${CT_PREFIX_DIR} may contain relative paths (e.g.
"${CT_TOP_DIR}/../x-tools/${CT_TARGET}"). Code added in commit d83a0036
("Add symlinks from the libraries' original location to the new one.")
didn't cope well with this. As we're already calculating
${cannon_prefix} make use of it when adding the symlinks. This avoids
any issues with ${CT_PREFIX_DIR}.
Fixes#1807
Signed-off-by: Chris Packham <judge.packham@gmail.com>
This commit adds architecture support for LoongArch.
The toolchain currently only supports the 64-bit target
loongarch64-unknown-linux-gnu.
It has been tested to build with GCC 12.1, GDB 12.1, Glibc 2.36, Linux
5.19 and Binutils 2.39 as of Aug 2022.
Signed-off-by: Jiajie Chen <c@jia.je>
libsanitizer is only supported on selected architectures. Add
ARCH_SUPPORTS_LIBSANITIZER and have architectures select this option
based on the list of supported configurations from GCC's
libsanitizer/configure.tgt. Support for mips64 was added in GCC12 so
this is an additional condition for the mips architecture.
Fixes#1733
Signed-off-by: Chris Packham <judge.packham@gmail.com>
CT_CC_{CORE_}SYSROOT_ARG is being used both as an array and string. Switch everything
to be used as an array for consistency.
Signed-off-by: Justin Chen <justinpopo6@gmail.com>
After building the cross toolchain and compile a simple C program,
objdump reports debug information even though -g modifier was not
used. These debug segments are in glibc library and CRT files. So
a new config entry: CT_GLIBC_ENABLE_DEBUG was added to control when
we want to add debug information to glibc, by default is enabled.
Signed-off-by: Guillermo E. Martinez <guillermo.e.martinez@oracle.com>
When core gcc older than v5 is compiled it shows the error message:
Build failed in step 'Installing core C gcc compiler'
called in step '(top-level)'
Error happened in: CT_DoExecLog[scripts/functions@376]
called from: do_gcc_core_backend[scripts/build/cc/gcc.sh@627]
called from: do_cc_core[scripts/build/cc/gcc.sh@210]
called from: main[scripts/crosstool-NG.sh@697]
configure: error: in
`.../build/build-cc-gcc-core/fixincludes':
configure: error: C compiler cannot create executables
This patch disable `all-build-libcpp' target when core gcc
older than v5 is configured.
Signed-off-by: Guillermo E. Martinez <guillermo.e.martinez@oracle.com>
When GMP is being "cross" compiled for the same architecture (i.e. build
== host) it does not pick the right compiler. Set CC_FOR_BUILD and
CPP_FOR_BUILD to override the default compiler.
Fixes#1328
Signed-off-by: Chris Packham <judge.packham@gmail.com>
Add GCC 12.1 https://gcc.gnu.org/gcc-12/
The following patches from GCC 11.3.0 are no longer needed:
- 0005-arc-Update-ZOL-pattern.patch
- 0006-arc-Update-u-maddhisi4-patterns.patch
- 0007-arc-Fix-maddhisi-patterns.patch
- 0008-Darwin-aarch64-Initial-support-for-the-self-host-dri.patch
- 0009-libstdc-Check-for-TLS-support-on-mingw-cross-compile.patch
One new patch is needed to avoid issues building sh-unknown-elf:
- 0006-sh-Avoid-mb-m1-multilib-combination.patch
It is also necessary to build all-build-libcpp. This target exists as
far back as GCC 6 so has been done unconditionally.
Signed-off-by: Chris Packham <judge.packham@gmail.com>
Commit 6d5227b6 ("Remove obsolete glibc 2.12.1") removed supports for
the separate glibc-ports but also removed GLIBC_USE_PORTS_ADDON. Glibc
versions up to 2.20 bundled support from some architectures in the ports
directory so GLIBC_USE_PORTS_ADDON is required to support these older
glibc versions.
Fixes#1736
Signed-off-by: Chris Packham <judge.packham@gmail.com>
Add duma 2.5.21 and mark 2.5.15 as obsolete. While we're at it use the
versions hosted on github which requres new checksums for the 2.5.15
version because the generated tarballs are different.
It appears we don't need any of the patches we've been carrying for the
older version but we do need to pass CC_FOR_BUILD in addition to HOSTCC.
When 2.5.15 is removed we can drop HOSTCC (and DUMA_CPP, DUMA_SO).
Signed-off-by: Chris Packham <judge.packham@gmail.com>
When `CT_NEWLIB_NANO_INSTALL_IN_TARGET=y`, the `nano.specs` file
emitted by the newlib-nano build script contains an invalid include
path, resulting in the full `newlib.h` being included instead of the
nano `newlib.h` by the application.
`=/include/newlib-nano` is not a valid path (`=` does not mean anything
and that string is taken as an include path as-is) and GCC ignores this
include path, resulting in application including the `newlib.h` from
`include/` which contains the newlib build configurations for the full
newlib.
This commit modifies the newlib-nano build script to emit a proper
newlib-nano include path relative to the `GCC_EXEC_PREFIX`.
Signed-off-by: Stephanos Ioannidis <root@stephanos.io>
Previously, cc/gcc.sh assumed that liblto_plugin would always be
installed with the ".so" file extension. However, this assumption is
incorrect when the host machine is Windows (".dll") or MacOS (".dylib").
This patch corrects this issue by probing the file extension in similar
fashion to the way adjacent code determines the file extension of
executable binaries.
Signed-off-by: Joel Holdsworth <jholdsworth@nvidia.com>
The "ext" variable is set with the file extension of executable binaries
for a given platform. To improve tidiness, this patch ensures the
variable is always set even when there is no file path.
Signed-off-by: Joel Holdsworth <jholdsworth@nvidia.com>
In do_gcc_core_backend and do_gcc_backend, variables "file" and "ext"
are used to store intermediate values. Previously, these were not
declared local. This patch corrects this issue.
Signed-off-by: Joel Holdsworth <jholdsworth@nvidia.com>
In Bryan Hundven's patch 1ad439907 from 2010, the author added -lstdc++
and -lm to the host gcc build's LDFLAGS, because at the time the linker
did not correctly include these libraries causing the build to fail.
In modern builds, this causes a problem for canadian gcc builds where
the host machine is mingw32-w64 Windows.
Within the gcc build there is the liblto_plugin module. On Windows this
must be built as the "liblto_plugin.dll" dynamic library. However,
libtool cannot produce a dynamic library unless every library dependency
is also a dynamic library, and falls back to producing a libstdc++.a
static library. liblto_plugin does not require libstdc++ - it
does not contain any C++ code, and the dependency would usually be
elided by the linker.
Unfortunately, in this case, crosstool-ng will never build a
libstdc++.dll dynamic library - only a libstdc++.a static library.
Therefore, there will never be a dynamic library version of stdc++ for
libtool to load and then discard.
This patch corrects the issue by removing "-lstdc++" from LDFLAGS,
because modern versions of gcc are able to correctly include libstdc++
where necessary. It also remove "-lm" for similar reasons.
Signed-off-by: Joel Holdsworth <jholdsworth@nvidia.com>
Drop gdb 7.11.1, 7.12.1, 8.0.1, 8.1.1 and 8.2.1. Cleanup milestones
related to these older versions.
Signed-off-by: Chris Packham <judge.packham@gmail.com>
glibc 2.12.1 was marked as obsolete. Now that the 1.25.0 release is out
this version can be removed completely. As glibc 2.12.1 was the last
remaining version supported by glibc-ports support for glibc-ports is
also removed.
Signed-off-by: Chris Packham <judge.packham@gmail.com>
The bionic libc support was out of date and relied on downloading
binaries from the internet. It was already marked as obsolete. Now that
the 1.25.0 release is out it can be completely removed.
Signed-off-by: Chris Packham <judge.packham@gmail.com>
Rather important option for arm cortex toolchains supporting c++,
avoids pulling in all printf/iostream code by default.
Signed-off-by: Norbert Lange <nolange79@gmail.com>
libdebuginfod is incompatible with static linking so pass
--without-debuginfod when CT_STATIC_TOOLCHAIN is selected.
Fixes#1683
Signed-off-by: Chris Packham <judge.packham@gmail.com>
With libc_headers step before pass-1, there is no need to distinguish
pass-1 and pass-2; they are configured identically (note that with the
current configuration, core pass-2 is only used for win32 - hence, uses
build_libgcc=yes and mode=static).
Signed-off-by: Alexey Neyman <stilor@att.net>
After 557b9d4, libc_start_files and libc_main steps are performed one
after another. It doesn't make sense, especially since some of the libcs
(glibc, uClibc-ng) go to great lengths to first install start files in
the first step, libc_start_files, only to remove them immediately in the
second step, libc_main.
Current build steps also break in the xtensa newlib configurations, as
it needs to install the custom xtensa headers before building the libgcc
and after 557b9d4, the headers are not installed before libgcc is built
in pass-1.
Therefore, finish what 557b9d4 mentioned but did not do: move header
installation into a new step, libc_headers, and combine libc_start_files
and libc_main into a single step.
This also allows to combine the core pass-1/pass-2 steps, to be done in
a subsequent commit.
Signed-off-by: Alexey Neyman <stilor@att.net>
... and the code dependent on them, after the latest wave of obsolete
package removals. This concludes the glorious history of the original
uClibc (non-NG) with lots of kludges removed.
There was a choice here, whether to call the resulting libc "uClibc" or
"uClibc-ng". I opted in favor of giving uClibc-ng the recognition it
deserves, although it had some ripple effect in the ct-ng code.
Signed-off-by: Alexey Neyman <stilor@att.net>
Per https://github.com/crosstool-ng/crosstool-ng/issues/808 build static
libgcc in the first pass which lets us skip the second one. Building
mingw-w64 requires header files in order to build C++ support so mingw
builds core pass 2. This could probably be cleaned up by splitting
libc_start_files into a separate libc_header step. But for now having
core 2 for mingw-w64 and core 1 for the other libcs will have to do.
Anything that previously selected CC_CORE_PASSES_NEEDED now selects
CC_CORE_PASS_1_NEEDED. The same goes for CC_CORE_PASS_2_NEEDED with the
exception of mingw-w64.
Fixes#808Fixes#217
Signed-off-by: Chris Packham <judge.packham@gmail.com>
The following versions were marked obsolete in crosstool-ng-1.24.0,
remove them.
- binutils-linaro-2.23.2-2013.10-4
- binutils-linaro-2.24.0-2014.11-2
- binutils-linaro-2.25.0-2015.01-2
- binutils-2.23.2
- binutils-2.24
- binutils-2.25.1
Adjust the milestones now that the old versions have been removed.
Signed-off-by: Chris Packham <judge.packham@gmail.com>
Now that the oldest supported version of gdb is 7.11.1 we can make some
parts of the build unconditional and remove the associated config vars.
Signed-off-by: Chris Packham <judge.packham@gmail.com>
Back in the day gdbserver was treated as a subproject of GDB and
even was located in "gdb/gdbserver" and so to build gdbserver we had
to go into "gdb/gdbserver" and there run configure. That way full GDB
was out of the picture.
Now starting from GDB 10.1 where gdbserver was promoted to the top-level
we're supposed to run top-level's configure script for all the tools
provided by the unified binutils-gdb tree.
That said if we only want to build gdbserver (and that's what we
want since we build one tool at a build step) we have to be explicit:
----------------->8----------------
--enable-gdbserver --disable--gdb
----------------->8----------------
Ah, and so far we used to build full native GDB when only wanted gdbserver
if it was GDB v10.x ;)
Ironically full native/target GDB also enabled gdbserver by default so
we need to also disable it explicitly with "--disable-gdbserver".
Signed-off-by: Alexey Brodkin <abrodkin@synopsys.com>
Since we have curses built for target anyway now, why don't allow
users to use very convenient pseudo-GUI operating mode?
And while at it, there's no use of TUI in naturally headless gdbserver.
Signed-off-by: Alexey Brodkin <abrodkin@synopsys.com>
"--with-host-libstdcxx" option was removed in GCC 6.x, see [1] because of [2].
So it makes no sense to use it with later GCC versions.
Frankly I don't like that implementation with yet another set of "if XXX",
but since we still support GCC down to 4.8.5 what else we may do?
Well, technically we may keep using things as they are now,
because surprisingly GCC's configure script doesn't mind accepting
meaningless options, but as a person who's looking at differences between
various builds and drill-down to peculiarities of various config
options, I'd prefer to not pollute configure with garbage.
But for all the rest... well, it works now and maybe nodody else cares.
[1] https://gcc.gnu.org/git/?p=gcc.git&a=commit;h=5dc85f7ec719a79ecfbcdd8563b07e5f0f365e65
[2] https://gcc.gnu.org/bugzilla/show_bug.cgi?id=67092
Signed-off-by: Alexey Brodkin <abrodkin@synopsys.com>
GCC 11+ requires compiler being used to support C++11 standard [1].
And while GCC starting from 6.x has C++11 support enabled by default [2],
older versions need to be forced to implement it with "-std=gnu++11" and luckily
GCC's build-system takes care of that:
1. For ${host} compiler - [1]
2. For ${build} compiler - [3, 4]
In a nutshell the configure script tries a couple of options and the one which
helps to build a test source gets appended to CXX (not CXXFLAGS!),
so on say CentOS 7.x with GCC 4.8.5 during cross-compilation of GCC
CXX="x86_64-build_pc-linux-gnu-g++ -std=gnu++11". And all is good.
But in case of canadian cross due to [5] we for some reason* force set
CXX_FOR_BUILD with just a compiler name, effectively overriding all the
magic done by GCC's internals described above.
This leads to a compilation failures like that:
------------------------------------->8----------------------------------
[ALL ] In file included from /usr/include/c++/4.8.2/type_traits:35:0,
[ALL ] from .../HOST-x86_64-apple-darwin14/arc-gcc11-elf/src/gcc/gcc/system.h:244,
[ALL ] from .../HOST-x86_64-apple-darwin14/arc-gcc11-elf/src/gcc/gcc/gengtype.c:26:
[ERROR] /usr/include/c++/4.8.2/bits/c++0x_warning.h:32:2: error: #error This file requires compiler and library support for the ISO C++ 2011 standard. This support is currently experimental, and must be enabled with the -std=c++11 or -std=gnu++11 compiler options.
[ALL ] #error This file requires compiler and library support for the ^
------------------------------------->8----------------------------------
* my guess that [5] was done because back in the day indeed we used to have
"--build=${CT_BUILD} --host=${CT_HOST}" in do_cc_core(). But now after [6]
this is no longer necessary as we use "--build=${CT_BUILD} --host=${CT_BUILD}"
and all is safe and clean. So yet another old quirk goes away - hooray!
[1] https://gcc.gnu.org/git/?p=gcc.git;a=commit;h=5329b59a2e13dabbe2038af0fe2e3cf5fc7f98ed
[2] https://gcc.gnu.org/gcc-6/changes.html
[3] https://gcc.gnu.org/bugzilla/show_bug.cgi?id=96612
[4] https://gcc.gnu.org/git/?p=gcc.git;a=commit;h=7ffcf5d61174dda1f39a623e15f7e5d6b98bbafc
[5] 9c6c090d7b
[6] 08161250ed
Signed-off-by: Alexey Brodkin <abrodkin@synopsys.com>
GCC can support using zstd compression for LTO object files. By default
GCC's configure will enable this if libzstd is installed on the machine
building the toolchain. This may be undesirable if the toolchain is to
be used on a different machine. Add an option to control zstd usage and
set the default to the same as the current behaviour (i.e. auto).
Fixes#1579
Signed-off-by: Chris Packham <judge.packham@gmail.com>
In GDB 10.x gdbserver was promoted to the top-level folder,
see https://sourceware.org/git/?p=binutils-gdb.git;a=commit;h=919adfe8409211c726c1d05b47ca59890ee648f1
Which means it is no longer a subfolder in "gdb" and so we have to
build gdbserver now exactly in the same way as normal native GDB.
One interesting detail is gdbserver doesn't need to deal with target
description in .xml so it doesn't depend on libexpat on target,
thus we need to move libexpat explicit selection from do_gdb_backend()
to its callers when building native [full] gdb as well as cross-gdb
for the host.
Signed-off-by: Alexey Brodkin <abrodkin@synopsys.com>
[cp: support old/new layout, regenerate patches]
Signed-off-by: Chris Packham <judge.packham@gmail.com>
Older ARC700 processors had atomic instructions (AKA llock/scond)
as an option and so quite some "atomic" operations were not possible
w/o OS support, which we implemented - see arc_usr_cmpxchg() in the
Linux kernel.
And in uClibc, which was the only Linux libc back in the day of ARC700
era, it is well supported. Well, uClibc could be configured to support it.
Which is done with CONFIG_ARC_HAS_ATOMICS Kconfig option.
But the problem is there's no check for ARC ISA version in uClibc when
this option gets enabled. That leads to a funny situation when even for
ARCv2 processors (ARC HS3x & HS4x) uClibc tries to utilize
arc_usr_cmpxchg() syscall which is not supported for this newer ISA since
ARCv2 processors have atomic instructions built-in all the time.
So what was happening here we didn't specify additional "-matomic"
CFLAG unless we were targeting exactly those ancient ARC770 processors
(ARC700 + MMUv3 + atomics) and so even for ARCv2 we forced uClibc
to not use built-in atomics.
And even though there're ways to add a smarter solution here to handle
that pretty rare by now case of ARC750 (ARC700 + MMUv2 - atomics),
I suggest we just remove this part completely, leaving a possibility
to add needed option in uClibc-ng's configuration
(I mean "packages/uClibc-ng/config").
Signed-off-by: Alexey Brodkin <abrodkin@synopsys.com>
Changes since v0.5.0:
* Add spec files for am64x SoCs.
* Require Binutils at least version 2.37.
* Require pru-gcc to be installed.
* Remove linker scripts. Instead set memory sizes from specs.
* Activate --gc-sections linker option by default.
* The "--host=pru" configure option must be used instead of "--target=pru.
Signed-off-by: Dimitar Dimitrov <dimitar@dinux.eu>
The spec file was missing replacing various libs like libc, libm, etc
with their nano equiv when CT_NEWLIB_NANO_INSTALL_IN_TARGET=y. Update
the nano.spec file that is generated to rename libc, libm, etc if
CT_NEWLIB_NANO_INSTALL_IN_TARGET=y
Signed-off-by: Kumar Gala <kumar.gala@linaro.org>
