I was trying to build static binaries for a range of Broadcom soft-float ARMv7
SoCs and kept getting SIGILL, although I was targeting Cortex A7 (and A5,
later), even on A9 and A15.
I found out that once I add -msoft-float, +mp+sec is to blame:
Attribute Section: aeabi
File Attributes
- Tag_CPU_name: "7VE"
+ Tag_CPU_name: "7"
Tag_CPU_arch: v7
- Tag_CPU_arch_profile: Application
- Tag_ARM_ISA_use: Yes
Tag_THUMB_ISA_use: Thumb-2
Tag_ABI_PCS_wchar_t: 4
Tag_ABI_FP_rounding: Needed
@@ -12,8 +10,5 @@ File Attributes
Tag_ABI_FP_number_model: IEEE 754
Tag_ABI_align_needed: 8-byte
Tag_ABI_enum_size: int
Tag_ABI_optimization_goals: Aggressive Size
Tag_CPU_unaligned_access: v6
- Tag_MPextension_use: Allowed
- Tag_DIV_use: Allowed in v7-A with integer division extension
- Tag_Virtualization_use: TrustZone and Virtualization Extensions
(This is the readelf -A diff, before and after armv7-a+nofp -> armv7+nofp).
I kept getting SIGILL even after building my application with a toolchain built
with the correct CFLAGS and found out that crosstool-ng doesn't pass the host
CFLAGS when building musl, which pollutes my binary with these ARMv7 extensions.
Signed-off-by: Dima Krasner <dima@dimakrasner.com>
The glibc will append the content of the CFLAGS variable,
overriding previous flags.
If unset, the CFLAGS variable is not empty, so explicitly set it.
Instead prepend the default CFLAGS flags.
Signed-off-by: Norbert Lange <nolange79@gmail.com>
Fixes: #887
On some systems the file command identifies a pie executable as a shared
object. Update do_finish() to handle this case so that they are stripped
as well.
Signed-off-by: Chris Packham <judge.packham@gmail.com>
From GCC's standpoint ARC's multilib items are defined by "mcpu" values
which we have quite a few and for all of them might be built optimized
cross-toolchain.
From Glibc's standpoint multilib is just multi-ABI [1] and so very limited
versions are supposed to co-exist (e.g. arc700 & archs).
Here we force Glibc to install libraries in GCC's multilib folder to create
a universal cross-toolchain that has libs optimized for multiple CPU types.
But note we only need to mess with installation paths in case of real
multilib, otherwise we keep default "lib/" paths so that GCC finds default
(the one and only) libs where it expects them to be.
Also here we add a sample which allows to build universal Glibc Linux
toolchain for ARC.
[1] https://sourceware.org/ml/libc-alpha/2019-06/msg00018.html
Signed-off-by: Alexey Brodkin <abrodkin@synopsys.com>
... resulted in an attempt to build libinproctrace.so whenever any
of the {gdbserver, native gdb} was enabled.
Signed-off-by: Alexey Neyman <stilor@att.net>
uclibc_backend_once tries to build dummy shared libraries regardless of
whether shared libraries support for target is enabled or not, resulting
in build failure in noMMU bFLT configuration.
Only build dummy shared libraries when shared library support for target
is enabled.
Signed-off-by: Max Filippov <jcmvbkbc@gmail.com>
scripts/build/kernel/linux.sh only provides suffix to known
architectures when building toolchain targeting noMMU linux.
Add support for xtensa and assign uclinux suffix to it.
Signed-off-by: Max Filippov <jcmvbkbc@gmail.com>
... when it is compiled without the native GDB.
Also, fix the gdbserver to be installed without a program prefix in this
case, as it was before the unification of the GDB backend.
Signed-off-by: Alexey Neyman <stilor@att.net>
- Force building make as a companion tool if host make is older than
4.0 (CentOS 7 currently has 3.82)
- Disable 2.29 as a choice if host python is older than 3.4
(CentOS 7 has 2.6 unless python from EPEL is installed)
- Python2 emits its version information to STDERR. Ugh.
While there, also use the detected host Python for GDB configuration.
Signed-off-by: Alexey Neyman <stilor@att.net>
This fixes following build error on Debian 9:
configure: error: Your local docbook2man was found to work with SGML rather
than XML. Please install docbook2X and use variable DOCBOOK_TO_MAN to point
configure to command docbook2x-man of docbook2X.
Or use DOCBOOK_TO_MAN="xmlto man --skip-validation" if you have xmlto around.
You can also configure using --without-docbook if you can do without a man
page for xmlwf.
Signed-off-by: Bernhard Walle <bernhard@bwalle.de>
uClibc-ng 1.0.31 enabled FDPIC as an option for ARM/no-MMU
configurations and defaults to that option if not set explicitly.
Signed-off-by: Alexey Neyman <stilor@att.net>
... in the backend function with ${CFLAGS_FOR_HOST}. The caller either
supplies them already, or (in case of pass-1/2 of the canadian cross)
passes ${CFLAGS_FOR_BUILD} there.
Signed-off-by: Alexey Neyman <stilor@att.net>
Otherwise, glibc 2.29 tries to use it - but we haven't built libstdc++ yet.
We really need to implement #808... Until now, pass empty CXX to make.
Signed-off-by: Alexey Neyman <stilor@att.net>
... which, after a recent change, is not reflected into CT_ALL_TARGET_CFLAGS
for non-multilib configurations.
Signed-off-by: Alexey Neyman <stilor@att.net>
... parts of the config tuple. While here, remove parts that are
setting portions of the target tuple to a value that's already
the default.
Signed-off-by: Alexey Neyman <stilor@att.net>
... in uClibc and glibc.
Fixes#681.
While here, relocate additional "sources" for uClibc/binutils into packages/
directory.
Signed-off-by: Alexey Neyman <stilor@att.net>
This required some rework of the libc selection, as moxiebox is a layer on
top of another libc - newlib.
Also, moxiebox'es host VM (`sandbox`) needs a libcrypto on the host. We will
not have it if we're cross-compiling a canadian cross. Fortunately, all moxiebox
needs from libcrypto is SHA256, and it already includes a standalone implementation
of SHA256 in its runtime. Provide a little wrapper that allows moxiebox use
that implementation for the host binary, too.
Also, automate collecting/printing the list of all packages in a given category
(e.g. LIBC or COMP_TOOLS), generate a list of all Kconfig symbols for a given
category.
Signed-off-by: Alexey Neyman <stilor@att.net>
... and use that ability to permanently set CTNG_LD_IS in case gold is default linker
and we're building glibc.
Fixes#988. This was a long-standing breakage in crosstool-NG (at least since it began
to run each step in a sub-shell).
Signed-off-by: Alexey Neyman <stilor@att.net>
There is a subtle difference when executable bit is a part of the umask.
And at least some versions (Debian/stretch) fail if the resulting mode
would've been different if not for the umask setting.
Fixes#998.
Although, with such chmods/umasks it is likely that some package installation
will break anyway. But I'll leave it until somebody complains.
Signed-off-by: Alexey Neyman <stilor@att.net>
Modify CT_TARGET_CFLAGS (which are passed to GCC's FOR_TARGET flags) rather
than CT_ALL_TARGET_CFLAGS.
Fixes#1006.
Signed-off-by: Alexey Neyman <stilor@att.net>
They're written in ARM dialect, and `ldmia r14, {r14, pc}` is not accepted in T2
encoding. GCC8 changed the list of multilibs for arm-*, which now includes a
variant with CPU that supports T2 but not A1 encoding.
Signed-off-by: Alexey Neyman <stilor@att.net>
In case we build for ARC core which has no support of atomic ops among
other things we need to configure libc to use Linux kernel helper to emulate
HS atomic ops. This is done with disabling of CONFIG_ARC_HAS_ATOMICS in uClibc.
Currently we __remove__ this option from .config but this makes no sense as
its default state is "y" so we need to explicitly disable it instead.
Signed-off-by: Alexey Brodkin <abrodkin@synopsys.com>
Synopsys' DesignWare ARC Processors are a family of 32-bit CPUs
that SoC designers can optimize for a wide range of uses,
from deeply embedded to high-performance host applications in a variety
of market segments.
Signed-off-by: Alexey Brodkin <abrodkin@synopsys.com>
- Incompatible function type for ifunc alias
- Multiple statements macro expansion in strftime
- if_nametoindex size checking
Signed-off-by: Alexey Neyman <stilor@att.net>
... in the corresponding /lib directory. Mingw-w64 installs it to /bin,
so multiple variants in a multilib configuration override each other.
Signed-off-by: Alexey Neyman <stilor@att.net>
... which now defaults to --enable-mpers=yes, which attempts to
invoke aarch64-*-gcc with -m32 and fails.
Signed-off-by: Alexey Neyman <stilor@att.net>
This fixes gccs LIMITS_H_TEST detection for baremetal targets
so limits.h will be installed correctly.
Signed-off-by: Michael Zimmermann <sigmaepsilon92@gmail.com>
- Update to 20180129
- Throw in --disable-db-install if database is disabled; otherwise
'make install' tries to run tic which is not built.
- Select appropriate strip utility for the host; otherwise non-x86
architectures fail to install (unless --disable-stripping is also
added)
Signed-off-by: Alexey Neyman <stilor@att.net>
These changes mainly fix static linking errors when building static
native gdb and gdbserver (tested with gcc 7.2.0 + uClibc-ng 1.0.27 +
binutils 2.29.1 for MIPS):
[ALL ] .../lib/libstdc++.a(eh_throw.o): In function `__cxa_throw':
[ALL ] (.text.__cxa_throw+0x64): undefined reference to `_Unwind_RaiseException'
[ALL ] (.text.__cxa_throw+0x6c): undefined reference to `_Unwind_RaiseException'
[ALL ] .../lib/libstdc++.a(eh_throw.o): In function `__cxa_rethrow':
[ALL ] (.text.__cxa_rethrow+0x78): undefined reference to `_Unwind_Resume_or_Rethrow'
[ALL ] (.text.__cxa_rethrow+0x80): undefined reference to `_Unwind_Resume_or_Rethrow'
...
The problem is in mixing of CPP, CC, CXX, and LD with CPPFLAGS, CFLAGS,
CXXFLAGS, and LDFLAGS before passing to configure scripts.
gcc is sensitive to argument order and the scripts are normally responsible
to combine the variables in a proper way.
Signed-off-by: Sergey Korolev <s.korolev@ndmsystems.com>
... when using musl to compile strace.
Also, honor CT_TARGET_CFLAGS in scripts compiling target libs/binaries.
Signed-off-by: Alexey Neyman <stilor@att.net>
- Glibc configure args and tuple need adjustment on SuperH
- Only allow "both endian" and "with CPU" for unspecified arch
variant. May reconsider endianness (was breaking things before
adjusting glibc tuple)
- Retire non-multilib sample, it should be a subset of the
multilib one now.
Signed-off-by: Alexey Neyman <stilor@att.net>
1. On SuperH, configuring GCC with explicit variant of the CPU
(like "sh4") limits the default set of multilibs to just that CPU
and requires --with-multilib-list to change. Allow for "unspecified"
variant, so that we can defer to GCC to determine the list.
2. Support toolchains with both endiannesses at the same time.
3. Add a SuperH/newlib sample
4. Add more flags processing for uClibc
Signed-off-by: Alexey Neyman <stilor@att.net>
Some users (like myself) may want to omit the crosstool-NG version
from the binaries' versioning output, as it can be incredibly long
and not too helpful. Add a config option to disable it. The possible
combinations are as follows:
- crosstool-NG version (default)
- crosstool-NG version - custom toolchain ID
- Custom toolchain ID
- No crosstool-NG version OR custom toolchain ID
Signed-off-by: Nathan Chancellor <natechancellor@gmail.com>
While here, also consider patched by anything other than "bundled patches"
as per-target sources. Add scary warnings in case of a failure.
Signed-off-by: Alexey Neyman <stilor@att.net>
... because there the symbols were constructer part by part.
Also, remove cc.sh and source $(CT_CC).sh directly - we only build
a single compiler at a time.
Signed-off-by: Alexey Neyman <stilor@att.net>
It fails to compile with the only locale version available (030818)
(on master too, with exactly the same error).
uClibc-ng does not use pregenerated locales.
Signed-off-by: Alexey Neyman <stilor@att.net>
- Need GDB8.0 milestone
- Make uClibc "master" package
- Rename bionic -> android-ndk to match the package name and
support suffixes for archives
Signed-off-by: Alexey Neyman <stilor@att.net>
Must have eabi suffix for GCC to accept it. Also:
- We only have one glibc now, no need to account for eglibc.
- Rename aarch64 samples, eabi suffix does not apply to them
(and ct-ng saveconfig was saving them into a different directory).
Fixes#772.
Signed-off-by: Alexey Neyman <stilor@att.net>
- Update .gitignore, do not place .gitignore into directories installed
in bulk
- Remove executable permissions and shebangs from the scripts that are
supposed to be invoked only via ct-ng frontent; prepend them with $(bash).
Despite what showSamples.sh said, it already has some bashisms.
- Remove --with autotools-dev and override dh_update_autotools_config
to avoid having config.{sub,guess} clobbered with older versions
- Install bash completion where Debian (now) expects it
- Update man page to use .\" as the comment delimiter, instead of
undefined macro (."); also, minor text edits.
- Install kconfig.mk without execute permission.
- Remove shell wrappers from 170-localedef-fix-trampoline.patch, we
do not use that for applying patches
- Revoke execute permissions on 210-expat.sh
- Get flags from dpkg-buildflags if available
Signed-off-by: Alexey Neyman <stilor@att.net>
Make this behavior default in case the core gcc backend is used
for final compiler (i.e., for baremetal configurations). Not
setting this option breaks canadian baremetal configurations,
and not setting it makes little sense at all in any baremetal
configuration (since in baremetal we don't have any libc to begin
with).
Signed-off-by: Alexey Neyman <stilor@att.net>
zlib refuses to run configure with mingw32 host and insists that
win32/Makefile.gcc is used instead.
This requires a change in this Makefile to support static-only builds.
Fixes#694.
Signed-off-by: Alexey Neyman <stilor@att.net>
The generated sysnum.h is different for o32/n32/64 ABIs.
This needs to be revisited; either do this for all architecutres or
perhaps, compare the headers for various multilibs and combine them
if the are identical.
Signed-off-by: Alexey Neyman <stilor@att.net>
Building with CT_MINGW_TOOLS unset before this change produces:
/usr/local/ct-ng/lib/crosstool-ng-1.23.0-rc2/scripts/build/libc/mingw.sh: line 212: [: =: unary operator expected
Check for python2/python3 and if found, pass them to --with-python.
Allow user to override the choice via a new config option. This
fixes systems where there is no "python", only "python2" or "python3".
Signed-off-by: Alexey Neyman <stilor@att.net>
If the build machine lacks tic, we need to build it in the first pass
even if host==build: ncurses Makefiles are not smart enough to build
'tic' first and use the just-built tic to compile fallback terminfo.
Signed-off-by: Alexey Neyman <stilor@att.net>
On cygwin, creating both "foo.exe" and "foo" results in 'ln -sf'
returning an error ("File exists"). However, ln silently removes
the "foo.exe" in this case, so an attempt to re-run the same command
manually then succeeds.
Hence, make binutils.sh also create symlinks with .exe prefix,
using the new & shiny routine.
Signed-off-by: Alexey Neyman <stilor@att.net>
- libpthread requires iteration over multilibs, unlike the core, it
does not detect and build multilibs by itself.
- Disable parallel builds for mingw-w64 components; until mingw-w64 core
builds clean, I am not trusting it.
- Make the list of tools to build configurable
- Turn on multilib in x86_64 sample.
- Make warnings about tuple less redundant. As in, "one WARN is enough,
no need to shout it three times".
- Messages about various steps/substeps are more aligned with the rest
of the components.
- Use 'make' instead of ${make} to invoke the companion make just built,
if applicable.
Signed-off-by: Alexey Neyman <stilor@att.net>
(see the comments in the code for details on the issue)
Old workaround in 100-gcc.sh stopped working (probably, due to one
of GCC version upgrades), so switch to the other approach originally
described there: adjust the list of multilibs to not include the
default target explicitly.
Signed-off-by: Alexey Neyman <stilor@att.net>
... enabled by default for multilib and disabled otherwise. Buildroot
has been complaining about /etc/ld.so.conf presence for almost a year
now and I missed that.
After the release, xldd will be modified to query the compiler for
the list of multilibs to search. This would be too invasive change
before 1.23, though.
Note that it may lead to configurations where xldd currently does not
find the libraries (if both DEMULTILIB and CREATE_LDSO_CONF are turned
off). This is not the default setting in Kconfig, though.
Signed-off-by: Alexey Neyman <stilor@att.net>
It turns out buildroot does not currently accept a toolchain where a dynamic
linker does not reside in the multi-os-directory. Unfortunately this is
how glibc installs itself on AArch64 without any extra tricks.
So, provide an option to force everything into /lib or /usr/lib; patch to
buildroot will be worked on separately.
Signed-off-by: Alexey Neyman <stilor@att.net>
Also a fix for CT_IterateMultilibs: it didn't pass multi_os_dir_gcc, so
it only worked if the caller did *not* declare it as a local variable.
Signed-off-by: Alexey Neyman <stilor@att.net>
Convert absolute targets to relative so that they are valid on the host,
too. The procedure is very similar to uclibc, so it is moved into a
common function.
Signed-off-by: Alexey Neyman <stilor@att.net>
... and in addition to final toolchain aliasing, use it when configuring
multilibs for glibc/musl. Note that uClibc does not need it, it is
explicitly selecting the tools using CROSS_PREFIX.
Signed-off-by: Alexey Neyman <stilor@att.net>
make's configure uses pkg-config to detect if Guile should be enabled;
on ArchLinux, this picks up Guile from build machine's pkgconfig and then
it fails to compile.
A better solution might be to create a ${CT_HOST}-pkg-config in
buildtools/bin that would report "unsupported" for all packages.
However a quick grep only showed pkg-config being used by GCJ
(not sure if it will build in canadian cross - we don't have any
samples with GCJ) and Blackfin simulator in GDB (Blackfin is not
currently supported by crosstool-ng). Hence, leave such pkg-config
implementation and testing for another day.
Signed-off-by: Alexey Neyman <stilor@att.net>
