial support Conflicts: makefile
2013-01-28 15:20:52 +00:00
ifeq ($(MSVS_ROOT),)
# Environment variable MSVS_ROOT not found. It should be something like
# "C:\$(programFiles)\Microsoft Visual Studio 11.0"
MSVS_ROOT = C:\$(programFiles)\Microsoft Visual Studio 11.0
endif
ifeq ($(MSVC_ROOT),)
# Environment variable MSVC_ROOT not found. It should be something like
# "C:\$(programFiles)\Microsoft Visual Studio 11.0\VC"
MSVC_ROOT = $(MSVS_ROOT)\VC
endif
ifeq ($(WP80_SDK),)
# Environment variable WP8_SDK not found. It should be something like
# "C:\Program Files[ (x86)]\Microsoft Visual Studio 11.0\VC\WPSDK\WP80"
# TODO: Lookup in SOFTWARE\Microsoft\Microsoft SDKs\WindowsPhone\v8.0
Fixes to WP8/WinRT support
2013-01-29 12:21:02 +00:00
WP80_SDK = $(MSVS_ROOT)\VC\WPSDK\WP80
Windows Phone 8 / Windows RT initial support Conflicts: makefile
2013-01-28 15:20:52 +00:00
endif
ifeq ($(WP80_KIT),)
# Environment variable WP8_KIT not found. It should be something like
# "c:\Program Files[ (x86)]\Windows Phone Kits\8.0"
# TODO: Lookup in SOFTWARE\Microsoft\Microsoft SDKs\WindowsPhone\v8.0
WP80_KIT = C:\$(programFiles)\Windows Phone Kits\8.0
endif
ifeq ($(WIN8_KIT),)
# Environment variable WIN8_KIT not found. It should be something like
# "c:\Program Files[ (x86)]\Windows Kits\8.0"
WIN8_KIT = C:\$(programFiles)\Windows Kits\8.0
endif
ifeq ($(build-platform),cygwin)
windows-path = cygpath -w
add use-clang option for building with LLVM Clang instead of GCC This also fixes several errors and warnings emitted by Clang.
2012-06-01 23:43:42 +00:00
else
Windows Phone 8 / Windows RT initial support Conflicts: makefile
2013-01-28 15:20:52 +00:00
windows-path = $(native-path)
add use-clang option for building with LLVM Clang instead of GCC This also fixes several errors and warnings emitted by Clang.
2012-06-01 23:43:42 +00:00
endif
Windows Phone 8 / Windows RT initial support Conflicts: makefile
2013-01-28 15:20:52 +00:00
windows-java-home := $(shell $(windows-path) "$(JAVA_HOME)")
target-format = pe
ms_cl_compiler = wp8
use-lto = false
supports_avian_executable = false
Makefile changes, added additional compiler flags
2013-01-30 05:31:02 +00:00
aot-only = true
WP8 target refine
2013-02-09 11:09:58 +00:00
ifneq ($(bootimage),true)
Lost assignment
2013-02-09 11:18:39 +00:00
x := $(error Windows Phone 8 target requires bootimage=true)
Windows Phone 8 / Windows RT initial support Conflicts: makefile
2013-01-28 15:20:52 +00:00
endif
system = windows
build-system = windows
static-prefix =
static-suffix = .lib
so-prefix =
so-suffix = .dll
exe-suffix = .exe
Allow output of exceptions to debugger. Generate WinMD file
2013-01-30 08:42:05 +00:00
manifest-flags = -MANIFEST:NO
enable link time optimization when using GCC 4.6 or greater
2011-03-26 20:44:08 +00:00
Windows Phone 8 / Windows RT initial support Conflicts: makefile
2013-01-28 15:20:52 +00:00
ifeq ($(arch),arm)
wp8_arch = \x86_arm
vc_arch = \arm
w8kit_arch = arm
deps_arch = ARM
as = "$$(cygpath -u "$(WP80_SDK)\bin\x86_arm\armasm.exe")"
cxx = "$$(cygpath -u "$(WP80_SDK)\bin\x86_arm\cl.exe")"
ld = "$$(cygpath -u "$(WP80_SDK)\bin\x86_arm\link.exe")"
Fix WP8/WinRT ARM build
2013-01-31 10:54:51 +00:00
asmflags = -machine ARM -32
Windows Phone 8 / Windows RT initial support Conflicts: makefile
2013-01-28 15:20:52 +00:00
asm-output = -o $(1)
asm-input = $(1)
machine_type = ARM
Fix WP8/WinRT ARM build
2013-01-31 10:54:51 +00:00
bootimage-symbols = binary_bootimage_bin_start:binary_bootimage_bin_end
codeimage-symbols = binary_codeimage_bin_start:binary_codeimage_bin_end
Windows Phone 8 / Windows RT initial support Conflicts: makefile
2013-01-28 15:20:52 +00:00
endif
ifeq ($(arch),i386)
wp8_arch =
vc_arch =
w8kit_arch = x86
deps_arch = x86
Makefile changes, added additional compiler flags
2013-01-30 05:31:02 +00:00
asmflags = $(target-cflags) -safeseh -nologo -Gd
Windows Phone 8 / Windows RT initial support Conflicts: makefile
2013-01-28 15:20:52 +00:00
as = "$$(cygpath -u "$(WP80_SDK)\bin\ml.exe")"
cxx = "$$(cygpath -u "$(WP80_SDK)\bin\cl.exe")"
ld = "$$(cygpath -u "$(WP80_SDK)\bin\link.exe")"
Makefile changes, added additional compiler flags
2013-01-30 05:31:02 +00:00
ifeq ($(mode),debug)
asmflags += -Zd
add use-clang option for building with LLVM Clang instead of GCC This also fixes several errors and warnings emitted by Clang.
2012-06-01 23:43:42 +00:00
endif
Makefile changes, added additional compiler flags
2013-01-30 05:31:02 +00:00
ifeq ($(mode),debug-fast)
asmflags += -Zd
add use-clang option for building with LLVM Clang instead of GCC This also fixes several errors and warnings emitted by Clang.
2012-06-01 23:43:42 +00:00
endif
Windows Phone 8 / Windows RT initial support Conflicts: makefile
2013-01-28 15:20:52 +00:00
asm-output = $(output)
machine_type = X86
enable link time optimization when using GCC 4.6 or greater
2011-03-26 20:44:08 +00:00
endif
copy makefile, stdc++.cpp, and type-generator.cpp from mess project and began customizing them for this project
2007-06-03 23:16:07 +00:00
Windows Phone 8 / Windows RT initial support Conflicts: makefile
2013-01-28 15:20:52 +00:00
PATH := $(shell cygpath -u "$(MSVS_ROOT)\Common7\IDE"):$(shell cygpath -u "$(WP80_SDK)\bin$(wp8_arch)"):$(shell cygpath -u "$(WP80_SDK)\bin"):${PATH}
enable standalone OpenJDK builds As described in readme.txt, a standalone OpenJDK build embeds all libraries, classes, and other files needed at runtime in the resulting binary, eliminating dependencies on external resources.
2010-11-04 17:02:09 +00:00
Windows Phone 8 / Windows RT initial support Conflicts: makefile
2013-01-28 15:20:52 +00:00
build-cflags = $(common-cflags) -I$(src) -I$(inc) -mthreads
build-lflags = -lz -lpthread
specify -march=i486 for x86_32 to support __sync_bool_compare_and_swap
2009-11-20 01:51:43 +00:00
Windows Phone 8 / Windows RT initial support Conflicts: makefile
2013-01-28 15:20:52 +00:00
cflags = -nologo \
Allow output of exceptions to debugger. Generate WinMD file
2013-01-30 08:42:05 +00:00
-AI"$(WP80_KIT)\Windows Metadata" \
Windows Phone 8 / Windows RT initial support Conflicts: makefile
2013-01-28 15:20:52 +00:00
-I"$(WP80_SDK)\include" -I"$(WP80_KIT)\Include" -I"$(WP80_KIT)\Include\minwin" -I"$(WP80_KIT)\Include\mincore" \
Allow output of exceptions to debugger. Generate WinMD file
2013-01-30 08:42:05 +00:00
-DWINAPI_FAMILY=WINAPI_FAMILY_PHONE_APP -D_USRDLL -D_WINDLL \
Windows Phone 8 / Windows RT initial support Conflicts: makefile
2013-01-28 15:20:52 +00:00
-DAVIAN_VERSION=\"$(version)\" -D_JNI_IMPLEMENTATION_ \
-DUSE_ATOMIC_OPERATIONS -DAVIAN_JAVA_HOME=\"$(javahome)\" \
-DAVIAN_EMBED_PREFIX=\"$(embed-prefix)\" \
Support new WinRT interop
2013-02-02 07:38:03 +00:00
-I"$(shell $(windows-path) "$(wp8)/zlib/upstream")" -I"$(shell $(windows-path) "$(wp8)/interop/avian-interop-client")" \
-I"$(shell $(windows-path) "$(wp8)/include")" -I$(src) -I$(classpath-src) \
Windows Phone 8 / Windows RT initial support Conflicts: makefile
2013-01-28 15:20:52 +00:00
-I"$(build)" \
-I"$(windows-java-home)/include" -I"$(windows-java-home)/include/win32" \
Makefile changes, added additional compiler flags
2013-01-30 05:31:02 +00:00
-DTARGET_BYTES_PER_WORD=$(pointer-size) \
Support new WinRT interop
2013-02-02 07:38:03 +00:00
-Gd -EHsc
add support for building with MSVC on Windows
2009-08-27 00:26:44 +00:00
various tweaks to get hello-ios working with the Android library
2013-03-04 21:24:41 +00:00
common-lflags =
Windows Phone 8 / Windows RT initial support Conflicts: makefile
2013-01-28 15:20:52 +00:00
Fix library overwriting ; proper assembler flags
2013-01-29 10:44:55 +00:00
ifeq ($(mode),debug)
build-type = Debug
endif
ifeq ($(mode),debug-fast)
build-type = Debug
endif
use stress-major consistently in makefile
2014-07-01 14:05:12 +00:00
ifeq ($(mode),stress-major)
Fix library overwriting ; proper assembler flags
2013-01-29 10:44:55 +00:00
build-type = Release
endif
ifeq ($(mode),fast)
build-type = Release
endif
ifeq ($(mode),fast)
build-type = Release
endif
ifeq ($(mode),small)
build-type = Release
endif
Windows Phone 8 / Windows RT initial support Conflicts: makefile
2013-01-28 15:20:52 +00:00
arflags = -MACHINE:$(machine_type)
lflags = $(common-lflags) -nologo \
-MACHINE:$(machine_type) \
Finally, proper jvm.dll linking for WP8
2013-01-30 09:31:06 +00:00
-LIBPATH:"$(WP80_KIT)\lib\$(w8kit_arch)" -LIBPATH:"$(WP80_SDK)\lib$(vc_arch)" -LIBPATH:"$(WIN8_KIT)\Lib\win8\um\$(w8kit_arch)" \
Windows Phone 8 / Windows RT initial support Conflicts: makefile
2013-01-28 15:20:52 +00:00
ws2_32.lib \
Support new WinRT interop
2013-02-02 07:38:03 +00:00
"$(shell $(windows-path) "$(wp8)\lib\$(deps_arch)\$(build-type)\zlib.lib")" "$(shell $(windows-path) "$(wp8)\lib\$(deps_arch)\$(build-type)\ThreadEmulation.lib")" \
"$(shell $(windows-path) "$(wp8)\lib\$(deps_arch)\$(build-type)\AvianInteropClient.lib")"
Allow output of exceptions to debugger. Generate WinMD file
2013-01-30 08:42:05 +00:00
lflags += -NXCOMPAT -DYNAMICBASE -SUBSYSTEM:CONSOLE -TLBID:1
Finally, proper jvm.dll linking for WP8
2013-01-30 09:31:06 +00:00
lflags += -NODEFAULTLIB:"ole32.lib" -NODEFAULTLIB:"kernel32.lib"
lflags += PhoneAppModelHost.lib WindowsPhoneCore.lib -WINMD -WINMDFILE:$(subst $(so-suffix),.winmd,$(@))
Windows Phone 8 / Windows RT initial support Conflicts: makefile
2013-01-28 15:20:52 +00:00
cc = $(cxx)
asm-format = masm
shared = -dll
ar = "$$(cygpath -u "$(WP80_SDK)\bin\lib.exe")"
arflags += -nologo
ifeq ($(build-platform),cygwin)
build-cxx = i686-w64-mingw32-g++
build-cc = i686-w64-mingw32-gcc
dlltool = i686-w64-mingw32-dlltool
ranlib =
strip =
endif
output = -Fo$(1)
Allow output of exceptions to debugger. Generate WinMD file
2013-01-30 08:42:05 +00:00
#TODO: -MT or -ZW?
Windows Phone 8 / Windows RT initial support Conflicts: makefile
2013-01-28 15:20:52 +00:00
cflags_debug = -Od -Zi -MDd
cflags_debug_fast = -Od -Zi -MDd
cflags_stress = -O0 -g3 -MD
cflags_stress_major = -O0 -g3 -MD
cflags_fast = -O2 -Zi -MD
cflags_small = -O1s -Zi -MD
# -GL [whole program optimization] in 'fast' and 'small' breaks compilation for some reason
ifeq ($(mode),debug)
cflags +=
lflags +=
endif
ifeq ($(mode),debug-fast)
cflags += -DNDEBUG
lflags +=
endif
use stress-major consistently in makefile
2014-07-01 14:05:12 +00:00
ifeq ($(mode),stress-major)
Windows Phone 8 / Windows RT initial support Conflicts: makefile
2013-01-28 15:20:52 +00:00
cflags +=
lflags +=
endif
ifeq ($(mode),fast)
cflags +=
lflags +=
endif
# -LTCG is needed only if -GL is used
ifeq ($(mode),fast)
cflags += -DNDEBUG
lflags += -LTCG
arflags +=
endif
ifeq ($(mode),small)
cflags += -DNDEBUG
lflags += -LTCG
arflags +=
endif
strip = :
endif
fix MSVC build regression
2012-12-05 22:30:49 +00:00
add support for building with MSVC on Windows
2009-08-27 00:26:44 +00:00
ifdef msvc
fix msvc build
2013-03-04 18:09:59 +00:00
target-format = pe
add support for LZMA on Windows
2012-06-02 21:43:42 +00:00
windows-path = $(native-path)
windows-java-home := $(shell $(windows-path) "$(JAVA_HOME)")
zlib := $(shell $(windows-path) "$(win32)/msvc")
Makefile changes, added additional compiler flags
2013-01-30 05:31:02 +00:00
ms_cl_compiler = regular
fix msvc build
2013-03-04 18:09:59 +00:00
as = $(build-cc)
add support for building with MSVC on Windows
2009-08-27 00:26:44 +00:00
cxx = "$(msvc)/BIN/cl.exe"
cc = $(cxx)
ld = "$(msvc)/BIN/link.exe"
add manifest to executables and DLLs when building with MSVC
2009-09-01 14:42:33 +00:00
mt = "mt.exe"
fix msvc build
2013-03-04 18:09:59 +00:00
ar = "$(msvc)/BIN/lib.exe"
Makefile changes, added additional compiler flags
2013-01-30 05:31:02 +00:00
manifest-flags = -MANIFEST -MANIFESTFILE:$(@).manifest
add support for building with MSVC on Windows
2009-08-27 00:26:44 +00:00
cflags = -nologo -DAVIAN_VERSION=\"$(version)\" -D_JNI_IMPLEMENTATION_ \
fix MSVC build regressions
2011-01-21 23:14:21 +00:00
-DUSE_ATOMIC_OPERATIONS -DAVIAN_JAVA_HOME=\"$(javahome)\" \
-DAVIAN_EMBED_PREFIX=\"$(embed-prefix)\" \
fix MSVC build of test JNI library
2012-08-15 23:42:31 +00:00
-Fd$(build)/$(name).pdb -I"$(zlib)/include" -I$(src) -I$(classpath-src) \
fix msvc build
2013-03-04 18:09:59 +00:00
-I"$(build)" -Iinclude \
fix MSVC build
2012-02-17 01:22:11 +00:00
-I"$(windows-java-home)/include" -I"$(windows-java-home)/include/win32" \
correctly define the target platform in the case of a bootimage build
2012-05-03 14:45:17 +00:00
-DTARGET_BYTES_PER_WORD=$(pointer-size)
add support for LZMA on Windows
2012-06-02 21:43:42 +00:00
Makefile changes, added additional compiler flags
2013-01-30 05:31:02 +00:00
ifneq ($(lzma),)
add support for LZMA on Windows
2012-06-02 21:43:42 +00:00
cflags += -I$(shell $(windows-path) "$(lzma)")
endif
add support for building with MSVC on Windows
2009-08-27 00:26:44 +00:00
shared = -dll
lflags = -nologo -LIBPATH:"$(zlib)/lib" -DEFAULTLIB:ws2_32 \
fix MSVC build regression
2012-12-05 22:30:49 +00:00
-DEFAULTLIB:zlib -DEFAULTLIB:user32 -MANIFEST -debug
fix OSX build regression
2009-08-27 15:30:34 +00:00
output = -Fo$(1)
add support for building with MSVC on Windows
2009-08-27 00:26:44 +00:00
fix msvc build
2013-03-04 18:09:59 +00:00
cflags_debug = -Od -Zi -MDd
cflags_debug_fast = -Od -Zi -DNDEBUG
cflags_fast = -O2 -GL -Zi -DNDEBUG
cflags_small = -O1s -Zi -GL -DNDEBUG
add support for building with MSVC on Windows
2009-08-27 00:26:44 +00:00
ifeq ($(mode),fast)
lflags += -LTCG
endif
ifeq ($(mode),small)
lflags += -LTCG
endif
fix msvc build
2013-03-04 18:09:59 +00:00
use-lto = false
add support for building with MSVC on Windows
2009-08-27 00:26:44 +00:00
strip = :
endif
implement primitive testing framework and provide for GC stress testing
2007-07-16 01:03:02 +00:00
ifeq ($(mode),debug)
Windows Phone 8 / Windows RT initial support Conflicts: makefile
2013-01-28 15:20:52 +00:00
optimization-cflags = $(cflags_debug)
converter-cflags += $(cflags_debug)
strip binaries if and only if optimizing
2009-08-24 22:38:00 +00:00
strip = :
implement primitive testing framework and provide for GC stress testing
2007-07-16 01:03:02 +00:00
endif
add debug-fast compilation mode
2008-01-12 00:15:34 +00:00
ifeq ($(mode),debug-fast)
Windows Phone 8 / Windows RT initial support Conflicts: makefile
2013-01-28 15:20:52 +00:00
optimization-cflags = $(cflags_debug_fast) -DNDEBUG
strip binaries if and only if optimizing
2009-08-24 22:38:00 +00:00
strip = :
add debug-fast compilation mode
2008-01-12 00:15:34 +00:00
endif
implement primitive testing framework and provide for GC stress testing
2007-07-16 01:03:02 +00:00
ifeq ($(mode),stress)
Windows Phone 8 / Windows RT initial support Conflicts: makefile
2013-01-28 15:20:52 +00:00
optimization-cflags = $(cflags_stress) -DVM_STRESS
strip binaries if and only if optimizing
2009-08-24 22:38:00 +00:00
strip = :
implement primitive testing framework and provide for GC stress testing
2007-07-16 01:03:02 +00:00
endif
ifeq ($(mode),stress-major)
Windows Phone 8 / Windows RT initial support Conflicts: makefile
2013-01-28 15:20:52 +00:00
optimization-cflags = $(cflags_stress_major) -DVM_STRESS -DVM_STRESS_MAJOR
strip binaries if and only if optimizing
2009-08-24 22:38:00 +00:00
strip = :
implement primitive testing framework and provide for GC stress testing
2007-07-16 01:03:02 +00:00
endif
ifeq ($(mode),fast)
Windows Phone 8 / Windows RT initial support Conflicts: makefile
2013-01-28 15:20:52 +00:00
optimization-cflags = $(cflags_fast) -DNDEBUG
set minimum value for initial nextGen2 capacity; profile support
2007-08-18 22:42:11 +00:00
endif
Merge branch 'master' into powerpc Conflicts: makefile src/assembler.h src/compile.cpp src/compiler.cpp src/compiler.h src/finder.cpp
2008-11-11 15:20:49 +00:00
ifeq ($(mode),small)
Windows Phone 8 / Windows RT initial support Conflicts: makefile
2013-01-28 15:20:52 +00:00
optimization-cflags = $(cflags_small) -DNDEBUG
enable link time optimization when using GCC 4.6 or greater
2011-03-26 20:44:08 +00:00
endif
ifeq ($(use-lto),true)
add use-clang option for building with LLVM Clang instead of GCC This also fixes several errors and warnings emitted by Clang.
2012-06-01 23:43:42 +00:00
ifeq ($(use-clang),true)
enable link time optimization when using GCC 4.6 or greater
2011-03-26 20:44:08 +00:00
optimization-cflags += -flto
lflags += $(optimization-cflags)
add use-clang option for building with LLVM Clang instead of GCC This also fixes several errors and warnings emitted by Clang.
2012-06-01 23:43:42 +00:00
else
# only try to use LTO when GCC 4.6.0 or greater is available
gcc-major := $(shell $(cc) -dumpversion | cut -f1 -d.)
gcc-minor := $(shell $(cc) -dumpversion | cut -f2 -d.)
ifeq ($(shell expr 4 \< $(gcc-major) \
\| \( 4 \<= $(gcc-major) \& 6 \<= $(gcc-minor) \)),1)
optimization-cflags += -flto
no-lto = -fno-lto
lflags += $(optimization-cflags)
endif
enable link time optimization when using GCC 4.6 or greater
2011-03-26 20:44:08 +00:00
endif
Merge branch 'master' into powerpc Conflicts: makefile src/assembler.h src/compile.cpp src/compiler.cpp src/compiler.h src/finder.cpp
2008-11-11 15:20:49 +00:00
endif
copy makefile, stdc++.cpp, and type-generator.cpp from mess project and began customizing them for this project
2007-06-03 23:16:07 +00:00
enable standalone OpenJDK builds As described in readme.txt, a standalone OpenJDK build embeds all libraries, classes, and other files needed at runtime in the resulting binary, eliminating dependencies on external resources.
2010-11-04 17:02:09 +00:00
cflags += $(optimization-cflags)
Windows Phone 8 / Windows RT initial support Conflicts: makefile
2013-01-28 15:20:52 +00:00
ifndef ms_cl_compiler
rename platform=darwin -> platform={ios,macosx}, fix ios simulator build
2014-06-22 18:30:52 +00:00
ifneq ($(kernel),darwin)
specify -march=i486 for x86_32 to support __sync_bool_compare_and_swap
2009-11-20 01:51:43 +00:00
ifeq ($(arch),i386)
# this is necessary to support __sync_bool_compare_and_swap:
pass -march=i586 to gcc when targetting 32-bit x86 Recent versions of GCC will only give us __sync_bool_compare_and_swap_8 if we target i586 and above, apparently.
2010-10-24 02:38:42 +00:00
cflags += -march=i586
add -march=i586 to lflags as well as cflags where appropriate When link time optimization is enabled, we need to remind the compiler that we're targeting i586 when linking so it can resolve atomic operations like __sync_bool_compare_and_swap.
2012-03-27 00:02:43 +00:00
lflags += -march=i586
specify -march=i486 for x86_32 to support __sync_bool_compare_and_swap
2009-11-20 01:51:43 +00:00
endif
don't specify -march=i486 on Darwin because a later generation is assumed by default
2009-11-20 18:40:55 +00:00
endif
Windows Phone 8 / Windows RT initial support Conflicts: makefile
2013-01-28 15:20:52 +00:00
endif
fix MSVC build regression
2012-12-05 22:30:49 +00:00
add optional LZMA support for compressing embedded JARs, boot images, and shared objects
2012-06-02 15:06:22 +00:00
c-objects = $(foreach x,$(1),$(patsubst $(2)/%.c,$(3)/%.o,$(x)))
refactor build system to support cross-compiling
2007-10-22 20:56:27 +00:00
cpp-objects = $(foreach x,$(1),$(patsubst $(2)/%.cpp,$(3)/%.o,$(x)))
update to more recent version of Android class library Lots has changed since we forked Android's libcore, so merging the latest upstream code has required extensive changes to the Avian/Android port. One big change is that we now use Avian's versions of java.lang.Object, java.lang.Class, java.lang.ClassLoader, some java.lang.reflect.* classes, etc. instead of the Android versions. The main reason is that the Android versions have become very Dex/Dalvik-specific, and since Avian is based on Java class files, not dex archives, that code doesn't make sense here. This has the side benefit that we can share more native code with classpath-avian.cpp and reduce the amount of Java/C++ code duplication.
2014-08-21 19:09:42 +00:00
cc-objects = $(foreach x,$(1),$(patsubst $(2)/%.cc,$(3)/%.o,$(x)))
Windows Phone 8 / Windows RT initial support Conflicts: makefile
2013-01-28 15:20:52 +00:00
asm-objects = $(foreach x,$(1),$(patsubst $(2)/%.$(asm-format),$(3)/%-asm.o,$(x)))
support classloading from jar files compiled into the executable
2007-10-25 22:06:05 +00:00
java-classes = $(foreach x,$(1),$(patsubst $(2)/%.java,$(3)/%.class,$(x)))
Fixed .properties publishing and patching
2014-03-30 19:41:26 +00:00
noop-files = $(foreach x,$(1),$(patsubst $(2)/%,$(3)/%,$(x)))
copy makefile, stdc++.cpp, and type-generator.cpp from mess project and began customizing them for this project
2007-06-03 23:16:07 +00:00
update generated type code and fix most link errors
2007-06-18 19:23:44 +00:00
generated-code = \
build everything (including classes) in options-specific directory Previously, we only built platform specific code in an options-specific directory (e.g. build/linux-x86_64-debug), but built the Java classes in the shared, top-level build directory. This minimized duplication when building with different sets of options, but now that we're supporting the option of selecting a third-party class library, we need to put the classes in the options-specific directory as well.
2010-09-17 22:03:37 +00:00
$(build)/type-enums.cpp \
$(build)/type-declarations.cpp \
$(build)/type-constructors.cpp \
$(build)/type-initializations.cpp \
Merge remote branch 'origin/master' into openjdk Conflicts: makefile src/compile.cpp src/compiler.cpp src/type-generator.cpp
2010-11-16 17:18:08 +00:00
$(build)/type-java-initializations.cpp \
progress towards cross-architecture bootimage builds This monster commit is the first step towards supporting cross-architecture bootimage builds. The challenge is to build a heap and code image for the target platform where the word size and endianess may differ from those of the build architecture. That means the memory layout of objects may differ due to alignment and size differences, so we can't just copy objects into the heap image unchanged; we must copy field by field, resizing values, reversing endianess and shifting offsets as necessary. This commit also removes POD (plain old data) type support from the type generator because it added a lot of complication and little value.
2011-08-30 01:00:17 +00:00
$(build)/type-name-initializations.cpp \
$(build)/type-maps.cpp
more refactoring
2007-07-06 23:50:26 +00:00
move tokenizer.h to include/avian/util/string.h, merge in the String declaration from object-writer/tools.h
2013-02-21 23:18:20 +00:00
vm-depends := $(generated-code) \
find headers regardless of directory layout in vm-depends definition
2013-02-21 00:20:10 +00:00
$(shell find src include -name '*.h' -or -name '*.inc.cpp')
more refactoring
2007-07-06 23:50:26 +00:00
don't include embedded boot classpath in libvm.a, only the executable
2008-01-19 00:54:36 +00:00
vm-sources = \
remove vm from code paths, to match namespacing
2014-02-07 21:24:56 +00:00
$(src)/system/$(system).cpp \
split avian::system::crash into its own file
2014-07-13 01:05:58 +00:00
$(wildcard $(src)/system/$(system)/*.cpp) \
flesh out resource URL scheme implementation
2007-08-10 23:45:47 +00:00
$(src)/finder.cpp \
more refactoring
2007-07-06 23:50:26 +00:00
$(src)/machine.cpp \
flesh out call tracing in new JIT code
2007-11-26 23:15:53 +00:00
$(src)/util.cpp \
add ArchitecturePlanTest
2013-02-15 14:35:17 +00:00
$(src)/heap/heap.cpp \
JIT works for Instructions.java
2007-09-25 23:53:11 +00:00
$(src)/$(process).cpp \
preliminary support for using OpenJDK's class library Whereas the GNU Classpath port used the strategy of patching Classpath with core classes from Avian so as to minimize changes to the VM, this port uses the opposite strategy: abstract and isolate classpath-specific features in the VM similar to how we abstract away platform-specific features in system.h. This allows us to use an unmodified copy of OpenJDK's class library, including its core classes and augmented by a few VM-specific classes in the "avian" package.
2010-09-10 21:05:29 +00:00
$(src)/classpath-$(classpath).cpp \
factor system.cpp and class-finder.cpp out of main.cpp
2007-07-20 14:36:31 +00:00
$(src)/builtin.cpp \
refactor native method resolution to be simpler and more robust
2008-01-28 17:27:02 +00:00
$(src)/jnienv.cpp \
always include heapdump code, to eliminate one dimension of the build matrix
2015-05-02 02:10:33 +00:00
$(src)/process.cpp \
$(src)/heapdump.cpp
more JNI work
2007-06-29 02:58:48 +00:00
split asm files by architecture (x86 -> i386,x86_64 ; arm -> arm,arm64)
2014-12-10 21:41:46 +00:00
vm-asm-sources = $(src)/$(arch).$(asm-format)
refactor build system to support cross-compiling
2007-10-22 20:56:27 +00:00
more progress towards cross-architecture bootimage builds This commit fixes a lot of bugs. All tests are now pass for Linux x86_64 to Linux i386 cross builds.
2011-09-01 03:18:00 +00:00
target-asm = $(asm)
Added new embed utility for Windows
2012-11-13 08:11:49 +00:00
build-embed = $(build)/embed
build-embed-loader = $(build)/embed-loader
Fixed embed.exe loader
2013-04-02 08:37:20 +00:00
embed-loader-sources = $(src)/embedded-loader.cpp
Added new embed utility for Windows
2012-11-13 08:11:49 +00:00
embed-loader-objects = $(call cpp-objects,$(embed-loader-sources),$(src),$(build-embed-loader))
embed-sources = $(src)/embed.cpp
embed-objects = $(call cpp-objects,$(embed-sources),$(src),$(build-embed))
fix unittests for process=interpret build
2013-02-16 03:44:55 +00:00
compiler-sources = \
$(src)/codegen/compiler.cpp \
begin splitting up x86 assembler
2013-02-17 02:50:34 +00:00
$(wildcard $(src)/codegen/compiler/*.cpp) \
add debug-util for printing java bytecode as it's compiled
2014-05-04 01:32:40 +00:00
$(src)/debug-util.cpp \
Extract independent runtime thunks into their own file
2013-12-19 05:46:58 +00:00
$(src)/codegen/runtime.cpp \
move fixed allocator out of machine.h
2014-02-25 22:15:37 +00:00
$(src)/codegen/targets.cpp \
$(src)/util/fixed-allocator.cpp
begin splitting up x86 assembler
2013-02-17 02:50:34 +00:00
group x86, arm, and powerpc target directories into 'target'
2013-02-24 05:47:52 +00:00
x86-assembler-sources = $(wildcard $(src)/codegen/target/x86/*.cpp)
begin splitting up x86 assembler
2013-02-17 02:50:34 +00:00
group x86, arm, and powerpc target directories into 'target'
2013-02-24 05:47:52 +00:00
arm-assembler-sources = $(wildcard $(src)/codegen/target/arm/*.cpp)
begin splitting up x86 assembler
2013-02-17 02:50:34 +00:00
fix unittests for process=interpret build
2013-02-16 03:44:55 +00:00
all-assembler-sources = \
begin splitting up x86 assembler
2013-02-17 02:50:34 +00:00
$(x86-assembler-sources) \
remove powerpc support
2014-04-29 19:26:40 +00:00
$(arm-assembler-sources)
fix unittests for process=interpret build
2013-02-16 03:44:55 +00:00
begin splitting up x86 assembler
2013-02-17 02:50:34 +00:00
native-assembler-sources = $($(target-asm)-assembler-sources)
fix unittests for process=interpret build
2013-02-16 03:44:55 +00:00
all-codegen-target-sources = \
$(compiler-sources) \
$(native-assembler-sources)
refactor build system to support cross-compiling
2007-10-22 20:56:27 +00:00
ifeq ($(process),compile)
fix unittests for process=interpret build
2013-02-16 03:44:55 +00:00
vm-sources += $(compiler-sources)
allow codegen targets (Architectures and Assemblers) to co-exist The primary motivation behind this is to allow all the different Assemblers to be built at once, on a single machine. This should dramatically reduce the time required to make sure that a particular change doesn't break the build for one of the not-so-common architectures (arm, powerpc) Simply pass "codegen-targets=all" to make to compile all src/codegen/<arch>/assembler.cpp. Note that while these architectures are built, they will not be fully- functional. Certain stuff is assumed to be the same across the entire build (such as TargetBytesPerWord), but this isn't the case anymore.
2013-02-12 15:15:30 +00:00
ifeq ($(codegen-targets),native)
fix unittests for process=interpret build
2013-02-16 03:44:55 +00:00
vm-sources += $(native-assembler-sources)
allow codegen targets (Architectures and Assemblers) to co-exist The primary motivation behind this is to allow all the different Assemblers to be built at once, on a single machine. This should dramatically reduce the time required to make sure that a particular change doesn't break the build for one of the not-so-common architectures (arm, powerpc) Simply pass "codegen-targets=all" to make to compile all src/codegen/<arch>/assembler.cpp. Note that while these architectures are built, they will not be fully- functional. Certain stuff is assumed to be the same across the entire build (such as TargetBytesPerWord), but this isn't the case anymore.
2013-02-12 15:15:30 +00:00
endif
ifeq ($(codegen-targets),all)
add process=interpret support for Linux/ARM64 This makes all the tests pass for the platform=linux arch=arm64 process=interpret build. Next step: process=compile support.
2014-11-26 01:36:34 +00:00
ifneq (,$(filter arm arm64,$(arch)))
Fix ci.sh tests on arm qemu systems There are two problems: * The x86 JIT compiler requires detectFeatures, defined in the x86 assembly. Thus it can't (currently) be built on non-x86 platforms. For the purposes of fixing test/ci.sh, it suffices to pretend codegen-targets=all means codegen-targets=native when on arm. * Qemu can introduce some extra latency which was regularly screwing up the LinkedBlockingQueueTest. Solution: increase the timeout to 1/10th seconds.
2014-04-29 20:14:44 +00:00
# The x86 jit has a dependency on the x86 assembly code,
# and thus can't be successfully built on non-x86 platforms.
vm-sources += $(native-assembler-sources)
else
vm-sources += $(all-assembler-sources)
endif
allow codegen targets (Architectures and Assemblers) to co-exist The primary motivation behind this is to allow all the different Assemblers to be built at once, on a single machine. This should dramatically reduce the time required to make sure that a particular change doesn't break the build for one of the not-so-common architectures (arm, powerpc) Simply pass "codegen-targets=all" to make to compile all src/codegen/<arch>/assembler.cpp. Note that while these architectures are built, they will not be fully- functional. Certain stuff is assumed to be the same across the entire build (such as TargetBytesPerWord), but this isn't the case anymore.
2013-02-12 15:15:30 +00:00
endif
more JIT compiler progress
2007-12-09 22:45:43 +00:00
split asm files by architecture (x86 -> i386,x86_64 ; arm -> arm,arm64)
2014-12-10 21:41:46 +00:00
vm-asm-sources += $(src)/compile-$(arch).$(asm-format)
refactor build system to support cross-compiling
2007-10-22 20:56:27 +00:00
endif
Makefile changes, added additional compiler flags
2013-01-30 05:31:02 +00:00
ifeq ($(aot-only),true)
Makefile fix
2013-01-28 17:15:29 +00:00
cflags += -DAVIAN_AOT_ONLY
refactor build system to support cross-compiling
2007-10-22 20:56:27 +00:00
endif
more JNI work
2007-06-29 02:58:48 +00:00
build everything (including classes) in options-specific directory Previously, we only built platform specific code in an options-specific directory (e.g. build/linux-x86_64-debug), but built the Java classes in the shared, top-level build directory. This minimized duplication when building with different sets of options, but now that we're supporting the option of selecting a third-party class library, we need to put the classes in the options-specific directory as well.
2010-09-17 22:03:37 +00:00
vm-cpp-objects = $(call cpp-objects,$(vm-sources),$(src),$(build))
fix unittests for process=interpret build
2013-02-16 03:44:55 +00:00
all-codegen-target-objects = $(call cpp-objects,$(all-codegen-target-sources),$(src),$(build))
build everything (including classes) in options-specific directory Previously, we only built platform specific code in an options-specific directory (e.g. build/linux-x86_64-debug), but built the Java classes in the shared, top-level build directory. This minimized duplication when building with different sets of options, but now that we're supporting the option of selecting a third-party class library, we need to put the classes in the options-specific directory as well.
2010-09-17 22:03:37 +00:00
vm-asm-objects = $(call asm-objects,$(vm-asm-sources),$(src),$(build))
don't include embedded boot classpath in libvm.a, only the executable
2008-01-19 00:54:36 +00:00
vm-objects = $(vm-cpp-objects) $(vm-asm-objects)
copy makefile, stdc++.cpp, and type-generator.cpp from mess project and began customizing them for this project
2007-06-03 23:16:07 +00:00
add process=interpret support for Linux/ARM64 This makes all the tests pass for the platform=linux arch=arm64 process=interpret build. Next step: process=compile support.
2014-11-26 01:36:34 +00:00
heapwalk-sources = $(src)/heapwalk.cpp
initial work to support boot image creation and use
2008-11-21 23:20:35 +00:00
heapwalk-objects = \
build everything (including classes) in options-specific directory Previously, we only built platform specific code in an options-specific directory (e.g. build/linux-x86_64-debug), but built the Java classes in the shared, top-level build directory. This minimized duplication when building with different sets of options, but now that we're supporting the option of selecting a third-party class library, we need to put the classes in the options-specific directory as well.
2010-09-17 22:03:37 +00:00
$(call cpp-objects,$(heapwalk-sources),$(src),$(build))
initial work to support boot image creation and use
2008-11-21 23:20:35 +00:00
find headers regardless of directory layout in vm-depends definition
2013-02-21 00:20:10 +00:00
unittest-objects = $(call cpp-objects,$(unittest-sources),$(unittest),$(build)/unittest)
add unittest framework
2013-02-15 03:54:20 +00:00
always include heapdump code, to eliminate one dimension of the build matrix
2015-05-02 02:10:33 +00:00
vm-heapwalk-objects = $(heapwalk-objects)
initial work to support boot image creation and use
2008-11-21 23:20:35 +00:00
support tail calls and continuations as build options
2009-05-26 05:27:10 +00:00
ifeq ($(tails),true)
cflags += -DAVIAN_TAILS
endif
ifeq ($(continuations),true)
cflags += -DAVIAN_CONTINUATIONS
pass -DAVIAN_CONTINUATIONS to assembler when continuations enabled
2009-05-29 00:51:53 +00:00
asmflags += -DAVIAN_CONTINUATIONS
support tail calls and continuations as build options
2009-05-26 05:27:10 +00:00
endif
first step towards supporting Android class library on Windows A majority of the tests are passing, but more work is needed to get them all to pass.
2013-03-14 21:33:05 +00:00
bootimage-generator-sources = $(src)/tools/bootimage-generator/main.cpp $(src)/util/arg-parser.cpp $(stub-sources)
Merge branch 'master' into wip Conflicts: makefile
2013-02-22 02:33:38 +00:00
add optional LZMA support for compressing embedded JARs, boot images, and shared objects
2012-06-02 15:06:22 +00:00
ifneq ($(lzma),)
bootimage-generator-sources += $(src)/lzma-encode.cpp
endif
first step towards supporting Android class library on Windows A majority of the tests are passing, but more work is needed to get them all to pass.
2013-03-14 21:33:05 +00:00
update makefile to optionally build and use a boot image; various bugfixes
2008-11-29 23:08:14 +00:00
bootimage-generator-objects = \
build everything (including classes) in options-specific directory Previously, we only built platform specific code in an options-specific directory (e.g. build/linux-x86_64-debug), but built the Java classes in the shared, top-level build directory. This minimized duplication when building with different sets of options, but now that we're supporting the option of selecting a third-party class library, we need to put the classes in the options-specific directory as well.
2010-09-17 22:03:37 +00:00
$(call cpp-objects,$(bootimage-generator-sources),$(src),$(build))
fix openjdk-src bootimage build The main change here is to use a lazily-populated vector to associate runtime data with classes instead of referencing them directly from the class which requires updating immutable references in the heap image. The other changes employ other strategies to avoid trying to update immutable references.
2010-11-26 19:41:31 +00:00
bootimage-generator = $(build)/bootimage-generator
update makefile to optionally build and use a boot image; various bugfixes
2008-11-29 23:08:14 +00:00
support AOT-compilation of Java 8 lambda expressions These expressions are tricky because they rely on invokedynamic, which normally implies runtime code generation. However, since lambdas don't actually use the "dynamicness" of invokedynamic, we can convert them into static calls to synthetic classes at compile time. Since I had already written code to synthesize such classes in Java and I didn't want to rewrite it in C++, I needed to add support for running Java code to the bootimage generator. And since the primary VM used by the generator is purpose-built to generate AOT-compiled code for a specific target architecture and is not capable of generating or running JIT-compiled code for the host architecture, I added support for loading a second, independent, host-specific VM for running Java code. The rest of the patch handles the fact that each method compilation might cause new, synthetic classes to be created, so we need to make sure those classes and their methods are included in the final heap and code images. This required breaking some giant code blocks out of makeCodeImage into their own methods, which makes the diff look scarier than it really is.
2015-09-13 02:15:46 +00:00
ifneq ($(mode),fast)
host-vm-options := -$(mode)
endif
host-vm = build/$(build-platform)-$(build-arch)-interpret$(host-vm-options)/libjvm.so
build everything (including classes) in options-specific directory Previously, we only built platform specific code in an options-specific directory (e.g. build/linux-x86_64-debug), but built the Java classes in the shared, top-level build directory. This minimized duplication when building with different sets of options, but now that we're supporting the option of selecting a third-party class library, we need to put the classes in the options-specific directory as well.
2010-09-17 22:03:37 +00:00
bootimage-object = $(build)/bootimage-bin.o
generate read-only code image in bootimage build This avoids the requirement of putting the code image in a section/segment which is both writable and executable, which is good for security and avoids trouble with systems like iOS which disallow such things. The implementation relies on relative addressing such that the offset of the desired address is fixed as a compile-time constant relative to the start of the memory area of interest (e.g. the code image, heap image, or thunk table). At runtime, the base pointer to the memory area is retrieved from the thread structure and added to the offset to compute the final address. Using the thread pointer allows us to generate read-only, position-independent code while avoiding the use of IP-relative addressing, which is not available on all architectures.
2011-09-20 22:30:30 +00:00
codeimage-object = $(build)/codeimage-bin.o
update makefile to optionally build and use a boot image; various bugfixes
2008-11-29 23:08:14 +00:00
ifeq ($(bootimage),true)
generate read-only code image in bootimage build This avoids the requirement of putting the code image in a section/segment which is both writable and executable, which is good for security and avoids trouble with systems like iOS which disallow such things. The implementation relies on relative addressing such that the offset of the desired address is fixed as a compile-time constant relative to the start of the memory area of interest (e.g. the code image, heap image, or thunk table). At runtime, the base pointer to the memory area is retrieved from the thread structure and added to the offset to compute the final address. Using the thread pointer allows us to generate read-only, position-independent code while avoiding the use of IP-relative addressing, which is not available on all architectures.
2011-09-20 22:30:30 +00:00
vm-classpath-objects = $(bootimage-object) $(codeimage-object)
cflags += -DBOOT_IMAGE -DAVIAN_CLASSPATH=\"\"
update makefile to optionally build and use a boot image; various bugfixes
2008-11-29 23:08:14 +00:00
else
generate read-only code image in bootimage build This avoids the requirement of putting the code image in a section/segment which is both writable and executable, which is good for security and avoids trouble with systems like iOS which disallow such things. The implementation relies on relative addressing such that the offset of the desired address is fixed as a compile-time constant relative to the start of the memory area of interest (e.g. the code image, heap image, or thunk table). At runtime, the base pointer to the memory area is retrieved from the thread structure and added to the offset to compute the final address. Using the thread pointer allows us to generate read-only, position-independent code while avoiding the use of IP-relative addressing, which is not available on all architectures.
2011-09-20 22:30:30 +00:00
vm-classpath-objects = $(classpath-object)
fix openjdk-src bootimage build The main change here is to use a lazily-populated vector to associate runtime data with classes instead of referencing them directly from the class which requires updating immutable references in the heap image. The other changes employ other strategies to avoid trying to update immutable references.
2010-11-26 19:41:31 +00:00
cflags += -DBOOT_CLASSPATH=\"[classpathJar]\" \
-DAVIAN_CLASSPATH=\"[classpathJar]\"
update makefile to optionally build and use a boot image; various bugfixes
2008-11-29 23:08:14 +00:00
endif
initial work to support boot image creation and use
2008-11-21 23:20:35 +00:00
support cross builds on OS X to non-Mac targets
2011-08-23 23:04:04 +00:00
cflags += $(extra-cflags)
lflags += $(extra-lflags)
fix Avian_sun_misc_Unsafe_compareAndSwapLong for platforms without atomicCompareAndSwap64 We never define atomicCompareAndSwap64 for ARM or PowerPC, and apparently only very recent ARM chips support it, so we must fall back to synchronization-based emulation.
2012-01-15 17:02:36 +00:00
openjdk-cflags += $(extra-cflags)
support building Avian as a self-contained dynamic library
2008-03-31 03:43:43 +00:00
driver-source = $(src)/main.cpp
build everything (including classes) in options-specific directory Previously, we only built platform specific code in an options-specific directory (e.g. build/linux-x86_64-debug), but built the Java classes in the shared, top-level build directory. This minimized duplication when building with different sets of options, but now that we're supporting the option of selecting a third-party class library, we need to put the classes in the options-specific directory as well.
2010-09-17 22:03:37 +00:00
driver-object = $(build)/main.o
implement -jar option in main.cpp
2010-11-27 21:46:07 +00:00
driver-dynamic-objects = \
fix GCC 4.6 LTO build On Ubuntu 11.10, the optimized build was breaking, apparently because it was eliminating most of the symbols defined in assembly code (e.g. vmJump) as unreachable when linking libjvm.so, which left avian-dynamic unlinkable due to an unresolved symbol. The solution in this commit is to export makeSystem and makeFinder from libjvm.so rather than build redundant versions of finder.cpp and posix.cpp/windows.cpp into avian-dynamic like we've been doing. This avoids the whole problem of vmJump reachability and reduces the size of avian-dynamic at the same time. This commit also turns off LTO for the avian-dynamic link since we get odd undefined symbol errors about libc-defined symbols otherwise. This may merit future investigation, but avian-dynamic is so small and simple that there's no need to optimize it anyway.
2011-11-20 02:26:25 +00:00
$(build)/main-dynamic.o
rewrite main.cpp to use jni.h interface for creating the VM and invoking the main method of the specified class
2007-10-25 15:04:13 +00:00
support building Avian as a self-contained dynamic library
2008-03-31 03:43:43 +00:00
boot-source = $(src)/boot.cpp
build everything (including classes) in options-specific directory Previously, we only built platform specific code in an options-specific directory (e.g. build/linux-x86_64-debug), but built the Java classes in the shared, top-level build directory. This minimized duplication when building with different sets of options, but now that we're supporting the option of selecting a third-party class library, we need to put the classes in the options-specific directory as well.
2010-09-17 22:03:37 +00:00
boot-object = $(build)/boot.o
rewrite main.cpp to use jni.h interface for creating the VM and invoking the main method of the specified class
2007-10-25 15:04:13 +00:00
rework OpenJDK build to derive classpath and library path from environment We now consult the JAVA_HOME environment variable to determine where to find the system library JARs and SOs. Ultimately, we'll want to support self-contained build, but this allows Avian to behave like a conventional libjvm.so.
2010-09-20 23:31:23 +00:00
generator-depends := $(wildcard $(src)/*.h)
generator-sources = \
move stream.h to include, and type-generator to src/tools
2013-02-21 04:26:34 +00:00
$(src)/tools/type-generator/main.cpp \
remove vm from code paths, to match namespacing
2014-02-07 21:24:56 +00:00
$(src)/system/$(build-system).cpp \
split avian::system::crash into its own file
2014-07-13 01:05:58 +00:00
$(wildcard $(src)/system/$(build-system)/*.cpp) \
use arg-parser in type-generator
2014-05-29 03:34:58 +00:00
$(src)/finder.cpp \
$(src)/util/arg-parser.cpp
add optional LZMA support for compressing embedded JARs, boot images, and shared objects
2012-06-02 15:06:22 +00:00
ifneq ($(lzma),)
fix lzma build with clang Clang doesn't like seeing -std=c++0x when building C code, and many of the options in cflags don't apply to the LZMA source code anyway, so we give LZMA its own set of compiler flags.
2014-07-28 20:26:33 +00:00
common-cflags += -I$(lzma) -DAVIAN_USE_LZMA
add optional LZMA support for compressing embedded JARs, boot images, and shared objects
2012-06-02 15:06:22 +00:00
vm-sources += \
$(src)/lzma-decode.cpp
generator-sources += \
$(src)/lzma-decode.cpp
lzma-decode-sources = \
$(lzma)/C/LzmaDec.c
lzma-decode-objects = \
$(call c-objects,$(lzma-decode-sources),$(lzma)/C,$(build))
lzma-encode-sources = \
$(lzma)/C/LzmaEnc.c \
$(lzma)/C/LzFind.c
lzma-encode-objects = \
$(call c-objects,$(lzma-encode-sources),$(lzma)/C,$(build))
lzma-encoder = $(build)/lzma/lzma
fix crash in Posix_getaddrinfo for ProGuarded Android build Posix_getaddrinfo needs to access fields in libcore.io.StructAddrinfo via JNI, so we tell ProGuard to preserve them. This commit also includes a minor indentation tweek in README.md and removes -fno-rtti from lzma-build-cflags to avoid a warning from GCC.
2014-08-07 16:11:45 +00:00
lzma-build-cflags = -D_7ZIP_ST -D__STDC_CONSTANT_MACROS \
-fno-exceptions -fPIC -I$(lzma)/C
fix lzma build with clang Clang doesn't like seeing -std=c++0x when building C code, and many of the options in cflags don't apply to the LZMA source code anyway, so we give LZMA its own set of compiler flags.
2014-07-28 20:26:33 +00:00
lzma-cflags = $(lzma-build-cflags) $(classpath-extra-cflags)
add optional LZMA support for compressing embedded JARs, boot images, and shared objects
2012-06-02 15:06:22 +00:00
lzma-encoder-sources = \
$(src)/lzma/main.cpp
lzma-encoder-objects = \
$(call cpp-objects,$(lzma-encoder-sources),$(src),$(build))
lzma-encoder-lzma-sources = $(lzma-encode-sources) $(lzma-decode-sources)
lzma-encoder-lzma-objects = \
add support for LZMA on Windows
2012-06-02 21:43:42 +00:00
$(call generator-c-objects,$(lzma-encoder-lzma-sources),$(lzma)/C,$(build))
add optional LZMA support for compressing embedded JARs, boot images, and shared objects
2012-06-02 15:06:22 +00:00
lzma-loader = $(build)/lzma/load.o
expand support for publishing LZMA artifacts
2015-04-07 22:55:45 +00:00
lzma-library = $(build)/libavian-lzma.a
add optional LZMA support for compressing embedded JARs, boot images, and shared objects
2012-06-02 15:06:22 +00:00
endif
rework OpenJDK build to derive classpath and library path from environment We now consult the JAVA_HOME environment variable to determine where to find the system library JARs and SOs. Ultimately, we'll want to support self-contained build, but this allows Avian to behave like a conventional libjvm.so.
2010-09-20 23:31:23 +00:00
generator-cpp-objects = \
$(foreach x,$(1),$(patsubst $(2)/%.cpp,$(3)/%-build.o,$(x)))
add optional LZMA support for compressing embedded JARs, boot images, and shared objects
2012-06-02 15:06:22 +00:00
generator-c-objects = \
$(foreach x,$(1),$(patsubst $(2)/%.c,$(3)/%-build.o,$(x)))
support classloading from jar files compiled into the executable
2007-10-25 22:06:05 +00:00
generator-objects = \
rework OpenJDK build to derive classpath and library path from environment We now consult the JAVA_HOME environment variable to determine where to find the system library JARs and SOs. Ultimately, we'll want to support self-contained build, but this allows Avian to behave like a conventional libjvm.so.
2010-09-20 23:31:23 +00:00
$(call generator-cpp-objects,$(generator-sources),$(src),$(build))
add optional LZMA support for compressing embedded JARs, boot images, and shared objects
2012-06-02 15:06:22 +00:00
generator-lzma-objects = \
$(call generator-c-objects,$(lzma-decode-sources),$(lzma)/C,$(build))
build everything (including classes) in options-specific directory Previously, we only built platform specific code in an options-specific directory (e.g. build/linux-x86_64-debug), but built the Java classes in the shared, top-level build directory. This minimized duplication when building with different sets of options, but now that we're supporting the option of selecting a third-party class library, we need to put the classes in the options-specific directory as well.
2010-09-17 22:03:37 +00:00
generator = $(build)/generator
copy makefile, stdc++.cpp, and type-generator.cpp from mess project and began customizing them for this project
2007-06-03 23:16:07 +00:00
move binary-to-object to new tree structure
2013-02-20 03:23:11 +00:00
all-depends = $(shell find include -name '*.h')
refactor binaryToObject to allow more flexibilty (in particular, allowing arbitrary symbols per object)
2012-04-27 18:08:44 +00:00
move binary-to-object to new tree structure
2013-02-20 03:23:11 +00:00
object-writer-depends = $(shell find $(src)/tools/object-writer -name '*.h')
object-writer-sources = $(shell find $(src)/tools/object-writer -name '*.cpp')
object-writer-objects = $(call cpp-objects,$(object-writer-sources),$(src),$(build))
improve build for binaryToObject
2012-04-25 15:09:26 +00:00
move binary-to-object to new tree structure
2013-02-20 03:23:11 +00:00
binary-to-object-depends = $(shell find $(src)/tools/binary-to-object/ -name '*.h')
binary-to-object-sources = $(shell find $(src)/tools/binary-to-object/ -name '*.cpp')
binary-to-object-objects = $(call cpp-objects,$(binary-to-object-sources),$(src),$(build))
converter-sources = $(object-writer-sources)
converter-tool-depends = $(binary-to-object-depends) $(all-depends)
converter-tool-sources = $(binary-to-object-sources)
directly emit codeimage as a object (binaryToObject is statically linked in), as a stepping stone to including extra symbols in said codeimage
2012-05-01 19:16:32 +00:00
improve build for binaryToObject
2012-04-25 15:09:26 +00:00
converter-objects = $(call cpp-objects,$(converter-sources),$(src),$(build))
directly emit codeimage as a object (binaryToObject is statically linked in), as a stepping stone to including extra symbols in said codeimage
2012-05-01 19:16:32 +00:00
converter-tool-objects = $(call cpp-objects,$(converter-tool-sources),$(src),$(build))
improve build for binaryToObject
2012-04-25 15:09:26 +00:00
converter = $(build)/binaryToObject/binaryToObject
build everything (including classes) in options-specific directory Previously, we only built platform specific code in an options-specific directory (e.g. build/linux-x86_64-debug), but built the Java classes in the shared, top-level build directory. This minimized duplication when building with different sets of options, but now that we're supporting the option of selecting a third-party class library, we need to put the classes in the options-specific directory as well.
2010-09-17 22:03:37 +00:00
Allow output of exceptions to debugger. Generate WinMD file
2013-01-30 08:42:05 +00:00
static-library = $(build)/$(static-prefix)$(name)$(static-suffix)
build everything (including classes) in options-specific directory Previously, we only built platform specific code in an options-specific directory (e.g. build/linux-x86_64-debug), but built the Java classes in the shared, top-level build directory. This minimized duplication when building with different sets of options, but now that we're supporting the option of selecting a third-party class library, we need to put the classes in the options-specific directory as well.
2010-09-17 22:03:37 +00:00
executable = $(build)/$(name)${exe-suffix}
dynamic-library = $(build)/$(so-prefix)jvm$(so-suffix)
Allow output of exceptions to debugger. Generate WinMD file
2013-01-30 08:42:05 +00:00
executable-dynamic = $(build)/$(name)-dynamic$(exe-suffix)
stack trace work
2007-06-30 02:39:01 +00:00
add unittest framework
2013-02-15 03:54:20 +00:00
unittest-executable = $(build)/$(name)-unittest${exe-suffix}
preliminary support for using OpenJDK's class library Whereas the GNU Classpath port used the strategy of patching Classpath with core classes from Avian so as to minimize changes to the VM, this port uses the opposite strategy: abstract and isolate classpath-specific features in the VM similar to how we abstract away platform-specific features in system.h. This allows us to use an unmodified copy of OpenJDK's class library, including its core classes and augmented by a few VM-specific classes in the "avian" package.
2010-09-10 21:05:29 +00:00
ifneq ($(classpath),avian)
fix Subroutine test for use with OpenJDK build
2011-02-17 21:28:51 +00:00
# Assembler, ConstantPool, and Stream are not technically needed for a
# working build, but we include them since our Subroutine test uses
# them to synthesize a class:
preliminary support for using OpenJDK's class library Whereas the GNU Classpath port used the strategy of patching Classpath with core classes from Avian so as to minimize changes to the VM, this port uses the opposite strategy: abstract and isolate classpath-specific features in the VM similar to how we abstract away platform-specific features in system.h. This allows us to use an unmodified copy of OpenJDK's class library, including its core classes and augmented by a few VM-specific classes in the "avian" package.
2010-09-10 21:05:29 +00:00
classpath-sources := \
fix Subroutine test for use with OpenJDK build
2011-02-17 21:28:51 +00:00
$(classpath-src)/avian/Addendum.java \
more Android class library work
2013-02-21 22:37:17 +00:00
$(classpath-src)/avian/AnnotationInvocationHandler.java \
fix Subroutine test for use with OpenJDK build
2011-02-17 21:28:51 +00:00
$(classpath-src)/avian/Assembler.java \
preliminary support for using OpenJDK's class library Whereas the GNU Classpath port used the strategy of patching Classpath with core classes from Avian so as to minimize changes to the VM, this port uses the opposite strategy: abstract and isolate classpath-specific features in the VM similar to how we abstract away platform-specific features in system.h. This allows us to use an unmodified copy of OpenJDK's class library, including its core classes and augmented by a few VM-specific classes in the "avian" package.
2010-09-10 21:05:29 +00:00
$(classpath-src)/avian/Callback.java \
fix OpenJDK and Android build regression c2bfba9 introduced a regression such that building against a non-Avian class library failed due to avian.Cell not being added to the library. Since avian.Continuations depends on that class, the build broke.
2014-03-25 21:58:46 +00:00
$(classpath-src)/avian/Cell.java \
fix Subroutine test for use with OpenJDK build
2011-02-17 21:28:51 +00:00
$(classpath-src)/avian/ClassAddendum.java \
implement JVM_ResolveClass This just reuses code in avian.Classes which was formerly only used in the Avian classpath build.
2011-03-27 05:24:48 +00:00
$(classpath-src)/avian/Classes.java \
fix all the bugs So there I was, planning to just fix one little bug: Thread.holdsLock and Thread.yield were missing for the Android class library. Easy enough, right? So, I added a test, got it passing, and figured I'd go ahead and run ci.sh with all three class libraries. Big mistake. Here's the stuff I found: * minor inconsistency in README.md about OpenSSL version * untested, broken Class.getEnclosingMethod (reported by Josh) * JNI test failed for tails=true Android build * Runtime.nativeExit missing for Android build * obsolete assertion in CallEvent broke tails=true Android build * obsolete superclass field offset padding broke bootimage=true Android build * runtime annotation parsing broke bootimage=true Android build (because we couldn't modify Addendum.annotationTable for classes in the heap image) * ci.sh tried building with both android=... and openjdk=..., which the makefile rightfully balked at Sorry this is all in a single commit; I didn't expect so many unrelated issues, and I'm too lazy to break them apart.
2014-07-12 22:03:11 +00:00
$(classpath-src)/avian/Code.java \
fix Subroutine test for use with OpenJDK build
2011-02-17 21:28:51 +00:00
$(classpath-src)/avian/ConstantPool.java \
$(classpath-src)/avian/Continuations.java \
$(classpath-src)/avian/FieldAddendum.java \
fix openjdk and android builds when continuations=true Also, update the whitespace padding for printing test results to accommodate long names like extra.ComposableContinuations.
2014-03-18 15:49:00 +00:00
$(classpath-src)/avian/Function.java \
preliminary support for using OpenJDK's class library Whereas the GNU Classpath port used the strategy of patching Classpath with core classes from Avian so as to minimize changes to the VM, this port uses the opposite strategy: abstract and isolate classpath-specific features in the VM similar to how we abstract away platform-specific features in system.h. This allows us to use an unmodified copy of OpenJDK's class library, including its core classes and augmented by a few VM-specific classes in the "avian" package.
2010-09-10 21:05:29 +00:00
$(classpath-src)/avian/IncompatibleContinuationException.java \
fix all the bugs So there I was, planning to just fix one little bug: Thread.holdsLock and Thread.yield were missing for the Android class library. Easy enough, right? So, I added a test, got it passing, and figured I'd go ahead and run ci.sh with all three class libraries. Big mistake. Here's the stuff I found: * minor inconsistency in README.md about OpenSSL version * untested, broken Class.getEnclosingMethod (reported by Josh) * JNI test failed for tails=true Android build * Runtime.nativeExit missing for Android build * obsolete assertion in CallEvent broke tails=true Android build * obsolete superclass field offset padding broke bootimage=true Android build * runtime annotation parsing broke bootimage=true Android build (because we couldn't modify Addendum.annotationTable for classes in the heap image) * ci.sh tried building with both android=... and openjdk=..., which the makefile rightfully balked at Sorry this is all in a single commit; I didn't expect so many unrelated issues, and I'm too lazy to break them apart.
2014-07-12 22:03:11 +00:00
$(classpath-src)/avian/InnerClassReference.java \
preliminary support for using OpenJDK's class library Whereas the GNU Classpath port used the strategy of patching Classpath with core classes from Avian so as to minimize changes to the VM, this port uses the opposite strategy: abstract and isolate classpath-specific features in the VM similar to how we abstract away platform-specific features in system.h. This allows us to use an unmodified copy of OpenJDK's class library, including its core classes and augmented by a few VM-specific classes in the "avian" package.
2010-09-10 21:05:29 +00:00
$(classpath-src)/avian/Machine.java \
$(classpath-src)/avian/MethodAddendum.java \
fix all the bugs So there I was, planning to just fix one little bug: Thread.holdsLock and Thread.yield were missing for the Android class library. Easy enough, right? So, I added a test, got it passing, and figured I'd go ahead and run ci.sh with all three class libraries. Big mistake. Here's the stuff I found: * minor inconsistency in README.md about OpenSSL version * untested, broken Class.getEnclosingMethod (reported by Josh) * JNI test failed for tails=true Android build * Runtime.nativeExit missing for Android build * obsolete assertion in CallEvent broke tails=true Android build * obsolete superclass field offset padding broke bootimage=true Android build * runtime annotation parsing broke bootimage=true Android build (because we couldn't modify Addendum.annotationTable for classes in the heap image) * ci.sh tried building with both android=... and openjdk=..., which the makefile rightfully balked at Sorry this is all in a single commit; I didn't expect so many unrelated issues, and I'm too lazy to break them apart.
2014-07-12 22:03:11 +00:00
$(classpath-src)/avian/Pair.java \
implement JVM_ResolveClass This just reuses code in avian.Classes which was formerly only used in the Avian classpath build.
2011-03-27 05:24:48 +00:00
$(classpath-src)/avian/Singleton.java \
fix Subroutine test for use with OpenJDK build
2011-02-17 21:28:51 +00:00
$(classpath-src)/avian/Stream.java \
$(classpath-src)/avian/SystemClassLoader.java \
add avian.trace.port property Setting this property (e.g. -Davian.trace.port=5555) will cause the VM to start an extra daemon thread which listens on the specified TCP port for incoming connections and dumps stack traces for all running threads to that socket. You can retrieve that dump using e.g. netcat: nc localhost 5555
2013-03-08 21:47:27 +00:00
$(classpath-src)/avian/Traces.java \
preliminary support for using OpenJDK's class library Whereas the GNU Classpath port used the strategy of patching Classpath with core classes from Avian so as to minimize changes to the VM, this port uses the opposite strategy: abstract and isolate classpath-specific features in the VM similar to how we abstract away platform-specific features in system.h. This allows us to use an unmodified copy of OpenJDK's class library, including its core classes and augmented by a few VM-specific classes in the "avian" package.
2010-09-10 21:05:29 +00:00
$(classpath-src)/avian/VMClass.java \
$(classpath-src)/avian/VMField.java \
$(classpath-src)/avian/VMMethod.java \
fix iOS OpenJDK and simulator builds With the iOS openjdk-src build, sun.misc.Launcher will throw an InternalError if avian.file.Handler is not present, so let's make it happy. Also, as of recent version(s) of Xcode, there are no utilities (e.g. ar and strip) in iPhoneSimulator.platform/usr/bin, so we can just use the system ones.
2014-05-19 01:23:21 +00:00
$(classpath-src)/avian/avianvmresource/Handler.java \
add support for openjdk=$JDK8_HOME All tests pass for the process=compile build. Next step: process=interpret.
2015-08-06 19:24:06 +00:00
$(classpath-src)/avian/file/Handler.java \
$(classpath-src)/java/lang/invoke/MethodHandle.java \
$(classpath-src)/java/lang/invoke/MethodHandles.java \
$(classpath-src)/java/lang/invoke/MethodType.java \
$(classpath-src)/java/lang/invoke/LambdaMetafactory.java \
allow building with Java 7
2015-08-06 23:23:14 +00:00
$(classpath-src)/java/lang/invoke/LambdaConversionException.java \
add support for openjdk=$JDK8_HOME All tests pass for the process=compile build. Next step: process=interpret.
2015-08-06 19:24:06 +00:00
$(classpath-src)/java/lang/invoke/CallSite.java
implement more JVM_* methods and avoid duplicate array class loading
2010-09-15 00:52:57 +00:00
move OpenJDK.getProtectionDomain into Classes.java This way, apps can access the CodeSource of a class whether they're using the OpenJDK class library or the Avian one.
2014-02-26 20:39:05 +00:00
ifeq ($(openjdk),)
first pass at Android classpath support Stuff compiles, but linking breaks spectacularly. Next step is to figure out how to build the dependencies without checking out and building the entire Android platform.
2013-02-19 16:36:19 +00:00
classpath-sources := $(classpath-sources) \
update to more recent version of Android class library Lots has changed since we forked Android's libcore, so merging the latest upstream code has required extensive changes to the Avian/Android port. One big change is that we now use Avian's versions of java.lang.Object, java.lang.Class, java.lang.ClassLoader, some java.lang.reflect.* classes, etc. instead of the Android versions. The main reason is that the Android versions have become very Dex/Dalvik-specific, and since Avian is based on Java class files, not dex archives, that code doesn't make sense here. This has the side benefit that we can share more native code with classpath-avian.cpp and reduce the amount of Java/C++ code duplication.
2014-08-21 19:09:42 +00:00
$(classpath-src)/dalvik/system/BaseDexClassLoader.java \
$(classpath-src)/libcore/reflect/AnnotationAccess.java \
first pass at Android classpath support Stuff compiles, but linking breaks spectacularly. Next step is to figure out how to build the dependencies without checking out and building the entire Android platform.
2013-02-19 16:36:19 +00:00
$(classpath-src)/sun/reflect/ConstantPool.java \
progress towards Android classpath support It now builds and links, but fails at runtime because register_libcore_icu_ICU can't find the file it wants. We'll probably need to replace register_libcore_icu_ICU with a better-behaved version.
2013-02-19 23:48:33 +00:00
$(classpath-src)/java/net/ProtocolFamily.java \
$(classpath-src)/java/net/StandardProtocolFamily.java \
$(classpath-src)/sun/misc/Cleaner.java \
more progress on Android class library port Hello.java works. Yay.
2013-02-20 17:22:40 +00:00
$(classpath-src)/sun/misc/Unsafe.java \
update to more recent version of Android class library Lots has changed since we forked Android's libcore, so merging the latest upstream code has required extensive changes to the Avian/Android port. One big change is that we now use Avian's versions of java.lang.Object, java.lang.Class, java.lang.ClassLoader, some java.lang.reflect.* classes, etc. instead of the Android versions. The main reason is that the Android versions have become very Dex/Dalvik-specific, and since Avian is based on Java class files, not dex archives, that code doesn't make sense here. This has the side benefit that we can share more native code with classpath-avian.cpp and reduce the amount of Java/C++ code duplication.
2014-08-21 19:09:42 +00:00
$(classpath-src)/java/lang/Object.java \
$(classpath-src)/java/lang/Class.java \
$(classpath-src)/java/lang/ClassLoader.java \
$(classpath-src)/java/lang/Package.java \
$(classpath-src)/java/lang/reflect/Proxy.java \
$(classpath-src)/java/lang/reflect/Field.java \
$(classpath-src)/java/lang/reflect/SignatureParser.java \
$(classpath-src)/java/lang/reflect/Constructor.java \
$(classpath-src)/java/lang/reflect/AccessibleObject.java \
$(classpath-src)/java/lang/reflect/Method.java
implement more JVM_* methods and avoid duplicate array class loading
2010-09-15 00:52:57 +00:00
endif
preliminary support for using OpenJDK's class library Whereas the GNU Classpath port used the strategy of patching Classpath with core classes from Avian so as to minimize changes to the VM, this port uses the opposite strategy: abstract and isolate classpath-specific features in the VM similar to how we abstract away platform-specific features in system.h. This allows us to use an unmodified copy of OpenJDK's class library, including its core classes and augmented by a few VM-specific classes in the "avian" package.
2010-09-10 21:05:29 +00:00
else
classpath-sources := $(shell find $(classpath-src) -name '*.java')
endif
support classloading from jar files compiled into the executable
2007-10-25 22:06:05 +00:00
classpath-classes = \
preliminary support for using OpenJDK's class library Whereas the GNU Classpath port used the strategy of patching Classpath with core classes from Avian so as to minimize changes to the VM, this port uses the opposite strategy: abstract and isolate classpath-specific features in the VM similar to how we abstract away platform-specific features in system.h. This allows us to use an unmodified copy of OpenJDK's class library, including its core classes and augmented by a few VM-specific classes in the "avian" package.
2010-09-10 21:05:29 +00:00
$(call java-classes,$(classpath-sources),$(classpath-src),$(classpath-build))
build everything (including classes) in options-specific directory Previously, we only built platform specific code in an options-specific directory (e.g. build/linux-x86_64-debug), but built the Java classes in the shared, top-level build directory. This minimized duplication when building with different sets of options, but now that we're supporting the option of selecting a third-party class library, we need to put the classes in the options-specific directory as well.
2010-09-17 22:03:37 +00:00
classpath-object = $(build)/classpath-jar.o
add build option for using GNU Classpath
2009-06-03 00:55:49 +00:00
classpath-dep = $(classpath-build).dep
preliminary support for using OpenJDK's class library Whereas the GNU Classpath port used the strategy of patching Classpath with core classes from Avian so as to minimize changes to the VM, this port uses the opposite strategy: abstract and isolate classpath-specific features in the VM similar to how we abstract away platform-specific features in system.h. This allows us to use an unmodified copy of OpenJDK's class library, including its core classes and augmented by a few VM-specific classes in the "avian" package.
2010-09-10 21:05:29 +00:00
vm-classes = \
more progress on GNU Classpath compatibility
2009-06-04 23:21:42 +00:00
avian/*.class \
preliminary support for using OpenJDK's class library Whereas the GNU Classpath port used the strategy of patching Classpath with core classes from Avian so as to minimize changes to the VM, this port uses the opposite strategy: abstract and isolate classpath-specific features in the VM similar to how we abstract away platform-specific features in system.h. This allows us to use an unmodified copy of OpenJDK's class library, including its core classes and augmented by a few VM-specific classes in the "avian" package.
2010-09-10 21:05:29 +00:00
avian/resource/*.class
implement primitive testing framework and provide for GC stress testing
2007-07-16 01:03:02 +00:00
Fixing some issues with runtime annotations within avian. We were not properly converting dots to slashes internally for package names and we did not properly handle Method.getAnnotations and Method.getAnnotation(Class<T>) on methods without any annotations. Added some tests to cover these cases.
2012-05-22 20:00:31 +00:00
test-support-sources = $(shell find $(test)/avian/ -name '*.java')
allow building with Java 7
2015-08-06 23:23:14 +00:00
test-sources := $(wildcard $(test)/*.java)
ifeq (7,$(java-version))
test-sources := $(subst $(test)/InvokeDynamic.java,,$(test-sources))
endif
add JNI test to test suite
2012-08-01 16:02:55 +00:00
test-cpp-sources = $(wildcard $(test)/*.cpp)
Fixing some issues with runtime annotations within avian. We were not properly converting dots to slashes internally for package names and we did not properly handle Method.getAnnotations and Method.getAnnotation(Class<T>) on methods without any annotations. Added some tests to cover these cases.
2012-05-22 20:00:31 +00:00
test-sources += $(test-support-sources)
test-support-classes = $(call java-classes, $(test-support-sources),$(test),$(test-build))
support classloading from jar files compiled into the executable
2007-10-25 22:06:05 +00:00
test-classes = $(call java-classes,$(test-sources),$(test),$(test-build))
add JNI test to test suite
2012-08-01 16:02:55 +00:00
test-cpp-objects = $(call cpp-objects,$(test-cpp-sources),$(test),$(test-build))
put test library in same directory as avian executable This is a workaround for the fact that there's no precise equivalent to LD_LIBRARY_PATH available on Windows.
2012-08-02 18:56:15 +00:00
test-library = $(build)/$(so-prefix)test$(so-suffix)
add build option for using GNU Classpath
2009-06-03 00:55:49 +00:00
test-dep = $(test-build).dep
implement primitive testing framework and provide for GC stress testing
2007-07-16 01:03:02 +00:00
build but don't run extra tests in makefile
2009-05-18 15:17:09 +00:00
test-extra-sources = $(wildcard $(test)/extra/*.java)
test-extra-classes = \
$(call java-classes,$(test-extra-sources),$(test),$(test-build))
add build option for using GNU Classpath
2009-06-03 00:55:49 +00:00
test-extra-dep = $(test-build)-extra.dep
build but don't run extra tests in makefile
2009-05-18 15:17:09 +00:00
add unittest framework
2013-02-15 03:54:20 +00:00
unittest-sources = \
$(wildcard $(unittest)/*.cpp) \
add unit tests for arg parser
2013-02-17 19:10:18 +00:00
$(wildcard $(unittest)/util/*.cpp) \
add unittest framework
2013-02-15 03:54:20 +00:00
$(wildcard $(unittest)/codegen/*.cpp)
unittest-depends = \
$(wildcard $(unittest)/*.h)
enable use-frame-pointer=true build Also, include Continuations, Coroutines, and DynamicWind tests in test suite for continuations=true build.
2011-01-28 04:06:01 +00:00
ifeq ($(continuations),true)
continuation-tests = \
quick sketch of composable continuation implementation I've been told by knowledgeable people that it is impossible to implement composable continuations (AKA delimited continuations AKA shift/reset) in terms of call-with-current-continuation. Since I don't yet understand why that is, I figured it would help my understanding to attempt it and see how it fails.
2013-12-15 03:06:51 +00:00
extra.ComposableContinuations \
enable use-frame-pointer=true build Also, include Continuations, Coroutines, and DynamicWind tests in test suite for continuations=true build.
2011-01-28 04:06:01 +00:00
extra.Continuations \
extra.Coroutines \
extra.DynamicWind
endif
add tail call test
2011-03-15 23:51:32 +00:00
ifeq ($(tails),true)
tail-tests = \
extra.Tails
endif
fix windows build
2012-05-02 19:38:00 +00:00
ifeq ($(target-arch),i386)
cflags += -DAVIAN_TARGET_ARCH=AVIAN_ARCH_X86
endif
ifeq ($(target-arch),x86_64)
cflags += -DAVIAN_TARGET_ARCH=AVIAN_ARCH_X86_64
endif
ifeq ($(target-arch),arm)
cflags += -DAVIAN_TARGET_ARCH=AVIAN_ARCH_ARM
endif
add process=interpret support for Linux/ARM64 This makes all the tests pass for the platform=linux arch=arm64 process=interpret build. Next step: process=compile support.
2014-11-26 01:36:34 +00:00
ifeq ($(target-arch),arm64)
cflags += -DAVIAN_TARGET_ARCH=AVIAN_ARCH_ARM64
endif
clean up binary format code Linux, FreeBSD, and QNX all use ELF, so no need to distinguish between them when generating object files. To avoid confusion, I've switch from using operating system names to using binary format names where applicable.
2012-08-02 18:15:15 +00:00
ifeq ($(target-format),elf)
cflags += -DAVIAN_TARGET_FORMAT=AVIAN_FORMAT_ELF
correctly define the target platform in the case of a bootimage build
2012-05-03 14:45:17 +00:00
endif
clean up binary format code Linux, FreeBSD, and QNX all use ELF, so no need to distinguish between them when generating object files. To avoid confusion, I've switch from using operating system names to using binary format names where applicable.
2012-08-02 18:15:15 +00:00
ifeq ($(target-format),pe)
cflags += -DAVIAN_TARGET_FORMAT=AVIAN_FORMAT_PE
correctly define the target platform in the case of a bootimage build
2012-05-03 14:45:17 +00:00
endif
clean up binary format code Linux, FreeBSD, and QNX all use ELF, so no need to distinguish between them when generating object files. To avoid confusion, I've switch from using operating system names to using binary format names where applicable.
2012-08-02 18:15:15 +00:00
ifeq ($(target-format),macho)
cflags += -DAVIAN_TARGET_FORMAT=AVIAN_FORMAT_MACHO
add support for FreeBSD
2012-08-02 16:36:16 +00:00
endif
support classloading from jar files compiled into the executable
2007-10-25 22:06:05 +00:00
class-name = $(patsubst $(1)/%.class,%,$(2))
implement fixed object support
2007-10-28 01:54:30 +00:00
class-names = $(foreach x,$(2),$(call class-name,$(1),$(x)))
implement primitive testing framework and provide for GC stress testing
2007-07-16 01:03:02 +00:00
fix Misc test failures The Misc test was failing when run as "make input=Misc run" since test-flags did not include $(build)/extra-dir in the class library, leading the ClassLoader.getResources test to fail. Also, the UnknownHostException test was not reliable -- some ISPs (mine included) return DNS matches for bogus hostnames, defaulting to the IP address of a webserver intended to help users with name resolution problems. That's dumb, I know, but I'm guessing I'm not the only person with a dumb ISP, and it seems better to just remove the test than make people think Avian is broken when it's really just their DNS server that's broken.
2014-04-21 01:11:15 +00:00
test-flags = -Djava.library.path=$(build) \
fix test-flags definition for Windows We need to quote a classpath with a Windows path separator (';') so it isn't interpreted by the shell.
2014-05-09 22:37:17 +00:00
-cp '$(build)/test$(target-path-separator)$(build)/extra-dir'
include boot classpath in vm arguments on darwin, since we don't build those classes into the executable
2007-12-27 15:46:17 +00:00
rework OpenJDK build to derive classpath and library path from environment We now consult the JAVA_HOME environment variable to determine where to find the system library JARs and SOs. Ultimately, we'll want to support self-contained build, but this allows Avian to behave like a conventional libjvm.so.
2010-09-20 23:31:23 +00:00
test-args = $(test-flags) $(input)
lots of bugfixes
2007-06-21 01:38:02 +00:00
Added targets for Eclipse environment descriptor. Someone on the Google Groups site asked how to set up an Eclipse project with Avian classpath. This patch creates the descriptor that you can you use to do that at `$(build)/eclipse/jdk/avian.ee`. The descriptor includes the Avian version, platform, architecture, and build options to allow for multiple versions to exist side by side. Users can import the descriptor into Eclipse via: Window >> Preferences >> Java >> Installed JREs >> Add >> Execution Environment Description Once the descriptor is imported, Avian can be used just like any other JVM installation for Eclipse projects. Personally I use this in conjunction with Eclim to gain code completion for Avian in vim. The new targets also create symlinks to loosely mimic OpenJDK's filenames and folder layout: build/linux-x86_64-tails-continuations/eclipse/jdk/ ├── avian.ee ├── bin │   └── java -> ../../../avian ├── jre │   └── lib │   └── rt.jar -> ../../../../classpath.jar └── src -> ../../classpath Annoyingly, Eclipse for some reason expects this layout to exist even though the descriptor format has required parameters for specifying these locations. I suppose that other software may look for this "standard" layout in a JVM installation so it may be generally useful. These artifacts are only built if the platform is one of `windows`, `linux`, or `macosx`. The symlinks might not actually work at all on Windows, I'm not sure how things like cygwin/msys handle that and I do not have the means to test it. If they do not work a fallback for windows might be to actually copy the files instead of symlinking. I realize this can be done outside of the makefile but it seemed useful to put it here to gain access to the information about the build location, platform, architecture, and other build options. For the record, this contribution is my original work and is released under the same license that Avian uses, found in the license.txt file in this repository.
2015-07-16 13:44:55 +00:00
ifneq ($(filter linux windows macosx,$(platform)),)
eclipse-exec-env = eclipse-ee
eclipse-jdk-dir = $(build)/eclipse/jdk
eclipse-ee-file = $(eclipse-jdk-dir)/avian.ee
eclipse-bin-dir = $(eclipse-jdk-dir)/bin
eclipse-lib-dir = $(eclipse-jdk-dir)/jre/lib
eclipse-src-dir = $(eclipse-jdk-dir)/src
define eclipse-ee-descriptor
# An Eclipse execution environment for the Avian JVM\
\n-Dee.executable=bin/java${exe-suffix}\
\n-Dee.bootclasspath=jre/lib/rt.jar\
\n-Dee.language.level=1.7\
\n-Dee.name=$(name)-$(version)-$(platform)-$(arch)$(options)\
\n-Dee.src=src\
\n-Dee.javadoc=file://$${ee.home}/doc\
\n-Djava.home=$${ee.home}\n
endef
else
eclipse-exec-env =
endif
copy makefile, stdc++.cpp, and type-generator.cpp from mess project and began customizing them for this project
2007-06-03 23:16:07 +00:00
.PHONY: build
Windows Phone 8 / Windows RT initial support Conflicts: makefile
2013-01-28 15:20:52 +00:00
ifneq ($(supports_avian_executable),false)
expand support for publishing LZMA artifacts
2015-04-07 22:55:45 +00:00
build: $(static-library) $(executable) $(dynamic-library) $(lzma-library) \
make LZMA encoder by default
2012-06-02 19:23:52 +00:00
$(lzma-encoder) $(executable-dynamic) $(classpath-dep) $(test-dep) \
Added targets for Eclipse environment descriptor. Someone on the Google Groups site asked how to set up an Eclipse project with Avian classpath. This patch creates the descriptor that you can you use to do that at `$(build)/eclipse/jdk/avian.ee`. The descriptor includes the Avian version, platform, architecture, and build options to allow for multiple versions to exist side by side. Users can import the descriptor into Eclipse via: Window >> Preferences >> Java >> Installed JREs >> Add >> Execution Environment Description Once the descriptor is imported, Avian can be used just like any other JVM installation for Eclipse projects. Personally I use this in conjunction with Eclim to gain code completion for Avian in vim. The new targets also create symlinks to loosely mimic OpenJDK's filenames and folder layout: build/linux-x86_64-tails-continuations/eclipse/jdk/ ├── avian.ee ├── bin │   └── java -> ../../../avian ├── jre │   └── lib │   └── rt.jar -> ../../../../classpath.jar └── src -> ../../classpath Annoyingly, Eclipse for some reason expects this layout to exist even though the descriptor format has required parameters for specifying these locations. I suppose that other software may look for this "standard" layout in a JVM installation so it may be generally useful. These artifacts are only built if the platform is one of `windows`, `linux`, or `macosx`. The symlinks might not actually work at all on Windows, I'm not sure how things like cygwin/msys handle that and I do not have the means to test it. If they do not work a fallback for windows might be to actually copy the files instead of symlinking. I realize this can be done outside of the makefile but it seemed useful to put it here to gain access to the information about the build location, platform, architecture, and other build options. For the record, this contribution is my original work and is released under the same license that Avian uses, found in the license.txt file in this repository.
2015-07-16 13:44:55 +00:00
$(test-extra-dep) $(embed) $(build)/classpath.jar $(eclipse-exec-env)
Windows Phone 8 / Windows RT initial support Conflicts: makefile
2013-01-28 15:20:52 +00:00
else
expand support for publishing LZMA artifacts
2015-04-07 22:55:45 +00:00
build: $(static-library) $(dynamic-library) $(lzma-library) \
Windows Phone 8 / Windows RT initial support Conflicts: makefile
2013-01-28 15:20:52 +00:00
$(lzma-encoder) $(classpath-dep) $(test-dep) \
fix LZMA/bootimage build This includes some tweaks to README.md and makefile to ensure the directions in the former for building a bootimage actually work.
2014-07-28 15:19:39 +00:00
$(test-extra-dep) $(embed) $(build)/classpath.jar
Windows Phone 8 / Windows RT initial support Conflicts: makefile
2013-01-28 15:20:52 +00:00
endif
copy makefile, stdc++.cpp, and type-generator.cpp from mess project and began customizing them for this project
2007-06-03 23:16:07 +00:00
fix parallel (make -jN) build
2009-09-05 22:04:43 +00:00
$(test-dep): $(classpath-dep)
$(test-extra-dep): $(classpath-dep)
lots of bugfixes
2007-06-21 01:38:02 +00:00
copy makefile, stdc++.cpp, and type-generator.cpp from mess project and began customizing them for this project
2007-06-03 23:16:07 +00:00
.PHONY: run
fixes for various recent regressions
2007-09-30 16:32:17 +00:00
run: build
enable standalone OpenJDK builds As described in readme.txt, a standalone OpenJDK build embeds all libraries, classes, and other files needed at runtime in the resulting binary, eliminating dependencies on external resources.
2010-11-04 17:02:09 +00:00
$(library-path) $(test-executable) $(test-args)
copy makefile, stdc++.cpp, and type-generator.cpp from mess project and began customizing them for this project
2007-06-03 23:16:07 +00:00
.PHONY: debug
fixes for various recent regressions
2007-09-30 16:32:17 +00:00
debug: build
fix parameterization of debugger
2014-05-07 19:01:53 +00:00
$(library-path) $(db) $(test-executable) $(test-args)
copy makefile, stdc++.cpp, and type-generator.cpp from mess project and began customizing them for this project
2007-06-03 23:16:07 +00:00
.PHONY: vg
fixes for various recent regressions
2007-09-30 16:32:17 +00:00
vg: build
enable standalone OpenJDK builds As described in readme.txt, a standalone OpenJDK build embeds all libraries, classes, and other files needed at runtime in the resulting binary, eliminating dependencies on external resources.
2010-11-04 17:02:09 +00:00
$(library-path) $(vg) $(test-executable) $(test-args)
copy makefile, stdc++.cpp, and type-generator.cpp from mess project and began customizing them for this project
2007-06-03 23:16:07 +00:00
.PHONY: test
assume we can't use ARMv7 instructions on non-iOS platforms by default The only Linux/ARM machine I have access to does not support ARMv7. I don't know how common that is in general, but this seems like the safe default. You can always override it on the command line. I've also broken the test target into build-test and run-test subtargets, which can be useful when you're building on a shared filesystem and running the tests on another machine (without using the remote-test-host option). If there's clock skew, the other machine might try to rebuild stuff unecessarily. Using run-test avoids that.
2014-04-22 17:09:41 +00:00
test: build-test run-test
.PHONY: build-test
build-test: build $(build)/run-tests.sh $(build)/test.sh $(unittest-executable)
.PHONY: run-test
run-test:
add remote-test capability To execute tests on a remote host (for instance, because you're cross-compiling), simply do: make remote-test=true remote-test-host=<host_to_test_on> test You can set several variables to control the functionality of remote-test. See them below, along with their default values: remote-test-host = localhost # host to ssh to remote-test-port = 22 remote-test-user = ${USER} # user to execute tests as remote-test-dir = /tmp/avian-test-${USER} # dir to rsync build output to
2013-02-15 00:21:57 +00:00
ifneq ($(remote-test),true)
/bin/sh $(build)/run-tests.sh
else
set remote-test variable based on the presence of remote-test-host or remote-test-port The new way to run a remote test is: make arch=<arch> remote-test-host=<host_to_test_on> test
2013-02-16 00:43:21 +00:00
@echo "running tests on $(remote-test-user)@$(remote-test-host):$(remote-test-port), in $(remote-test-dir)"
add remote-test capability To execute tests on a remote host (for instance, because you're cross-compiling), simply do: make remote-test=true remote-test-host=<host_to_test_on> test You can set several variables to control the functionality of remote-test. See them below, along with their default values: remote-test-host = localhost # host to ssh to remote-test-port = 22 remote-test-user = ${USER} # user to execute tests as remote-test-dir = /tmp/avian-test-${USER} # dir to rsync build output to
2013-02-15 00:21:57 +00:00
rsync $(build) -rav --exclude '*.o' --rsh="ssh -p$(remote-test-port)" $(remote-test-user)@$(remote-test-host):$(remote-test-dir)
ssh -p$(remote-test-port) $(remote-test-user)@$(remote-test-host) sh "$(remote-test-dir)/$(platform)-$(arch)$(options)/run-tests.sh"
endif
copy makefile, stdc++.cpp, and type-generator.cpp from mess project and began customizing them for this project
2007-06-03 23:16:07 +00:00
add jdk-test target, and fix failures The intent of this target is to run our test suite against the installed jre. This should help prevent our VM from diverging in implementation from the jdk. The remainder of this commit fixes the problems that this exposes.
2013-12-06 21:56:02 +00:00
.PHONY: jdk-test
fix jdk-test failures
2013-12-07 02:30:04 +00:00
jdk-test: $(test-dep) $(build)/classpath.jar $(build)/jdk-run-tests.sh $(build)/test.sh
add jdk-test target, and fix failures The intent of this target is to run our test suite against the installed jre. This should help prevent our VM from diverging in implementation from the jdk. The remainder of this commit fixes the problems that this exposes.
2013-12-06 21:56:02 +00:00
/bin/sh $(build)/jdk-run-tests.sh
add tarball target to makefile
2009-04-11 17:54:43 +00:00
.PHONY: tarball
tarball:
@echo "creating build/avian-$(version).tar.bz2"
@mkdir -p build
delete cmake-build and distrib in clean target
2015-03-13 21:21:29 +00:00
(cd .. && tar --exclude=build --exclude=cmake-build --exclude=distrib \
delete lib dir in clean target
2015-03-16 16:39:11 +00:00
--exclude=lib --exclude='.*' --exclude='*~' \
delete cmake-build and distrib in clean target
2015-03-13 21:21:29 +00:00
-cjf avian/build/avian-$(version).tar.bz2 avian)
add tarball target to makefile
2009-04-11 17:54:43 +00:00
rename clean to clean-current, and clean-all back to clean This way, the clean target continues to do what it always did: delete the whole build directory. You can use clean-current to just delete the currently-configured build directory.
2013-02-06 18:38:30 +00:00
.PHONY: clean-current
clean-current:
Make clean target more robust Conflicts: .gitignore
2013-02-03 10:27:30 +00:00
@echo "removing $(build)"
rm -rf $(build)
copy makefile, stdc++.cpp, and type-generator.cpp from mess project and began customizing them for this project
2007-06-03 23:16:07 +00:00
.PHONY: clean
clean:
delete lib dir in clean target
2015-03-16 16:39:11 +00:00
@echo "removing build directories"
rm -rf build cmake-build distrib lib
copy makefile, stdc++.cpp, and type-generator.cpp from mess project and began customizing them for this project
2007-06-03 23:16:07 +00:00
Added targets for Eclipse environment descriptor. Someone on the Google Groups site asked how to set up an Eclipse project with Avian classpath. This patch creates the descriptor that you can you use to do that at `$(build)/eclipse/jdk/avian.ee`. The descriptor includes the Avian version, platform, architecture, and build options to allow for multiple versions to exist side by side. Users can import the descriptor into Eclipse via: Window >> Preferences >> Java >> Installed JREs >> Add >> Execution Environment Description Once the descriptor is imported, Avian can be used just like any other JVM installation for Eclipse projects. Personally I use this in conjunction with Eclim to gain code completion for Avian in vim. The new targets also create symlinks to loosely mimic OpenJDK's filenames and folder layout: build/linux-x86_64-tails-continuations/eclipse/jdk/ ├── avian.ee ├── bin │   └── java -> ../../../avian ├── jre │   └── lib │   └── rt.jar -> ../../../../classpath.jar └── src -> ../../classpath Annoyingly, Eclipse for some reason expects this layout to exist even though the descriptor format has required parameters for specifying these locations. I suppose that other software may look for this "standard" layout in a JVM installation so it may be generally useful. These artifacts are only built if the platform is one of `windows`, `linux`, or `macosx`. The symlinks might not actually work at all on Windows, I'm not sure how things like cygwin/msys handle that and I do not have the means to test it. If they do not work a fallback for windows might be to actually copy the files instead of symlinking. I realize this can be done outside of the makefile but it seemed useful to put it here to gain access to the information about the build location, platform, architecture, and other build options. For the record, this contribution is my original work and is released under the same license that Avian uses, found in the license.txt file in this repository.
2015-07-16 13:44:55 +00:00
.PHONY: eclipse-ee
ifneq ($(strip $(eclipse-exec-env)),)
eclipse-ee: $(eclipse-ee-file) $(eclipse-lib-dir)/rt.jar $(eclipse-bin-dir)/java${exe-suffix} $(eclipse-src-dir)
$(eclipse-bin-dir):
@mkdir -p $(@)
$(eclipse-lib-dir):
@mkdir -p $(@)
$(eclipse-jdk-dir):
@mkdir -p $(@)
$(eclipse-ee-file): $(eclipse-jdk-dir)
@echo "writing eclipse execution environment descriptor to $(@)"
@printf '${eclipse-ee-descriptor}' > $(@)
$(eclipse-src-dir): $(eclipse-jdk-dir)
@echo "symlinking classpath for $(@)"
Fixed src symlink in eclipse config to actually point to source.
2015-07-17 03:59:27 +00:00
@ln -sf ../../../../classpath $(@)
Added targets for Eclipse environment descriptor. Someone on the Google Groups site asked how to set up an Eclipse project with Avian classpath. This patch creates the descriptor that you can you use to do that at `$(build)/eclipse/jdk/avian.ee`. The descriptor includes the Avian version, platform, architecture, and build options to allow for multiple versions to exist side by side. Users can import the descriptor into Eclipse via: Window >> Preferences >> Java >> Installed JREs >> Add >> Execution Environment Description Once the descriptor is imported, Avian can be used just like any other JVM installation for Eclipse projects. Personally I use this in conjunction with Eclim to gain code completion for Avian in vim. The new targets also create symlinks to loosely mimic OpenJDK's filenames and folder layout: build/linux-x86_64-tails-continuations/eclipse/jdk/ ├── avian.ee ├── bin │   └── java -> ../../../avian ├── jre │   └── lib │   └── rt.jar -> ../../../../classpath.jar └── src -> ../../classpath Annoyingly, Eclipse for some reason expects this layout to exist even though the descriptor format has required parameters for specifying these locations. I suppose that other software may look for this "standard" layout in a JVM installation so it may be generally useful. These artifacts are only built if the platform is one of `windows`, `linux`, or `macosx`. The symlinks might not actually work at all on Windows, I'm not sure how things like cygwin/msys handle that and I do not have the means to test it. If they do not work a fallback for windows might be to actually copy the files instead of symlinking. I realize this can be done outside of the makefile but it seemed useful to put it here to gain access to the information about the build location, platform, architecture, and other build options. For the record, this contribution is my original work and is released under the same license that Avian uses, found in the license.txt file in this repository.
2015-07-16 13:44:55 +00:00
$(eclipse-bin-dir)/java$(exe-suffix): $(eclipse-bin-dir) $(executable)
@echo "symlinking $(executable) for $(@)"
@ln -sf ../../../$(name)${exe-suffix} $(@)
$(eclipse-lib-dir)/rt.jar: $(eclipse-lib-dir) $(build)/classpath.jar
@echo "symlinking $(build)/classpath.jar for $(@)"
@ln -sf ../../../../classpath.jar $(@)
else
eclipse-ee:
$(error "Eclipse execution environment for platform '$(platform)' is not supported")
endif
WinRT/WP8 process=compile
2013-01-29 07:40:29 +00:00
ifeq ($(continuations),true)
Windows Phone 8 / Windows RT initial support Conflicts: makefile
2013-01-28 15:20:52 +00:00
$(build)/compile-x86-asm.o: $(src)/continuations-x86.$(asm-format)
WinRT/WP8 process=compile
2013-01-29 07:40:29 +00:00
endif
various performance tweaks and bugfixes
2007-08-19 19:45:51 +00:00
Add a pseudo-integration test for getResources() This adds an extra class path element to the VM running the unit tests, writes files with identical file names into both directories and then verifies that SystemClassLoader#getResources can find them. Signed-off-by: Johannes Schindelin <johannes.schindelin@gmx.de>
2013-11-04 21:28:19 +00:00
$(build)/run-tests.sh: $(test-classes) makefile $(build)/extra-dir/multi-classpath-test.txt $(build)/test/multi-classpath-test.txt
add remote-test capability To execute tests on a remote host (for instance, because you're cross-compiling), simply do: make remote-test=true remote-test-host=<host_to_test_on> test You can set several variables to control the functionality of remote-test. See them below, along with their default values: remote-test-host = localhost # host to ssh to remote-test-port = 22 remote-test-user = ${USER} # user to execute tests as remote-test-dir = /tmp/avian-test-${USER} # dir to rsync build output to
2013-02-15 00:21:57 +00:00
echo 'cd $$(dirname $$0)' > $(@)
echo "sh ./test.sh 2>/dev/null \\" >> $(@)
fix remote-test functionality for windows
2014-02-22 22:35:12 +00:00
echo "$(shell echo $(library-path) | sed 's|$(build)|\.|g') ./$(name)-unittest${exe-suffix} ./$(notdir $(test-executable)) $(mode) \"-Djava.library.path=. -cp test$(target-path-separator)extra-dir\" \\" >> $(@)
add remote-test capability To execute tests on a remote host (for instance, because you're cross-compiling), simply do: make remote-test=true remote-test-host=<host_to_test_on> test You can set several variables to control the functionality of remote-test. See them below, along with their default values: remote-test-host = localhost # host to ssh to remote-test-port = 22 remote-test-user = ${USER} # user to execute tests as remote-test-dir = /tmp/avian-test-${USER} # dir to rsync build output to
2013-02-15 00:21:57 +00:00
echo "$(call class-names,$(test-build),$(filter-out $(test-support-classes), $(test-classes))) \\" >> $(@)
echo "$(continuation-tests) $(tail-tests)" >> $(@)
add jdk-test target, and fix failures The intent of this target is to run our test suite against the installed jre. This should help prevent our VM from diverging in implementation from the jdk. The remainder of this commit fixes the problems that this exposes.
2013-12-06 21:56:02 +00:00
$(build)/jdk-run-tests.sh: $(test-classes) makefile $(build)/extra-dir/multi-classpath-test.txt $(build)/test/multi-classpath-test.txt
echo 'cd $$(dirname $$0)' > $(@)
echo "sh ./test.sh 2>/dev/null \\" >> $(@)
use ';' instead of ':' as path separator on windows when running tests
2014-01-08 22:00:19 +00:00
echo "'' true $(JAVA_HOME)/bin/java $(mode) \"-Xmx128m -Djava.library.path=. -cp test$(path-separator)extra-dir$(path-separator)classpath\" \\" >> $(@)
add jdk-test target, and fix failures The intent of this target is to run our test suite against the installed jre. This should help prevent our VM from diverging in implementation from the jdk. The remainder of this commit fixes the problems that this exposes.
2013-12-06 21:56:02 +00:00
echo "$(call class-names,$(test-build),$(filter-out $(test-support-classes), $(test-classes))) \\" >> $(@)
echo "$(continuation-tests) $(tail-tests)" >> $(@)
Add a pseudo-integration test for getResources() This adds an extra class path element to the VM running the unit tests, writes files with identical file names into both directories and then verifies that SystemClassLoader#getResources can find them. Signed-off-by: Johannes Schindelin <johannes.schindelin@gmx.de>
2013-11-04 21:28:19 +00:00
$(build)/extra-dir/multi-classpath-test.txt:
mkdir -p $(build)/extra-dir
echo "$@" > $@
$(build)/test/multi-classpath-test.txt:
echo "$@" > $@
add remote-test capability To execute tests on a remote host (for instance, because you're cross-compiling), simply do: make remote-test=true remote-test-host=<host_to_test_on> test You can set several variables to control the functionality of remote-test. See them below, along with their default values: remote-test-host = localhost # host to ssh to remote-test-port = 22 remote-test-user = ${USER} # user to execute tests as remote-test-dir = /tmp/avian-test-${USER} # dir to rsync build output to
2013-02-15 00:21:57 +00:00
$(build)/test.sh: $(test)/test.sh
cp $(<) $(@)
build everything (including classes) in options-specific directory Previously, we only built platform specific code in an options-specific directory (e.g. build/linux-x86_64-debug), but built the Java classes in the shared, top-level build directory. This minimized duplication when building with different sets of options, but now that we're supporting the option of selecting a third-party class library, we need to put the classes in the options-specific directory as well.
2010-09-17 22:03:37 +00:00
gen-arg = $(shell echo $(1) | sed -e 's:$(build)/type-\(.*\)\.cpp:\1:')
clean up bootstrap type generation to eliminate redundancy (broken)
2007-11-04 21:15:28 +00:00
$(generated-code): %.cpp: $(src)/types.def $(generator) $(classpath-dep)
copy makefile, stdc++.cpp, and type-generator.cpp from mess project and began customizing them for this project
2007-06-03 23:16:07 +00:00
@echo "generating $(@)"
implement fixed object support
2007-10-28 01:54:30 +00:00
@mkdir -p $(dir $(@))
use arg-parser in type-generator
2014-05-29 03:34:58 +00:00
$(generator) -cp $(boot-classpath) -i $(<) -o $(@) -t $(call gen-arg,$(@))
copy makefile, stdc++.cpp, and type-generator.cpp from mess project and began customizing them for this project
2007-06-03 23:16:07 +00:00
preliminary support for using OpenJDK's class library Whereas the GNU Classpath port used the strategy of patching Classpath with core classes from Avian so as to minimize changes to the VM, this port uses the opposite strategy: abstract and isolate classpath-specific features in the VM similar to how we abstract away platform-specific features in system.h. This allows us to use an unmodified copy of OpenJDK's class library, including its core classes and augmented by a few VM-specific classes in the "avian" package.
2010-09-10 21:05:29 +00:00
$(classpath-build)/%.class: $(classpath-src)/%.java
tweak makefile to recompile all dirty java files in a single javac invocation for efficiency
2007-11-02 14:01:10 +00:00
@echo $(<)
first pass at Android classpath support Stuff compiles, but linking breaks spectacularly. Next step is to figure out how to build the dependencies without checking out and building the entire Android platform.
2013-02-19 16:36:19 +00:00
$(classpath-dep): $(classpath-sources) $(classpath-jar-dep)
tweak makefile to recompile all dirty java files in a single javac invocation for efficiency
2007-11-02 14:01:10 +00:00
@echo "compiling classpath classes"
fix rebuild when using GNU Classpath
2009-06-04 23:54:06 +00:00
@mkdir -p $(classpath-build)
various fixes to get avian-swt-examples working with the Android class library
2013-02-26 23:24:02 +00:00
classes="$(shell $(MAKE) -s --no-print-directory build=$(build) \
fix Cygwin build 80ce92a introduced a regression which caused the Cygwin build to fail, since it tried to set platform=$(build-platform) when calling make recursively, and platform=cygwin is rejected by the makefile. Instead, we need to use $(bootimage-platform), which normalizes "cygwin" and "mingw32" to "windows".
2015-01-22 16:57:15 +00:00
$(classpath-classes) arch=$(build-arch) platform=$(bootimage-platform))"; \
fix iOS ARM64 build
2014-12-31 00:03:58 +00:00
if [ -n "$${classes}" ]; then \
allow building with Java 7
2015-08-06 23:23:14 +00:00
$(javac) -source 1.$(java-version) -target 1.$(java-version) \
add process=interpret support for Linux/ARM64 This makes all the tests pass for the platform=linux arch=arm64 process=interpret build. Next step: process=compile support.
2014-11-26 01:36:34 +00:00
-d $(classpath-build) -bootclasspath $(boot-classpath) \
various fixes to get avian-swt-examples working with the Android class library
2013-02-26 23:24:02 +00:00
$${classes}; fi
support classloading from jar files compiled into the executable
2007-10-25 22:06:05 +00:00
@touch $(@)
more JNI work
2007-06-29 02:58:48 +00:00
more progress on Android class library port Hello.java works. Yay.
2013-02-20 17:22:40 +00:00
$(build)/android-src/%.cpp: $(luni-native)/%.cpp
first step towards supporting Android class library on Windows A majority of the tests are passing, but more work is needed to get them all to pass.
2013-03-14 21:33:05 +00:00
cp $(<) $(@)
more progress on Android class library port Hello.java works. Yay.
2013-02-20 17:22:40 +00:00
fix Android classpath build Hopefully it will stay fixed this time, since we now specify commit hashes for each dependency which are known to work.
2013-07-05 20:34:16 +00:00
$(build)/android-src/%.cpp: $(libnativehelper-native)/%.cpp
cp $(<) $(@)
$(build)/android-src/%.cpp: $(crypto-native)/%.cpp
cp $(<) $(@)
update to more recent version of Android class library Lots has changed since we forked Android's libcore, so merging the latest upstream code has required extensive changes to the Avian/Android port. One big change is that we now use Avian's versions of java.lang.Object, java.lang.Class, java.lang.ClassLoader, some java.lang.reflect.* classes, etc. instead of the Android versions. The main reason is that the Android versions have become very Dex/Dalvik-specific, and since Avian is based on Java class files, not dex archives, that code doesn't make sense here. This has the side benefit that we can share more native code with classpath-avian.cpp and reduce the amount of Java/C++ code duplication.
2014-08-21 19:09:42 +00:00
$(build)/android-src/%.cpp: $(libziparchive-native)/%.cc
cp $(<) $(@)
$(build)/android-src/%.cpp: $(libutils-native)/%.cpp
fix Android classpath build Hopefully it will stay fixed this time, since we now specify commit hashes for each dependency which are known to work.
2013-07-05 20:34:16 +00:00
cp $(<) $(@)
more progress on Android class library port Hello.java works. Yay.
2013-02-20 17:22:40 +00:00
$(build)/%.o: $(build)/android-src/%.cpp $(build)/android.dep
first pass at Android classpath support Stuff compiles, but linking breaks spectacularly. Next step is to figure out how to build the dependencies without checking out and building the entire Android platform.
2013-02-19 16:36:19 +00:00
@echo "compiling $(@)"
@mkdir -p $(dir $(@))
various tweaks to get hello-ios working with the Android library
2013-03-04 21:24:41 +00:00
$(cxx) $(android-cflags) $(classpath-extra-cflags) -c \
$$($(windows-path) $(<)) $(call output,$(@))
first pass at Android classpath support Stuff compiles, but linking breaks spectacularly. Next step is to figure out how to build the dependencies without checking out and building the entire Android platform.
2013-02-19 16:36:19 +00:00
update to more recent version of Android class library Lots has changed since we forked Android's libcore, so merging the latest upstream code has required extensive changes to the Avian/Android port. One big change is that we now use Avian's versions of java.lang.Object, java.lang.Class, java.lang.ClassLoader, some java.lang.reflect.* classes, etc. instead of the Android versions. The main reason is that the Android versions have become very Dex/Dalvik-specific, and since Avian is based on Java class files, not dex archives, that code doesn't make sense here. This has the side benefit that we can share more native code with classpath-avian.cpp and reduce the amount of Java/C++ code duplication.
2014-08-21 19:09:42 +00:00
$(build)/android.dep: $(luni-javas) $(dalvik-javas) $(libart-javas) \
Makefile improvement to support okhttp, and Bouncy Castle for Android version
2014-11-23 09:55:09 +00:00
$(xml-javas) $(okhttp-android-javas) $(okhttp-javas) $(okio-javas) \
$(bcpkix-javas) $(bcprov-javas) $(luni-nonjavas) $(crypto-javas) $(crypto-platform-javas)
first pass at Android classpath support Stuff compiles, but linking breaks spectacularly. Next step is to figure out how to build the dependencies without checking out and building the entire Android platform.
2013-02-19 16:36:19 +00:00
@echo "compiling luni classes"
@mkdir -p $(classpath-build)
@mkdir -p $(build)/android
@mkdir -p $(build)/android-src/external/fdlibm
@mkdir -p $(build)/android-src/libexpat
use Android directory layout when accessing class library code Per https://github.com/ReadyTalk/avian/issues/53, Avian should build against a standard AOSP checkout, which means we should look for subprojects in the directories the repo utility would place them.
2013-04-22 23:10:23 +00:00
cp $(android)/external/fdlibm/fdlibm.h $(build)/android-src/external/fdlibm/
cp $(android)/external/expat/lib/expat*.h $(build)/android-src/libexpat/
Makefile improvement to support okhttp, and Bouncy Castle for Android version
2014-11-23 09:55:09 +00:00
cp -a $(luni-java)/* $(xml-java)/* $(okhttp-android-java)/* $(okhttp-java)/* $(okio-java)/* $(bcpkix-java)/* $(bcprov-java)/* $(build)/android-src/
update to more recent version of Android class library Lots has changed since we forked Android's libcore, so merging the latest upstream code has required extensive changes to the Avian/Android port. One big change is that we now use Avian's versions of java.lang.Object, java.lang.Class, java.lang.ClassLoader, some java.lang.reflect.* classes, etc. instead of the Android versions. The main reason is that the Android versions have become very Dex/Dalvik-specific, and since Avian is based on Java class files, not dex archives, that code doesn't make sense here. This has the side benefit that we can share more native code with classpath-avian.cpp and reduce the amount of Java/C++ code duplication.
2014-08-21 19:09:42 +00:00
rm $(call noop-files,$(luni-blacklist),$(luni-java),$(build)/android-src)
(cd $(dalvik-java) && \
Fixed jar tool usage - we should call "$(jar)" instead of simply "jar"
2014-08-31 21:13:14 +00:00
$(jar) c $(call noop-files,$(dalvik-javas),$(dalvik-java),.)) \
| (cd $(build)/android-src && $(jar) x)
update to more recent version of Android class library Lots has changed since we forked Android's libcore, so merging the latest upstream code has required extensive changes to the Avian/Android port. One big change is that we now use Avian's versions of java.lang.Object, java.lang.Class, java.lang.ClassLoader, some java.lang.reflect.* classes, etc. instead of the Android versions. The main reason is that the Android versions have become very Dex/Dalvik-specific, and since Avian is based on Java class files, not dex archives, that code doesn't make sense here. This has the side benefit that we can share more native code with classpath-avian.cpp and reduce the amount of Java/C++ code duplication.
2014-08-21 19:09:42 +00:00
(cd $(libart-java) && \
Fixed jar tool usage - we should call "$(jar)" instead of simply "jar"
2014-08-31 21:13:14 +00:00
$(jar) c $(call noop-files,$(libart-javas),$(libart-java),.)) \
| (cd $(build)/android-src && $(jar) x)
fix SSL stack and Android JAR resource loading
2014-08-22 13:05:58 +00:00
(cd $(crypto-java) && \
Fixed jar tool usage - we should call "$(jar)" instead of simply "jar"
2014-08-31 21:13:14 +00:00
$(jar) c $(call noop-files,$(crypto-javas),$(crypto-java),.)) \
| (cd $(build)/android-src && $(jar) x)
fix SSL stack and Android JAR resource loading
2014-08-22 13:05:58 +00:00
(cd $(crypto-platform-java) && \
Fixed jar tool usage - we should call "$(jar)" instead of simply "jar"
2014-08-31 21:13:14 +00:00
$(jar) c $(call noop-files,$(crypto-platform-javas),$(crypto-platform-java),.)) \
| (cd $(build)/android-src && $(jar) x)
update to more recent version of Android class library Lots has changed since we forked Android's libcore, so merging the latest upstream code has required extensive changes to the Avian/Android port. One big change is that we now use Avian's versions of java.lang.Object, java.lang.Class, java.lang.ClassLoader, some java.lang.reflect.* classes, etc. instead of the Android versions. The main reason is that the Android versions have become very Dex/Dalvik-specific, and since Avian is based on Java class files, not dex archives, that code doesn't make sense here. This has the side benefit that we can share more native code with classpath-avian.cpp and reduce the amount of Java/C++ code duplication.
2014-08-21 19:09:42 +00:00
(cd $(classpath-src) && \
Fixed jar tool usage - we should call "$(jar)" instead of simply "jar"
2014-08-31 21:13:14 +00:00
$(jar) c $(call noop-files,$(classpath-sources),$(classpath-src),.)) \
| (cd $(build)/android-src && $(jar) x)
# (cd android && $(jar) c *) | (cd $(build)/android-src && $(jar) x)
first pass at Android classpath support Stuff compiles, but linking breaks spectacularly. Next step is to figure out how to build the dependencies without checking out and building the entire Android platform.
2013-02-19 16:36:19 +00:00
find $(build)/android-src -name '*.java' > $(build)/android.txt
update to more recent version of Android class library Lots has changed since we forked Android's libcore, so merging the latest upstream code has required extensive changes to the Avian/Android port. One big change is that we now use Avian's versions of java.lang.Object, java.lang.Class, java.lang.ClassLoader, some java.lang.reflect.* classes, etc. instead of the Android versions. The main reason is that the Android versions have become very Dex/Dalvik-specific, and since Avian is based on Java class files, not dex archives, that code doesn't make sense here. This has the side benefit that we can share more native code with classpath-avian.cpp and reduce the amount of Java/C++ code duplication.
2014-08-21 19:09:42 +00:00
$(javac) -Xmaxerrs 1000 -d $(build)/android @$(build)/android.txt
more Android class library work
2013-02-21 22:37:17 +00:00
rm $(build)/android/sun/misc/Unsafe* \
$(build)/android/java/lang/reflect/Proxy*
Made it so all non-.java files from luni/src/main will be added to classpath directory
2014-03-18 15:22:52 +00:00
for x in $(luni-copied-nonjavas); \
Small refactoring of the non-class file copying
2014-03-30 19:10:48 +00:00
do cp $(luni-java)$${x} $(build)/android$${x} ; \
Made .properties be copied from Android Java source tree
2014-03-16 20:15:33 +00:00
done
Added patching of security.properties - removed "com.android." in front of all "com.android.org.*" classes
2014-03-18 19:13:35 +00:00
# fix security.properties - get rid of "com.android" in front of classes starting with "org"
sed -i -e 's/\(.*=\)com\.android\.\(org\..*\)/\1\2/g' \
Small refactoring of the non-class file copying
2014-03-30 19:10:48 +00:00
$(build)/android/java/security/security.properties
chmod +w $(build)/android/java/security/security.properties
first pass at Android classpath support Stuff compiles, but linking breaks spectacularly. Next step is to figure out how to build the dependencies without checking out and building the entire Android platform.
2013-02-19 16:36:19 +00:00
cp -r $(build)/android/* $(classpath-build)
add process=interpret support for Linux/ARM64 This makes all the tests pass for the platform=linux arch=arm64 process=interpret build. Next step: process=compile support.
2014-11-26 01:36:34 +00:00
@touch $(@)
first pass at Android classpath support Stuff compiles, but linking breaks spectacularly. Next step is to figure out how to build the dependencies without checking out and building the entire Android platform.
2013-02-19 16:36:19 +00:00
support classloading from jar files compiled into the executable
2007-10-25 22:06:05 +00:00
$(test-build)/%.class: $(test)/%.java
sketch of singleton support
2007-11-02 21:08:14 +00:00
@echo $(<)
add JNI test to test suite
2012-08-01 16:02:55 +00:00
$(test-dep): $(test-sources) $(test-library)
tweak makefile to recompile all dirty java files in a single javac invocation for efficiency
2007-11-02 14:01:10 +00:00
@echo "compiling test classes"
add build option for using GNU Classpath
2009-06-03 00:55:49 +00:00
@mkdir -p $(test-build)
build everything (including classes) in options-specific directory Previously, we only built platform specific code in an options-specific directory (e.g. build/linux-x86_64-debug), but built the Java classes in the shared, top-level build directory. This minimized duplication when building with different sets of options, but now that we're supporting the option of selecting a third-party class library, we need to put the classes in the options-specific directory as well.
2010-09-17 22:03:37 +00:00
files="$(shell $(MAKE) -s --no-print-directory build=$(build) $(test-classes))"; \
fix test class build when boot classes change
2009-09-17 17:50:35 +00:00
if test -n "$${files}"; then \
allow building with Java 7
2015-08-06 23:23:14 +00:00
$(javac) -source 1.$(java-version) -target 1.$(java-version) \
add process=interpret support for Linux/ARM64 This makes all the tests pass for the platform=linux arch=arm64 process=interpret build. Next step: process=compile support.
2014-11-26 01:36:34 +00:00
-classpath $(test-build) -d $(test-build) -bootclasspath $(boot-classpath) $${files}; \
fix test class build when boot classes change
2009-09-17 17:50:35 +00:00
fi
implement NewDirectByteBuffer etc. properly when building against Classpath; call JNI_OnLoad if found in newly-loaded libraries
2009-06-11 00:15:00 +00:00
$(javac) -source 1.2 -target 1.1 -XDjsrlimit=0 -d $(test-build) \
add new subroutine test This test covers the case where a local stack slot is first used to store an object reference and later to store a subroutine return address. Unfortunately, this confuses the VM's stack mapping code; I'll be working on a fix for that next. The new test requires generating bytecode from scratch, since there's no reliable way to get javac to generate the code we want. Since we already had primitive bytecode construction code in Proxy.java, I factored it out so we can reuse it in Subroutine.java.
2011-02-16 18:41:33 +00:00
-bootclasspath $(boot-classpath) test/Subroutine.java
support classloading from jar files compiled into the executable
2007-10-25 22:06:05 +00:00
@touch $(@)
copy makefile, stdc++.cpp, and type-generator.cpp from mess project and began customizing them for this project
2007-06-03 23:16:07 +00:00
build but don't run extra tests in makefile
2009-05-18 15:17:09 +00:00
$(test-extra-dep): $(test-extra-sources)
@echo "compiling extra test classes"
add build option for using GNU Classpath
2009-06-03 00:55:49 +00:00
@mkdir -p $(test-build)
build everything (including classes) in options-specific directory Previously, we only built platform specific code in an options-specific directory (e.g. build/linux-x86_64-debug), but built the Java classes in the shared, top-level build directory. This minimized duplication when building with different sets of options, but now that we're supporting the option of selecting a third-party class library, we need to put the classes in the options-specific directory as well.
2010-09-17 22:03:37 +00:00
files="$(shell $(MAKE) -s --no-print-directory build=$(build) $(test-extra-classes))"; \
fix test class build when boot classes change
2009-09-17 17:50:35 +00:00
if test -n "$${files}"; then \
allow building with Java 7
2015-08-06 23:23:14 +00:00
$(javac) -source 1.$(java-version) -target 1.$(java-version) \
add process=interpret support for Linux/ARM64 This makes all the tests pass for the platform=linux arch=arm64 process=interpret build. Next step: process=compile support.
2014-11-26 01:36:34 +00:00
-d $(test-build) -bootclasspath $(boot-classpath) $${files}; \
fix test class build when boot classes change
2009-09-17 17:50:35 +00:00
fi
build but don't run extra tests in makefile
2009-05-18 15:17:09 +00:00
@touch $(@)
implement primitive testing framework and provide for GC stress testing
2007-07-16 01:03:02 +00:00
define compile-object
copy makefile, stdc++.cpp, and type-generator.cpp from mess project and began customizing them for this project
2007-06-03 23:16:07 +00:00
@echo "compiling $(@)"
implement fixed object support
2007-10-28 01:54:30 +00:00
@mkdir -p $(dir $(@))
add support for LZMA on Windows
2012-06-02 21:43:42 +00:00
$(cxx) $(cflags) -c $$($(windows-path) $(<)) $(call output,$(@))
implement primitive testing framework and provide for GC stress testing
2007-07-16 01:03:02 +00:00
endef
copy makefile, stdc++.cpp, and type-generator.cpp from mess project and began customizing them for this project
2007-06-03 23:16:07 +00:00
Merge branch 'master' into powerpc Conflicts: makefile src/assembler.h src/compile.cpp src/compiler.cpp src/compiler.h src/finder.cpp
2008-11-11 15:20:49 +00:00
define compile-asm-object
@echo "compiling $(@)"
@mkdir -p $(dir $(@))
Windows Phone 8 / Windows RT initial support Conflicts: makefile
2013-01-28 15:20:52 +00:00
$(as) $(asmflags) $(call asm-output,$(@)) $(call asm-input,$(<))
Merge branch 'master' into powerpc Conflicts: makefile src/assembler.h src/compile.cpp src/compiler.cpp src/compiler.h src/finder.cpp
2008-11-11 15:20:49 +00:00
endef
add unittest framework
2013-02-15 03:54:20 +00:00
define compile-unittest-object
@echo "compiling $(@)"
@mkdir -p $(dir $(@))
fix msvc build
2013-03-04 18:09:59 +00:00
$(cxx) $(cflags) -c $$($(windows-path) $(<)) -I$(unittest) $(call output,$(@))
add unittest framework
2013-02-15 03:54:20 +00:00
endef
build everything (including classes) in options-specific directory Previously, we only built platform specific code in an options-specific directory (e.g. build/linux-x86_64-debug), but built the Java classes in the shared, top-level build directory. This minimized duplication when building with different sets of options, but now that we're supporting the option of selecting a third-party class library, we need to put the classes in the options-specific directory as well.
2010-09-17 22:03:37 +00:00
$(vm-cpp-objects): $(build)/%.o: $(src)/%.cpp $(vm-depends)
support classloading from jar files compiled into the executable
2007-10-25 22:06:05 +00:00
$(compile-object)
fix unittests for process=interpret build
2013-02-16 03:44:55 +00:00
ifeq ($(process),interpret)
$(all-codegen-target-objects): $(build)/%.o: $(src)/%.cpp $(vm-depends)
$(compile-object)
endif
add unittest framework
2013-02-15 03:54:20 +00:00
$(unittest-objects): $(build)/unittest/%.o: $(unittest)/%.cpp $(vm-depends) $(unittest-depends)
$(compile-unittest-object)
add JNI test to test suite
2012-08-01 16:02:55 +00:00
$(test-cpp-objects): $(test-build)/%.o: $(test)/%.cpp $(vm-depends)
$(compile-object)
$(test-library): $(test-cpp-objects)
@echo "linking $(@)"
Windows Phone 8 / Windows RT initial support Conflicts: makefile
2013-01-28 15:20:52 +00:00
ifdef ms_cl_compiler
Allow output of exceptions to debugger. Generate WinMD file
2013-01-30 08:42:05 +00:00
$(ld) $(shared) $(lflags) $(^) -out:$(@) \
-debug -PDB:$(subst $(so-suffix),.pdb,$(@)) \
Windows Phone 8 / Windows RT initial support Conflicts: makefile
2013-01-28 15:20:52 +00:00
-IMPLIB:$(test-build)/$(name).lib $(manifest-flags)
ifdef mt
$(mt) -nologo -manifest $(@).manifest -outputresource:"$(@);2"
endif
add JNI test to test suite
2012-08-01 16:02:55 +00:00
else
$(ld) $(^) $(shared) $(lflags) -o $(@)
endif
Added new embed utility for Windows
2012-11-13 08:11:49 +00:00
ifdef embed
$(embed): $(embed-objects) $(embed-loader-o)
@echo "building $(embed)"
Windows Phone 8 / Windows RT initial support Conflicts: makefile
2013-01-28 15:20:52 +00:00
ifdef ms_cl_compiler
Allow output of exceptions to debugger. Generate WinMD file
2013-01-30 08:42:05 +00:00
$(ld) $(lflags) $(^) -out:$(@) \
-debug -PDB:$(subst $(exe-suffix),.pdb,$(@)) $(manifest-flags)
Windows Phone 8 / Windows RT initial support Conflicts: makefile
2013-01-28 15:20:52 +00:00
ifdef mt
$(mt) -nologo -manifest $(@).manifest -outputresource:"$(@);1"
endif
fix MSVC build regression
2012-12-05 22:30:49 +00:00
else
$(cxx) $(^) $(lflags) $(static) $(call output,$(@))
endif
fix Windows cross-compile build
2012-12-05 00:37:02 +00:00
Added new embed utility for Windows
2012-11-13 08:11:49 +00:00
$(build-embed)/%.o: $(src)/%.cpp
@echo "compiling $(@)"
@mkdir -p $(dir $(@))
fix MSVC build regression
2012-12-05 22:30:49 +00:00
$(cxx) $(cflags) -c $(<) $(call output,$(@))
fix Windows cross-compile build
2012-12-05 00:37:02 +00:00
Added new embed utility for Windows
2012-11-13 08:11:49 +00:00
$(embed-loader-o): $(embed-loader) $(converter)
@mkdir -p $(dir $(@))
$(converter) $(<) $(@) _binary_loader_start \
_binary_loader_end $(target-format) $(arch)
Fixed embed.exe loader
2013-04-02 08:37:20 +00:00
$(embed-loader): $(embed-loader-objects) $(vm-objects) $(classpath-objects) \
first step towards supporting Android class library on Windows A majority of the tests are passing, but more work is needed to get them all to pass.
2013-03-14 21:33:05 +00:00
$(heapwalk-objects) $(lzma-decode-objects)
Windows Phone 8 / Windows RT initial support Conflicts: makefile
2013-01-28 15:20:52 +00:00
ifdef ms_cl_compiler
fix windows embed-loader build
2013-03-01 21:52:27 +00:00
$(ld) $(lflags) $(^) -out:$(@) \
Allow output of exceptions to debugger. Generate WinMD file
2013-01-30 08:42:05 +00:00
-debug -PDB:$(subst $(exe-suffix),.pdb,$(@)) $(manifest-flags)
Windows Phone 8 / Windows RT initial support Conflicts: makefile
2013-01-28 15:20:52 +00:00
ifdef mt
$(mt) -nologo -manifest $(@).manifest -outputresource:"$(@);1"
endif
fix MSVC build regression
2012-12-05 22:30:49 +00:00
else
fix windows embed-loader build
2013-03-01 21:52:27 +00:00
$(dlltool) -z $(addsuffix .def,$(basename $(@))) $(^)
add process=interpret support for Linux/ARM64 This makes all the tests pass for the platform=linux arch=arm64 process=interpret build. Next step: process=compile support.
2014-11-26 01:36:34 +00:00
$(dlltool) -d $(addsuffix .def,$(basename $(@))) -e $(addsuffix .exp,$(basename $(@)))
fix windows embed-loader build
2013-03-01 21:52:27 +00:00
$(ld) $(addsuffix .exp,$(basename $(@))) $(^) \
Fixes for avian-pack
2015-06-02 21:01:03 +00:00
$(lflags) $(classpath-lflags) $(bootimage-lflags) -o $(@)
fix MSVC build regression
2012-12-05 22:30:49 +00:00
endif
$(strip) $(strip-all) $(@)
Added new embed utility for Windows
2012-11-13 08:11:49 +00:00
$(build-embed-loader)/%.o: $(src)/%.cpp
@echo "compiling $(@)"
@mkdir -p $(dir $(@))
fix MSVC build regression
2012-12-05 22:30:49 +00:00
$(cxx) $(cflags) -c $(<) $(call output,$(@))
Added new embed utility for Windows
2012-11-13 08:11:49 +00:00
endif
add optional LZMA support for compressing embedded JARs, boot images, and shared objects
2012-06-02 15:06:22 +00:00
$(build)/%.o: $(lzma)/C/%.c
fix OS X build regressions
2012-06-03 00:56:07 +00:00
@echo "compiling $(@)"
@mkdir -p $(dir $(@))
fix lzma build with clang Clang doesn't like seeing -std=c++0x when building C code, and many of the options in cflags don't apply to the LZMA source code anyway, so we give LZMA its own set of compiler flags.
2014-07-28 20:26:33 +00:00
$(cc) $(lzma-cflags) -c $$($(windows-path) $(<)) $(call output,$(@))
add optional LZMA support for compressing embedded JARs, boot images, and shared objects
2012-06-02 15:06:22 +00:00
Windows Phone 8 / Windows RT initial support Conflicts: makefile
2013-01-28 15:20:52 +00:00
$(vm-asm-objects): $(build)/%-asm.o: $(src)/%.$(asm-format)
Merge branch 'master' into powerpc Conflicts: makefile src/assembler.h src/compile.cpp src/compiler.cpp src/compiler.h src/finder.cpp
2008-11-11 15:20:49 +00:00
$(compile-asm-object)
fix test, stress, and fast makefile targets
2007-06-29 17:45:14 +00:00
build everything (including classes) in options-specific directory Previously, we only built platform specific code in an options-specific directory (e.g. build/linux-x86_64-debug), but built the Java classes in the shared, top-level build directory. This minimized duplication when building with different sets of options, but now that we're supporting the option of selecting a third-party class library, we need to put the classes in the options-specific directory as well.
2010-09-17 22:03:37 +00:00
$(bootimage-generator-objects): $(build)/%.o: $(src)/%.cpp $(vm-depends)
initial work to support boot image creation and use
2008-11-21 23:20:35 +00:00
$(compile-object)
build everything (including classes) in options-specific directory Previously, we only built platform specific code in an options-specific directory (e.g. build/linux-x86_64-debug), but built the Java classes in the shared, top-level build directory. This minimized duplication when building with different sets of options, but now that we're supporting the option of selecting a third-party class library, we need to put the classes in the options-specific directory as well.
2010-09-17 22:03:37 +00:00
$(heapwalk-objects): $(build)/%.o: $(src)/%.cpp $(vm-depends)
initial work to support boot image creation and use
2008-11-21 23:20:35 +00:00
$(compile-object)
support building Avian as a self-contained dynamic library
2008-03-31 03:43:43 +00:00
$(driver-object): $(driver-source)
$(compile-object)
implement -jar option in main.cpp
2010-11-27 21:46:07 +00:00
$(build)/main-dynamic.o: $(driver-source)
support building Avian as a self-contained dynamic library
2008-03-31 03:43:43 +00:00
@echo "compiling $(@)"
@mkdir -p $(dir $(@))
preliminary support for using OpenJDK's class library Whereas the GNU Classpath port used the strategy of patching Classpath with core classes from Avian so as to minimize changes to the VM, this port uses the opposite strategy: abstract and isolate classpath-specific features in the VM similar to how we abstract away platform-specific features in system.h. This allows us to use an unmodified copy of OpenJDK's class library, including its core classes and augmented by a few VM-specific classes in the "avian" package.
2010-09-10 21:05:29 +00:00
$(cxx) $(cflags) -DBOOT_LIBRARY=\"$(so-prefix)jvm$(so-suffix)\" \
fix OSX build regression
2009-08-27 15:30:34 +00:00
-c $(<) $(call output,$(@))
support building Avian as a self-contained dynamic library
2008-03-31 03:43:43 +00:00
$(boot-object): $(boot-source)
implement primitive testing framework and provide for GC stress testing
2007-07-16 01:03:02 +00:00
$(compile-object)
copy makefile, stdc++.cpp, and type-generator.cpp from mess project and began customizing them for this project
2007-06-03 23:16:07 +00:00
various fixes for embedded resource loading in OpenJDK build
2010-12-06 00:40:50 +00:00
$(boot-javahome-object): $(src)/boot-javahome.cpp
$(compile-object)
move binary-to-object to new tree structure
2013-02-20 03:23:11 +00:00
$(object-writer-objects) $(binary-to-object-objects): $(build)/%.o: $(src)/%.cpp $(binary-to-object-depends) $(object-writer-depends) $(all-depends)
improve build for binaryToObject
2012-04-25 15:09:26 +00:00
@mkdir -p $(dir $(@))
$(build-cxx) $(converter-cflags) -c $(<) -o $(@)
add binaryToObject utility We use this utility instead of objcopy to embed data into object files because it offers more control over e.g. section alignment, which is important for bootimage builds.
2009-10-14 01:00:16 +00:00
directly emit codeimage as a object (binaryToObject is statically linked in), as a stepping stone to including extra symbols in said codeimage
2012-05-01 19:16:32 +00:00
$(converter): $(converter-objects) $(converter-tool-objects)
move binary-to-object to new tree structure
2013-02-20 03:23:11 +00:00
@mkdir -p $(dir $(@))
refactor binaryToObject to allow more flexibilty (in particular, allowing arbitrary symbols per object)
2012-04-27 18:08:44 +00:00
$(build-cc) $(^) -g -o $(@)
add binaryToObject utility We use this utility instead of objcopy to embed data into object files because it offers more control over e.g. section alignment, which is important for bootimage builds.
2009-10-14 01:00:16 +00:00
add optional LZMA support for compressing embedded JARs, boot images, and shared objects
2012-06-02 15:06:22 +00:00
$(lzma-encoder-objects): $(build)/lzma/%.o: $(src)/lzma/%.cpp
@mkdir -p $(dir $(@))
fix lzma build with clang Clang doesn't like seeing -std=c++0x when building C code, and many of the options in cflags don't apply to the LZMA source code anyway, so we give LZMA its own set of compiler flags.
2014-07-28 20:26:33 +00:00
$(build-cxx) $(lzma-build-cflags) -c $(<) -o $(@)
add optional LZMA support for compressing embedded JARs, boot images, and shared objects
2012-06-02 15:06:22 +00:00
$(lzma-encoder): $(lzma-encoder-objects) $(lzma-encoder-lzma-objects)
$(build-cc) $(^) -g -o $(@)
expand support for publishing LZMA artifacts
2015-04-07 22:55:45 +00:00
$(lzma-library): $(lzma-loader) $(lzma-decode-objects)
@echo "creating $(@)"
@rm -rf $(build)/libavian-lzma
@mkdir -p $(build)/libavian-lzma
rm -rf $(@)
for x in $(^); \
do cp $${x} $(build)/libavian-lzma/$$(echo $${x} | sed s:/:_:g); \
done
ifdef ms_cl_compiler
$(ar) $(arflags) $(build)/libavian-lzma/*.o -out:$(@)
else
$(ar) cru $(@) $(build)/libavian-lzma/*.o
$(ranlib) $(@)
endif
add optional LZMA support for compressing embedded JARs, boot images, and shared objects
2012-06-02 15:06:22 +00:00
$(lzma-loader): $(src)/lzma/load.cpp
$(compile-object)
add support for accessing embedded JARs as if they were directories This allows OpenJDK to access time zone data which is normally found under java.home, but which we must embed in the executable itself to create a self-contained build. The VM intercepts various file operations, looking for paths which start with a prefix specified by the avian.embed.prefix property and redirecting those operations to an embedded JAR. For example, if avian.embed.prefix is "/avian-embedded", and code calls File.exists() with a path of "/avian-embedded/javahomeJar/foo.txt", the VM looks for a function named javahomeJar via dlsym, calls the function to find the memory region containing the embeded JAR, and finally consults the JAR to see if the file "foo.txt" exists.
2010-11-05 19:18:28 +00:00
$(build)/classpath.jar: $(classpath-dep) $(classpath-jar-dep)
@echo "creating $(@)"
(wd=$$(pwd) && \
cd $(classpath-build) && \
$(jar) c0f "$$($(native-path) "$${wd}/$(@)")" .)
add binaryToObject utility We use this utility instead of objcopy to embed data into object files because it offers more control over e.g. section alignment, which is important for bootimage builds.
2009-10-14 01:00:16 +00:00
$(classpath-object): $(build)/classpath.jar $(converter)
fix darwin build
2008-03-31 04:23:51 +00:00
@echo "creating $(@)"
add binaryToObject utility We use this utility instead of objcopy to embed data into object files because it offers more control over e.g. section alignment, which is important for bootimage builds.
2009-10-14 01:00:16 +00:00
$(converter) $(<) $(@) _binary_classpath_jar_start \
clean up binary format code Linux, FreeBSD, and QNX all use ELF, so no need to distinguish between them when generating object files. To avoid confusion, I've switch from using operating system names to using binary format names where applicable.
2012-08-02 18:15:15 +00:00
_binary_classpath_jar_end $(target-format) $(arch)
support classloading from jar files compiled into the executable
2007-10-25 22:06:05 +00:00
fix Cygwin/MSYS openjdk-src builds
2010-11-15 23:27:00 +00:00
$(build)/javahome.jar:
add support for accessing embedded JARs as if they were directories This allows OpenJDK to access time zone data which is normally found under java.home, but which we must embed in the executable itself to create a self-contained build. The VM intercepts various file operations, looking for paths which start with a prefix specified by the avian.embed.prefix property and redirecting those operations to an embedded JAR. For example, if avian.embed.prefix is "/avian-embedded", and code calls File.exists() with a path of "/avian-embedded/javahomeJar/foo.txt", the VM looks for a function named javahomeJar via dlsym, calls the function to find the memory region containing the embeded JAR, and finally consults the JAR to see if the file "foo.txt" exists.
2010-11-05 19:18:28 +00:00
@echo "creating $(@)"
(wd=$$(pwd) && \
fix Cygwin/MSYS openjdk-src builds
2010-11-15 23:27:00 +00:00
cd "$(build-javahome)" && \
add support for accessing embedded JARs as if they were directories This allows OpenJDK to access time zone data which is normally found under java.home, but which we must embed in the executable itself to create a self-contained build. The VM intercepts various file operations, looking for paths which start with a prefix specified by the avian.embed.prefix property and redirecting those operations to an embedded JAR. For example, if avian.embed.prefix is "/avian-embedded", and code calls File.exists() with a path of "/avian-embedded/javahomeJar/foo.txt", the VM looks for a function named javahomeJar via dlsym, calls the function to find the memory region containing the embeded JAR, and finally consults the JAR to see if the file "foo.txt" exists.
2010-11-05 19:18:28 +00:00
$(jar) c0f "$$($(native-path) "$${wd}/$(@)")" $(javahome-files))
$(javahome-object): $(build)/javahome.jar $(converter)
@echo "creating $(@)"
$(converter) $(<) $(@) _binary_javahome_jar_start \
clean up binary format code Linux, FreeBSD, and QNX all use ELF, so no need to distinguish between them when generating object files. To avoid confusion, I've switch from using operating system names to using binary format names where applicable.
2012-08-02 18:15:15 +00:00
_binary_javahome_jar_end $(target-format) $(arch)
add support for accessing embedded JARs as if they were directories This allows OpenJDK to access time zone data which is normally found under java.home, but which we must embed in the executable itself to create a self-contained build. The VM intercepts various file operations, looking for paths which start with a prefix specified by the avian.embed.prefix property and redirecting those operations to an embedded JAR. For example, if avian.embed.prefix is "/avian-embedded", and code calls File.exists() with a path of "/avian-embedded/javahomeJar/foo.txt", the VM looks for a function named javahomeJar via dlsym, calls the function to find the memory region containing the embeded JAR, and finally consults the JAR to see if the file "foo.txt" exists.
2010-11-05 19:18:28 +00:00
add optional LZMA support for compressing embedded JARs, boot images, and shared objects
2012-06-02 15:06:22 +00:00
define compile-generator-object
refactor build system to support cross-compiling
2007-10-22 20:56:27 +00:00
@echo "compiling $(@)"
implement fixed object support
2007-10-28 01:54:30 +00:00
@mkdir -p $(dir $(@))
clean up bootstrap type generation to eliminate redundancy (broken)
2007-11-04 21:15:28 +00:00
$(build-cxx) -DPOINTER_SIZE=$(pointer-size) -O0 -g3 $(build-cflags) \
-c $(<) -o $(@)
add optional LZMA support for compressing embedded JARs, boot images, and shared objects
2012-06-02 15:06:22 +00:00
endef
$(generator-objects): $(generator-depends)
$(generator-objects): $(build)/%-build.o: $(src)/%.cpp
$(compile-generator-object)
add support for LZMA on Windows
2012-06-02 21:43:42 +00:00
$(build)/%-build.o: $(lzma)/C/%.c
fix OS X build regressions
2012-06-03 00:56:07 +00:00
@echo "compiling $(@)"
@mkdir -p $(dir $(@))
fix lzma build with clang Clang doesn't like seeing -std=c++0x when building C code, and many of the options in cflags don't apply to the LZMA source code anyway, so we give LZMA its own set of compiler flags.
2014-07-28 20:26:33 +00:00
$(build-cc) -DPOINTER_SIZE=$(pointer-size) -O0 -g3 $(lzma-build-cflags) \
-c $(<) -o $(@)
copy makefile, stdc++.cpp, and type-generator.cpp from mess project and began customizing them for this project
2007-06-03 23:16:07 +00:00
make classpath native sources depend on vm-depends in makefile (fixes #340)
2014-09-24 19:15:41 +00:00
$(jni-objects): $(build)/%.o: $(classpath-src)/%.cpp $(vm-depends)
use compile-object to build jni-objects
2008-01-25 23:45:52 +00:00
$(compile-object)
more JNI work
2007-06-25 01:34:07 +00:00
various fixes for embedded resource loading in OpenJDK build
2010-12-06 00:40:50 +00:00
$(static-library): $(vm-objects) $(classpath-objects) $(vm-heapwalk-objects) \
add optional LZMA support for compressing embedded JARs, boot images, and shared objects
2012-06-02 15:06:22 +00:00
$(javahome-object) $(boot-javahome-object) $(lzma-decode-objects)
build standalone executable (no seperate .so) on all platforms; build libvm.a too
2007-10-23 23:23:55 +00:00
@echo "creating $(@)"
various tweaks to get hello-ios working with the Android library
2013-03-04 21:24:41 +00:00
@rm -rf $(build)/libavian
@mkdir -p $(build)/libavian
build fixes and refactoring
2007-10-26 21:02:39 +00:00
rm -rf $(@)
fix build when let expressions aren't available
2013-03-04 21:42:18 +00:00
for x in $(^); \
do cp $${x} $(build)/libavian/$$(echo $${x} | sed s:/:_:g); \
various tweaks to get hello-ios working with the Android library
2013-03-04 21:24:41 +00:00
done
Windows Phone 8 / Windows RT initial support Conflicts: makefile
2013-01-28 15:20:52 +00:00
ifdef ms_cl_compiler
various tweaks to get hello-ios working with the Android library
2013-03-04 21:24:41 +00:00
$(ar) $(arflags) $(build)/libavian/*.o -out:$(@)
Windows Phone 8 / Windows RT initial support Conflicts: makefile
2013-01-28 15:20:52 +00:00
else
various tweaks to get hello-ios working with the Android library
2013-03-04 21:24:41 +00:00
$(ar) cru $(@) $(build)/libavian/*.o
build standalone executable (no seperate .so) on all platforms; build libvm.a too
2007-10-23 23:23:55 +00:00
$(ranlib) $(@)
Windows Phone 8 / Windows RT initial support Conflicts: makefile
2013-01-28 15:20:52 +00:00
endif
build standalone executable (no seperate .so) on all platforms; build libvm.a too
2007-10-23 23:23:55 +00:00
fix openjdk bootimage cross builds We need to extract the OpenJDK classes into the build classpath directory for the target platform before running the bootimage-generator, or else it won't be able to find the classes.
2013-01-27 01:01:34 +00:00
$(bootimage-object) $(codeimage-object): $(bootimage-generator) \
add bootimage-test option to makefile This option specifies that the test classes should be AOT-compiled along with the class library, which allows us to test that everything works in AOT-compiled form as well as JIT-compiled form. This is primarily motivated by the need to test d906db6 (support for AOT-compilation of Java 8 lambda expressions). Note that I had to tweak Misc because it tested something that couldn't be done in an AOT build without a lot of extra work, and SystemClassLoader.getPackage because it was returning null when the requested package could not be populated with JAR manifest metadata. Technically, we probably *should* return null for packages that don't exist at all (in the sense that no classes have been loaded from such a package), but tracking that kind of thing seems like more trouble than it's worth unless someone complains about it.
2015-10-19 21:10:53 +00:00
$(classpath-jar-dep) $(test-dep)
Makefile fix
2013-01-28 17:15:29 +00:00
@echo "generating bootimage and codeimage binaries from $(classpath-build) using $(<)"
add bootimage-test option to makefile This option specifies that the test classes should be AOT-compiled along with the class library, which allows us to test that everything works in AOT-compiled form as well as JIT-compiled form. This is primarily motivated by the need to test d906db6 (support for AOT-compilation of Java 8 lambda expressions). Note that I had to tweak Misc because it tested something that couldn't be done in an AOT build without a lot of extra work, and SystemClassLoader.getPackage because it was returning null when the requested package could not be populated with JAR manifest metadata. Technically, we probably *should* return null for packages that don't exist at all (in the sense that no classes have been loaded from such a package), but tracking that kind of thing seems like more trouble than it's worth unless someone complains about it.
2015-10-19 21:10:53 +00:00
$(<) -cp $(bootimage-classpath) -bootimage $(bootimage-object) -codeimage $(codeimage-object) \
Windows Phone 8 / Windows RT initial support Conflicts: makefile
2013-01-28 15:20:52 +00:00
-bootimage-symbols $(bootimage-symbols) \
support AOT-compilation of Java 8 lambda expressions These expressions are tricky because they rely on invokedynamic, which normally implies runtime code generation. However, since lambdas don't actually use the "dynamicness" of invokedynamic, we can convert them into static calls to synthetic classes at compile time. Since I had already written code to synthesize such classes in Java and I didn't want to rewrite it in C++, I needed to add support for running Java code to the bootimage generator. And since the primary VM used by the generator is purpose-built to generate AOT-compiled code for a specific target architecture and is not capable of generating or running JIT-compiled code for the host architecture, I added support for loading a second, independent, host-specific VM for running Java code. The rest of the patch handles the fact that each method compilation might cause new, synthetic classes to be created, so we need to make sure those classes and their methods are included in the final heap and code images. This required breaking some giant code blocks out of makeCodeImage into their own methods, which makes the diff look scarier than it really is.
2015-09-13 02:15:46 +00:00
-codeimage-symbols $(codeimage-symbols) \
-hostvm $(host-vm)
generate read-only code image in bootimage build This avoids the requirement of putting the code image in a section/segment which is both writable and executable, which is good for security and avoids trouble with systems like iOS which disallow such things. The implementation relies on relative addressing such that the offset of the desired address is fixed as a compile-time constant relative to the start of the memory area of interest (e.g. the code image, heap image, or thunk table). At runtime, the base pointer to the memory area is retrieved from the thread structure and added to the offset to compute the final address. Using the thread pointer allows us to generate read-only, position-independent code while avoiding the use of IP-relative addressing, which is not available on all architectures.
2011-09-20 22:30:30 +00:00
enable standalone OpenJDK builds As described in readme.txt, a standalone OpenJDK build embeds all libraries, classes, and other files needed at runtime in the resulting binary, eliminating dependencies on external resources.
2010-11-04 17:02:09 +00:00
executable-objects = $(vm-objects) $(classpath-objects) $(driver-object) \
generate read-only code image in bootimage build This avoids the requirement of putting the code image in a section/segment which is both writable and executable, which is good for security and avoids trouble with systems like iOS which disallow such things. The implementation relies on relative addressing such that the offset of the desired address is fixed as a compile-time constant relative to the start of the memory area of interest (e.g. the code image, heap image, or thunk table). At runtime, the base pointer to the memory area is retrieved from the thread structure and added to the offset to compute the final address. Using the thread pointer allows us to generate read-only, position-independent code while avoiding the use of IP-relative addressing, which is not available on all architectures.
2011-09-20 22:30:30 +00:00
$(vm-heapwalk-objects) $(boot-object) $(vm-classpath-objects) \
add optional LZMA support for compressing embedded JARs, boot images, and shared objects
2012-06-02 15:06:22 +00:00
$(javahome-object) $(boot-javahome-object) $(lzma-decode-objects)
fix build breakage due to makefile syntax confusion Consider the following makefile construct: target: dep0 target: dep1 dep2 command -o $(@) $(^) With some versions of make, $(^) will expand to "dep1 dep2", but on others it will expand to "dep0 dep1 dep2". The commit ensures that we no longer rely on the former behavior (nor the latter, for that matter).
2010-10-23 00:20:19 +00:00
more progress on Android class library port Hello.java works. Yay.
2013-02-20 17:22:40 +00:00
unittest-executable-objects = $(unittest-objects) $(vm-objects) \
fix heapdump=true build
2014-07-16 20:18:42 +00:00
$(vm-heapwalk-objects) $(build)/util/arg-parser.o $(stub-objects) \
$(lzma-decode-objects)
add unittest framework
2013-02-15 03:54:20 +00:00
fix unittests for process=interpret build
2013-02-16 03:44:55 +00:00
ifeq ($(process),interpret)
unittest-executable-objects += $(all-codegen-target-objects)
endif
begin work on audit-codegen
2013-02-18 14:50:37 +00:00
# apparently, make does poorly with ifs inside of defines, and indented defines.
# I suggest re-indenting the following before making edits (and unindenting afterwards):
ifneq ($(platform),windows)
define link-executable
@echo linking $(@)
various tweaks to get hello-ios working with the Android library
2013-03-04 21:24:41 +00:00
$(ld) $(^) $(rdynamic) $(lflags) $(classpath-lflags) $(bootimage-lflags) \
-o $(@)
begin work on audit-codegen
2013-02-18 14:50:37 +00:00
endef
else
Windows Phone 8 / Windows RT initial support Conflicts: makefile
2013-01-28 15:20:52 +00:00
ifdef ms_cl_compiler
ifdef mt
begin work on audit-codegen
2013-02-18 14:50:37 +00:00
define link-executable
@echo linking $(@)
$(ld) $(lflags) $(^) -out:$(@) \
-debug -PDB:$(subst $(exe-suffix),.pdb,$(@)) $(manifest-flags)
Windows Phone 8 / Windows RT initial support Conflicts: makefile
2013-01-28 15:20:52 +00:00
$(mt) -nologo -manifest $(@).manifest -outputresource:"$(@);1"
begin work on audit-codegen
2013-02-18 14:50:37 +00:00
endef
else
define link-executable
@echo linking $(@)
Windows Phone 8 / Windows RT initial support Conflicts: makefile
2013-01-28 15:20:52 +00:00
$(mt) -nologo -manifest $(@).manifest -outputresource:"$(@);1"
begin work on audit-codegen
2013-02-18 14:50:37 +00:00
endef
Windows Phone 8 / Windows RT initial support Conflicts: makefile
2013-01-28 15:20:52 +00:00
endif
add support for building with MSVC on Windows
2009-08-27 00:26:44 +00:00
else
begin work on audit-codegen
2013-02-18 14:50:37 +00:00
define link-executable
@echo linking $(@)
$(dlltool) -z $(@).def $(^)
Merge branch 'master' into powerpc Conflicts: makefile src/assembler.h src/compile.cpp src/compiler.cpp src/compiler.h src/finder.cpp
2008-11-11 15:20:49 +00:00
$(dlltool) -d $(@).def -e $(@).exp
first step towards supporting Android class library on Windows A majority of the tests are passing, but more work is needed to get them all to pass.
2013-03-14 21:33:05 +00:00
$(ld) $(@).exp $(^) $(lflags) $(classpath-lflags) -o $(@)
begin work on audit-codegen
2013-02-18 14:50:37 +00:00
endef
add support for building with MSVC on Windows
2009-08-27 00:26:44 +00:00
endif
build fixes and refactoring
2007-10-26 21:02:39 +00:00
endif
copy makefile, stdc++.cpp, and type-generator.cpp from mess project and began customizing them for this project
2007-06-03 23:16:07 +00:00
begin work on audit-codegen
2013-02-18 14:50:37 +00:00
$(executable): $(executable-objects)
$(link-executable)
add unittest framework
2013-02-15 03:54:20 +00:00
$(unittest-executable): $(unittest-executable-objects)
begin work on audit-codegen
2013-02-18 14:50:37 +00:00
$(link-executable)
rebuild bootimage-generator when vm-objects have changed
2013-03-15 19:16:20 +00:00
$(bootimage-generator): $(bootimage-generator-objects) $(vm-objects)
Makefile fix
2013-01-28 17:15:29 +00:00
echo building $(bootimage-generator) arch=$(build-arch) platform=$(bootimage-platform)
support AOT-compilation of Java 8 lambda expressions These expressions are tricky because they rely on invokedynamic, which normally implies runtime code generation. However, since lambdas don't actually use the "dynamicness" of invokedynamic, we can convert them into static calls to synthetic classes at compile time. Since I had already written code to synthesize such classes in Java and I didn't want to rewrite it in C++, I needed to add support for running Java code to the bootimage generator. And since the primary VM used by the generator is purpose-built to generate AOT-compiled code for a specific target architecture and is not capable of generating or running JIT-compiled code for the host architecture, I added support for loading a second, independent, host-specific VM for running Java code. The rest of the patch handles the fact that each method compilation might cause new, synthetic classes to be created, so we need to make sure those classes and their methods are included in the final heap and code images. This required breaking some giant code blocks out of makeCodeImage into their own methods, which makes the diff look scarier than it really is.
2015-09-13 02:15:46 +00:00
$(MAKE) process=interpret bootimage= mode=$(mode)
fix parallel bootimage build
2009-09-06 00:26:13 +00:00
$(MAKE) mode=$(mode) \
Makefile fix
2013-01-28 17:15:29 +00:00
build=$(host-build-root) \
support cross-platform bootimage build (as long as the build arch matches the target arch
2008-12-05 00:24:38 +00:00
arch=$(build-arch) \
Makefile changes, added additional compiler flags
2013-01-30 05:31:02 +00:00
aot-only=false \
directly emit codeimage as a object (binaryToObject is statically linked in), as a stepping stone to including extra symbols in said codeimage
2012-05-01 19:16:32 +00:00
target-arch=$(arch) \
use armv7 memory barriers by default armv7 and later provide weaker cache coherency models than armv6 and earlier, so we cannot just implement memory barriers as no-ops. This patch uses the DMB instruction (or the equivalent OS-provided barrier function) to implement barriers. This should fix concurrency issues on newer chips such as the Apple A6 and A7. If you still need to support ARMv6 devices, you should pass "armv6=true" to make when building Avian. Ideally, the VM would detect what kind of CPU it was executing on at runtime and direct the JIT compiler accordingly, but I don't know how to do that on ARM. Patches are welcome, though!
2014-01-10 21:38:57 +00:00
armv6=$(armv6) \
fix cygwin/msys bootimage build
2009-06-01 18:12:29 +00:00
platform=$(bootimage-platform) \
clean up binary format code Linux, FreeBSD, and QNX all use ELF, so no need to distinguish between them when generating object files. To avoid confusion, I've switch from using operating system names to using binary format names where applicable.
2012-08-02 18:15:15 +00:00
target-format=$(target-format) \
fix all the bugs So there I was, planning to just fix one little bug: Thread.holdsLock and Thread.yield were missing for the Android class library. Easy enough, right? So, I added a test, got it passing, and figured I'd go ahead and run ci.sh with all three class libraries. Big mistake. Here's the stuff I found: * minor inconsistency in README.md about OpenSSL version * untested, broken Class.getEnclosingMethod (reported by Josh) * JNI test failed for tails=true Android build * Runtime.nativeExit missing for Android build * obsolete assertion in CallEvent broke tails=true Android build * obsolete superclass field offset padding broke bootimage=true Android build * runtime annotation parsing broke bootimage=true Android build (because we couldn't modify Addendum.annotationTable for classes in the heap image) * ci.sh tried building with both android=... and openjdk=..., which the makefile rightfully balked at Sorry this is all in a single commit; I didn't expect so many unrelated issues, and I'm too lazy to break them apart.
2014-07-12 22:03:11 +00:00
android=$(android) \
fix openjdk-src bootimage build The main change here is to use a lazily-populated vector to associate runtime data with classes instead of referencing them directly from the class which requires updating immutable references in the heap image. The other changes employ other strategies to avoid trying to update immutable references.
2010-11-26 19:41:31 +00:00
openjdk=$(openjdk) \
openjdk-src=$(openjdk-src) \
support cross-platform bootimage build (as long as the build arch matches the target arch
2008-12-05 00:24:38 +00:00
bootimage-generator= \
build-bootimage-generator=$(bootimage-generator) \
more progress towards cross-architecture bootimage builds This commit fixes a lot of bugs. All tests are now pass for Linux x86_64 to Linux i386 cross builds.
2011-09-01 03:18:00 +00:00
target-cflags="$(bootimage-cflags)" \
target-asm=$(asm) \
fix parallel bootimage build
2009-09-06 00:26:13 +00:00
$(bootimage-generator)
support cross-platform bootimage build (as long as the build arch matches the target arch
2008-12-05 00:24:38 +00:00
$(build-bootimage-generator): \
various tweaks to get hello-ios working with the Android library
2013-03-04 21:24:41 +00:00
$(vm-objects) $(classpath-object) \
add optional LZMA support for compressing embedded JARs, boot images, and shared objects
2012-06-02 15:06:22 +00:00
$(heapwalk-objects) $(bootimage-generator-objects) $(converter-objects) \
$(lzma-decode-objects) $(lzma-encode-objects)
copy makefile, stdc++.cpp, and type-generator.cpp from mess project and began customizing them for this project
2007-06-03 23:16:07 +00:00
@echo "linking $(@)"
build fixes and refactoring
2007-10-26 21:02:39 +00:00
ifeq ($(platform),windows)
Windows Phone 8 / Windows RT initial support Conflicts: makefile
2013-01-28 15:20:52 +00:00
ifdef ms_cl_compiler
Allow output of exceptions to debugger. Generate WinMD file
2013-01-30 08:42:05 +00:00
$(ld) $(bootimage-generator-lflags) $(lflags) $(^) -out:$(@) \
-debug -PDB:$(subst $(exe-suffix),.pdb,$(@)) $(manifest-flags)
Windows Phone 8 / Windows RT initial support Conflicts: makefile
2013-01-28 15:20:52 +00:00
ifdef mt
$(mt) -nologo -manifest $(@).manifest -outputresource:"$(@);1"
endif
add support for building with MSVC on Windows
2009-08-27 00:26:44 +00:00
else
build fixes and refactoring
2007-10-26 21:02:39 +00:00
$(dlltool) -z $(@).def $(^)
look for %s@%d format symbols when resolving native methods on Windows, which obviates the need for -k flag to dlltool
2008-10-10 23:37:36 +00:00
$(dlltool) -d $(@).def -e $(@).exp
Windows Phone 8 / Windows RT initial support Conflicts: makefile
2013-01-28 15:20:52 +00:00
$(ld) $(@).exp $(^) $(bootimage-generator-lflags) $(lflags) -o $(@)
add support for building with MSVC on Windows
2009-08-27 00:26:44 +00:00
endif
build fixes and refactoring
2007-10-26 21:02:39 +00:00
else
Windows Phone 8 / Windows RT initial support Conflicts: makefile
2013-01-28 15:20:52 +00:00
$(ld) $(^) $(rdynamic) $(bootimage-generator-lflags) $(lflags) -o $(@)
build fixes and refactoring
2007-10-26 21:02:39 +00:00
endif
copy makefile, stdc++.cpp, and type-generator.cpp from mess project and began customizing them for this project
2007-06-03 23:16:07 +00:00
add support for accessing embedded JARs as if they were directories This allows OpenJDK to access time zone data which is normally found under java.home, but which we must embed in the executable itself to create a self-contained build. The VM intercepts various file operations, looking for paths which start with a prefix specified by the avian.embed.prefix property and redirecting those operations to an embedded JAR. For example, if avian.embed.prefix is "/avian-embedded", and code calls File.exists() with a path of "/avian-embedded/javahomeJar/foo.txt", the VM looks for a function named javahomeJar via dlsym, calls the function to find the memory region containing the embeded JAR, and finally consults the JAR to see if the file "foo.txt" exists.
2010-11-05 19:18:28 +00:00
$(dynamic-library): $(vm-objects) $(dynamic-object) $(classpath-objects) \
generate read-only code image in bootimage build This avoids the requirement of putting the code image in a section/segment which is both writable and executable, which is good for security and avoids trouble with systems like iOS which disallow such things. The implementation relies on relative addressing such that the offset of the desired address is fixed as a compile-time constant relative to the start of the memory area of interest (e.g. the code image, heap image, or thunk table). At runtime, the base pointer to the memory area is retrieved from the thread structure and added to the offset to compute the final address. Using the thread pointer allows us to generate read-only, position-independent code while avoiding the use of IP-relative addressing, which is not available on all architectures.
2011-09-20 22:30:30 +00:00
$(vm-heapwalk-objects) $(boot-object) $(vm-classpath-objects) \
add optional LZMA support for compressing embedded JARs, boot images, and shared objects
2012-06-02 15:06:22 +00:00
$(classpath-libraries) $(javahome-object) $(boot-javahome-object) \
$(lzma-decode-objects)
support building Avian as a self-contained dynamic library
2008-03-31 03:43:43 +00:00
@echo "linking $(@)"
Windows Phone 8 / Windows RT initial support Conflicts: makefile
2013-01-28 15:20:52 +00:00
ifdef ms_cl_compiler
Allow output of exceptions to debugger. Generate WinMD file
2013-01-30 08:42:05 +00:00
$(ld) $(shared) $(lflags) $(^) -out:$(@) \
-debug -PDB:$(subst $(so-suffix),.pdb,$(@)) \
-IMPLIB:$(subst $(so-suffix),.lib,$(@)) $(manifest-flags)
Windows Phone 8 / Windows RT initial support Conflicts: makefile
2013-01-28 15:20:52 +00:00
ifdef mt
$(mt) -nologo -manifest $(@).manifest -outputresource:"$(@);2"
endif
add support for building with MSVC on Windows
2009-08-27 00:26:44 +00:00
else
Add a non-versioned SONAME to Avian libjvm.so that matches the Hotspot Server/Client libjvm.so SONAME, this allow libjava.so in OpenJDK 7 to find the Avian libjvm.so during ldopen.
2012-09-05 08:29:12 +00:00
$(ld) $(^) $(version-script-flag) $(soname-flag) \
various tweaks to get hello-ios working with the Android library
2013-03-04 21:24:41 +00:00
$(shared) $(lflags) $(classpath-lflags) $(bootimage-lflags) \
fix OS X build
2010-11-08 04:23:25 +00:00
-o $(@)
add support for building with MSVC on Windows
2009-08-27 00:26:44 +00:00
endif
fix darwin build
2008-03-31 04:23:51 +00:00
$(strip) $(strip-all) $(@)
support building Avian as a self-contained dynamic library
2008-03-31 03:43:43 +00:00
fix GCC 4.6 LTO build On Ubuntu 11.10, the optimized build was breaking, apparently because it was eliminating most of the symbols defined in assembly code (e.g. vmJump) as unreachable when linking libjvm.so, which left avian-dynamic unlinkable due to an unresolved symbol. The solution in this commit is to export makeSystem and makeFinder from libjvm.so rather than build redundant versions of finder.cpp and posix.cpp/windows.cpp into avian-dynamic like we've been doing. This avoids the whole problem of vmJump reachability and reduces the size of avian-dynamic at the same time. This commit also turns off LTO for the avian-dynamic link since we get odd undefined symbol errors about libc-defined symbols otherwise. This may merit future investigation, but avian-dynamic is so small and simple that there's no need to optimize it anyway.
2011-11-20 02:26:25 +00:00
# todo: the $(no-lto) flag below is due to odd undefined reference errors on
# Ubuntu 11.10 which may be fixable without disabling LTO.
implement -jar option in main.cpp
2010-11-27 21:46:07 +00:00
$(executable-dynamic): $(driver-dynamic-objects) $(dynamic-library)
support building Avian as a self-contained dynamic library
2008-03-31 03:43:43 +00:00
@echo "linking $(@)"
Windows Phone 8 / Windows RT initial support Conflicts: makefile
2013-01-28 15:20:52 +00:00
ifdef ms_cl_compiler
build everything (including classes) in options-specific directory Previously, we only built platform specific code in an options-specific directory (e.g. build/linux-x86_64-debug), but built the Java classes in the shared, top-level build directory. This minimized duplication when building with different sets of options, but now that we're supporting the option of selecting a third-party class library, we need to put the classes in the options-specific directory as well.
2010-09-17 22:03:37 +00:00
$(ld) $(lflags) -LIBPATH:$(build) -DEFAULTLIB:$(name) \
fix msvc build
2013-03-04 18:09:59 +00:00
-debug -PDB:$(subst $(exe-suffix),.pdb,$(@)) \
Allow output of exceptions to debugger. Generate WinMD file
2013-01-30 08:42:05 +00:00
$(driver-dynamic-objects) -out:$(@) $(manifest-flags)
Windows Phone 8 / Windows RT initial support Conflicts: makefile
2013-01-28 15:20:52 +00:00
ifdef mt
$(mt) -nologo -manifest $(@).manifest -outputresource:"$(@);1"
endif
add support for building with MSVC on Windows
2009-08-27 00:26:44 +00:00
else
don't specify -Wl,-rpath argument on Darwin or Windows (take 2)
2013-02-12 16:27:47 +00:00
$(ld) $(driver-dynamic-objects) -L$(build) -ljvm $(lflags) $(no-lto) $(rpath) -o $(@)
add support for building with MSVC on Windows
2009-08-27 00:26:44 +00:00
endif
support building Avian as a self-contained dynamic library
2008-03-31 03:43:43 +00:00
$(strip) $(strip-all) $(@)
add optional LZMA support for compressing embedded JARs, boot images, and shared objects
2012-06-02 15:06:22 +00:00
$(generator): $(generator-objects) $(generator-lzma-objects)
support classloading from jar files compiled into the executable
2007-10-25 22:06:05 +00:00
@echo "linking $(@)"
statically link libstdc++ to type-generator on windows
2014-07-14 15:37:37 +00:00
$(build-ld-cpp) $(^) $(build-lflags) $(static-on-windows) -o $(@)
enable standalone OpenJDK builds As described in readme.txt, a standalone OpenJDK build embeds all libraries, classes, and other files needed at runtime in the resulting binary, eliminating dependencies on external resources.
2010-11-04 17:02:09 +00:00
avoid naming conflicts in OpenJDK build of libavian.a OpenJDK's compile.c and Avian's compile.cpp were both being compiled to compile.o, which led to a conflict when building libavian.a. We now append "-openjdk" to the object file name for OpenJDK code to avoid such conflicts.
2011-04-14 22:12:12 +00:00
$(openjdk-objects): $(build)/openjdk/%-openjdk.o: $(openjdk-src)/%.c \
enable standalone OpenJDK builds As described in readme.txt, a standalone OpenJDK build embeds all libraries, classes, and other files needed at runtime in the resulting binary, eliminating dependencies on external resources.
2010-11-04 17:02:09 +00:00
$(openjdk-headers-dep)
@echo "compiling $(@)"
@mkdir -p $(dir $(@))
fix windows cross openjdk-src build This mainly involves some makefile ugliness to work around bugs in the native Windows OpenJDK code involving conflicting static and not-static declarations which GCC 4.0 and later justifiably reject but MSVC tolerates.
2010-12-02 03:05:22 +00:00
sed 's/^static jclass ia_class;//' < $(<) > $(build)/openjdk/$(notdir $(<))
rename platform=darwin -> platform={ios,macosx}, fix ios simulator build
2014-06-22 18:30:52 +00:00
ifeq ($(platform),ios)
heretofore untested iOS OpenJDK support
2012-03-11 23:46:36 +00:00
sed \
-e 's/^#ifndef __APPLE__/#if 1/' \
-e 's/^#ifdef __APPLE__/#if 0/' \
< "$(openjdk-src)/solaris/native/java/lang/ProcessEnvironment_md.c" \
> $(build)/openjdk/ProcessEnvironment_md.c
sed \
-e 's/^#ifndef __APPLE__/#if 1/' \
-e 's/^#ifdef __APPLE__/#if 0/' \
< "$(openjdk-src)/solaris/native/java/lang/UNIXProcess_md.c" \
> $(build)/openjdk/UNIXProcess_md.c
fix iOS build when using recent versions of OpenJDK 7 This is based on recent feedback on the Avian group. I've tested it using both OpenJDK 7u40 and 7u60.
2014-07-24 17:20:17 +00:00
if [ -e "$(openjdk-src)/solaris/native/java/lang/childproc.h" ]; then \
sed \
-e 's/^#ifndef __APPLE__/#if 1/' \
-e 's/^#ifdef __APPLE__/#if 0/' \
< "$(openjdk-src)/solaris/native/java/lang/childproc.h" \
> $(build)/openjdk/childproc.h; \
fi
heretofore untested iOS OpenJDK support
2012-03-11 23:46:36 +00:00
endif
prevent openjdk-src build from trying to load the installed 'JavaRuntimeSupport' (fixes #128)
2013-12-08 01:49:23 +00:00
if [ -f openjdk-patches/$(notdir $(<)).patch ]; then \
( cd $(build) && patch -p0 ) < openjdk-patches/$(notdir $(<)).patch; \
fi
fix Cygwin/MSYS openjdk-src builds
2010-11-15 23:27:00 +00:00
$(cc) -fPIC $(openjdk-extra-cflags) $(openjdk-cflags) \
fix windows cross openjdk-src build This mainly involves some makefile ugliness to work around bugs in the native Windows OpenJDK code involving conflicting static and not-static declarations which GCC 4.0 and later justifiably reject but MSVC tolerates.
2010-12-02 03:05:22 +00:00
$(optimization-cflags) -w -c $(build)/openjdk/$(notdir $(<)) \
fix openjdk-src build failures on osx 10.9
2013-12-07 23:38:00 +00:00
$(call output,$(@)) -Wno-return-type
enable standalone OpenJDK builds As described in readme.txt, a standalone OpenJDK build embeds all libraries, classes, and other files needed at runtime in the resulting binary, eliminating dependencies on external resources.
2010-11-04 17:02:09 +00:00
avoid naming conflicts in OpenJDK build of libavian.a OpenJDK's compile.c and Avian's compile.cpp were both being compiled to compile.o, which led to a conflict when building libavian.a. We now append "-openjdk" to the object file name for OpenJDK code to avoid such conflicts.
2011-04-14 22:12:12 +00:00
$(openjdk-local-objects): $(build)/openjdk/%-openjdk.o: $(src)/openjdk/%.c \
various fixes to get Eclipse 3.6 working * add libnet.so and libnio.so to built-in libraries for openjdk-src build * implement sun.misc.Unsafe.park/unpark * implement JVM_SetClassSigners/JVM_GetClassSigners * etc.
2010-11-27 18:25:02 +00:00
$(openjdk-headers-dep)
@echo "compiling $(@)"
@mkdir -p $(dir $(@))
$(cc) -fPIC $(openjdk-extra-cflags) $(openjdk-cflags) \
$(optimization-cflags) -w -c $(<) $(call output,$(@))
fix Cygwin/MSYS openjdk-src builds
2010-11-15 23:27:00 +00:00
$(openjdk-headers-dep):
enable standalone OpenJDK builds As described in readme.txt, a standalone OpenJDK build embeds all libraries, classes, and other files needed at runtime in the resulting binary, eliminating dependencies on external resources.
2010-11-04 17:02:09 +00:00
@echo "generating openjdk headers"
@mkdir -p $(dir $(@))
$(javah) -d $(build)/openjdk -bootclasspath $(boot-classpath) \
$(openjdk-headers-classes)
fix Windows openjdk-src build
2010-11-30 00:39:41 +00:00
ifeq ($(platform),windows)
sed 's/^#ifdef _WIN64/#if 1/' \
< "$(openjdk-src)/windows/native/java/net/net_util_md.h" \
> $(build)/openjdk/net_util_md.h
fix Cygwin 1.7.x build Recent versions of Cygwin's GCC no longer support the -mno-cygwin flag (compilation works, but linking fails). However, Cygwin now includes mingw64-i686-gcc, mingw64-x86_64-gcc, and related packages, so we use those compilers instead. This has the added benefit that we can build native x86_64 binaries on Cygwin now. This will break the build for old Cygwin installs, so it will be necessary to upgrade Cygwin and install the aforementioned packages to build Avian going forward.
2011-07-12 22:58:51 +00:00
sed \
fix MinGW (non-w64) build Commit a2dca0dc fixed the MinGW-w64 openjdk-src build but broke the traditional MinGW build. This should ensure that both work.
2011-09-17 00:42:18 +00:00
-e 's/\(^#include "net_util.h"\)/\1\n#if (defined _INC_NLDEF) || (defined _WS2DEF_)\n#define HIDE(x) hide_##x\n#else\n#define HIDE(x) x\n#define _WINSOCK2API_\n#endif/' \
-e 's/\(IpPrefix[a-zA-Z_]*\)/HIDE(\1)/' \
-e 's/\(IpSuffix[a-zA-Z_]*\)/HIDE(\1)/' \
-e 's/\(IpDad[a-zA-Z_]*\)/HIDE(\1)/' \
-e 's/\(ScopeLevel[a-zA-Z_]*\)/HIDE(\1)/' \
-e 's/\(SCOPE_LEVEL[a-zA-Z_]*\)/HIDE(\1)/' \
fix Cygwin 1.7.x build Recent versions of Cygwin's GCC no longer support the -mno-cygwin flag (compilation works, but linking fails). However, Cygwin now includes mingw64-i686-gcc, mingw64-x86_64-gcc, and related packages, so we use those compilers instead. This has the added benefit that we can build native x86_64 binaries on Cygwin now. This will break the build for old Cygwin installs, so it will be necessary to upgrade Cygwin and install the aforementioned packages to build Avian going forward.
2011-07-12 22:58:51 +00:00
< "$(openjdk-src)/windows/native/java/net/NetworkInterface.h" \
> $(build)/openjdk/NetworkInterface.h
fix windows cross openjdk-src build This mainly involves some makefile ugliness to work around bugs in the native Windows OpenJDK code involving conflicting static and not-static declarations which GCC 4.0 and later justifiably reject but MSVC tolerates.
2010-12-02 03:05:22 +00:00
echo 'static int getAddrsFromAdapter(IP_ADAPTER_ADDRESSES *ptr, netaddr **netaddrPP);' >> $(build)/openjdk/NetworkInterface.h
heretofore untested iOS OpenJDK support
2012-03-11 23:46:36 +00:00
endif
fix openjdk-src build failures on osx 10.9
2013-12-07 23:38:00 +00:00
rename platform=darwin -> platform={ios,macosx}, fix ios simulator build
2014-06-22 18:30:52 +00:00
ifeq ($(kernel),darwin)
heretofore untested iOS OpenJDK support
2012-03-11 23:46:36 +00:00
mkdir -p $(build)/openjdk/netinet
for file in \
fix openjdk-src build failures on osx 10.9
2013-12-07 23:38:00 +00:00
$(sysroot)/usr/include/netinet/ip.h \
$(sysroot)/usr/include/netinet/in_systm.h \
$(sysroot)/usr/include/netinet/ip_icmp.h \
$(sysroot)/usr/include/netinet/in_var.h \
$(sysroot)/usr/include/netinet/icmp6.h \
$(sysroot)/usr/include/netinet/ip_var.h; do \
heretofore untested iOS OpenJDK support
2012-03-11 23:46:36 +00:00
if [ ! -f "$(build)/openjdk/netinet/$$(basename $${file})" ]; then \
ln "$${file}" "$(build)/openjdk/netinet/$$(basename $${file})"; \
fi; \
done
mkdir -p $(build)/openjdk/netinet6
for file in \
fix openjdk-src build failures on osx 10.9
2013-12-07 23:38:00 +00:00
$(sysroot)/usr/include/netinet6/in6_var.h; do \
heretofore untested iOS OpenJDK support
2012-03-11 23:46:36 +00:00
if [ ! -f "$(build)/openjdk/netinet6/$$(basename $${file})" ]; then \
ln "$${file}" "$(build)/openjdk/netinet6/$$(basename $${file})"; \
fi; \
done
mkdir -p $(build)/openjdk/net
for file in \
fix openjdk-src build failures on osx 10.9
2013-12-07 23:38:00 +00:00
$(sysroot)/usr/include/net/if_arp.h; do \
heretofore untested iOS OpenJDK support
2012-03-11 23:46:36 +00:00
if [ ! -f "$(build)/openjdk/net/$$(basename $${file})" ]; then \
ln "$${file}" "$(build)/openjdk/net/$$(basename $${file})"; \
fi; \
done
mkdir -p $(build)/openjdk/sys
for file in \
fix openjdk-src build failures on osx 10.9
2013-12-07 23:38:00 +00:00
$(sysroot)/usr/include/sys/kern_event.h \
$(sysroot)/usr/include/sys/sys_domain.h; do \
heretofore untested iOS OpenJDK support
2012-03-11 23:46:36 +00:00
if [ ! -f "$(build)/openjdk/sys/$$(basename $${file})" ]; then \
ln "$${file}" "$(build)/openjdk/sys/$$(basename $${file})"; \
fi; \
done
fix Windows openjdk-src build
2010-11-30 00:39:41 +00:00
endif
enable standalone OpenJDK builds As described in readme.txt, a standalone OpenJDK build embeds all libraries, classes, and other files needed at runtime in the resulting binary, eliminating dependencies on external resources.
2010-11-04 17:02:09 +00:00
@touch $(@)
fix Cygwin/MSYS openjdk-src builds
2010-11-15 23:27:00 +00:00
$(openjdk-jar-dep):
enable standalone OpenJDK builds As described in readme.txt, a standalone OpenJDK build embeds all libraries, classes, and other files needed at runtime in the resulting binary, eliminating dependencies on external resources.
2010-11-04 17:02:09 +00:00
@echo "extracting openjdk classes"
@mkdir -p $(dir $(@))
@mkdir -p $(classpath-build)
fix Cygwin/MSYS openjdk-src builds
2010-11-15 23:27:00 +00:00
(cd $(classpath-build) && \
$(jar) xf "$$($(native-path) "$(openjdk)/jre/lib/rt.jar")" && \
$(jar) xf "$$($(native-path) "$(openjdk)/jre/lib/jsse.jar")" && \
$(jar) xf "$$($(native-path) "$(openjdk)/jre/lib/jce.jar")" && \
include charsets.jar in openjdk-src build The other JARs we were including refer to classes in this jar, so we need to include it as well.
2012-03-27 23:58:52 +00:00
$(jar) xf "$$($(native-path) "$(openjdk)/jre/lib/charsets.jar")" && \
fix OpenJDK JCE support This primarily required additions to classpath-openjdk.cpp to intercept ZipFile, ZipEntry, and JarFile native methods to consult embedded encryption policy jars when required.
2011-03-04 23:55:31 +00:00
$(jar) xf "$$($(native-path) "$(openjdk)/jre/lib/ext/sunjce_provider.jar")" && \
fix Cygwin/MSYS openjdk-src builds
2010-11-15 23:27:00 +00:00
$(jar) xf "$$($(native-path) "$(openjdk)/jre/lib/resources.jar")")
enable standalone OpenJDK builds As described in readme.txt, a standalone OpenJDK build embeds all libraries, classes, and other files needed at runtime in the resulting binary, eliminating dependencies on external resources.
2010-11-04 17:02:09 +00:00
@touch $(@)
Reference in New Issue Copy Permalink