libstdc++ uses assertion macros which use __FILE__ macros and
if (__builtin_expect(!bool(_Condition), false)) \
std::__replacement_assert(__FILE__, __LINE__, __PRETTY_FUNCTION__, \
#_Condition)
This ends up using absolute paths into build tree for the cases where
the charconv header is used, therefore replace the file prefix paths
with on-target paths to make them build dir independent
(From OE-Core rev: 972c50d6e46ee9dfba8b8ea3867ebdbf24001e6e)
Signed-off-by: Khem Raj <raj.khem@gmail.com>
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>
This helps binutils provided tools ar/ranlib/nm to find the LTO linker
plugin automatically as well which makes it equivalent to gcc-ar/gcc-nm/gcc-ranlib
(From OE-Core rev: 7d8d0b90bea7ea01e1e9ab0ff98f22431f68a506)
Signed-off-by: Khem Raj <raj.khem@gmail.com>
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>
This will ensure that ar/ranlib/nm can load the lto linker plugin like
gcc-ar, gcc-nm, gcc-ranlib does, this will let the behaviour match
between gcc wrappers for these tools, this should help LTO builds for
packages
(From OE-Core rev: d6658505089234476c1b35fc08fef1eb4f121e85)
Signed-off-by: Khem Raj <raj.khem@gmail.com>
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>
This directory is scanned by binutils provided ar,ranlib,nm for plugins
that it can load automatically, putting liblto_plugin.so in their means
we do not need gcc-ar, gcc-nm, gcc-ranlib particularly as normal
ar/ranlib/nm tools will work equally well as they can now use this
linker plugin by default
This also mean we can revert back to using ar/ranlib/nm as default
providers for AR/NM/RANLIB on target
(From OE-Core rev: 5aae5812223792d5e5bd57e024de50fbcd1e6da5)
Signed-off-by: Khem Raj <raj.khem@gmail.com>
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>
Current definition of SLIB is actually equal to S but is hardcoded, this
means when we have altered location of S, then the regexp for
DEBUG_PREFIX_MAP will not be effective, which could result in S being
emitted into debug_line sections. Simplify the maps to use S variable
instead of SLIB
Secondly, rename SLIB_NEW to REL_S to make it more appropritate to what
it represents
(From OE-Core rev: 2c8e130adb5d4d55ba732a042ec157498460ee29)
Signed-off-by: Khem Raj <raj.khem@gmail.com>
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>
This fixes an error in do_patch when using "PATCHTOOL = "patch""
error message:
checking file gcc/configure.ac Hunk #1 FAILED at 6602.
(From OE-Core rev: f75a1eddc7c01d9a83d0a1114fc0d6aaa79d8806)
Signed-off-by: Wang Mingyu <wangmy@cn.fujitsu.com>
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>
When building the cross GCC, configure it so that including the host
system include directories is fatal, not just a warning.
(From OE-Core rev: d039d6fbfc9076f081173f6e5cc63c56fe3a5adb)
Signed-off-by: Ross Burton <ross.burton@arm.com>
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>
Extend --enable-poison-system-directories so that it can be an error
instead of a warning by default.
(From OE-Core rev: 1fcaa5f452b2c6b93a6f3dfd57f894cbef5e3379)
Signed-off-by: Ross Burton <ross.burton@arm.com>
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>
In e9e5744ba8b0d43c8b874d365f83071ce20bf0a1, Khem Raj wrote:
> OE does not use the traditional /usr/lib/gcc prefix to store
> gcc-runtime it basically is moved into libdir, however some newer
> files were installed by newer versions of gcc especially libgomp (
> omp.h openacc.h ) into gcclibdir, so we have content in both
> directories, this confuses other tools which are trying to guess the
> gcc installation and its runtime location, since now we have two
> directories, the tools either choose one or other and we get
> inconsistent behavior, e.g. clang for aarch64 uses /usr/lib but same
> clang for riscv64 chose /usr/lib/gcc
> This change ensures that OE ends up with single valid location for gcc
> runtime files
I think that the same thing needs to happen in gcc-sanitizers.inc,
otherwise I get errors like:
| .../recipe-sysroot/usr/include/gpg-error-64.h:884:11: fatal error: sanitizer/lsan_interface.h: No such file or directory
when attempting to compile with sanitizers enabled.
FILES_${PN} needs updating to match too.
(From OE-Core rev: 862b4fac3ee7d951758c8c93462331ad52bf0190)
Signed-off-by: Mike Crowe <mac@mcrowe.com>
Cc: Khem Raj <raj.khem@gmail.com>
Cc: Alexandre Belloni <alexandre.belloni@bootlin.com>
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>
Added HOMEPAGE and DESCRIPTION for recipes with missing decriptions or homepage
[YOCTO #13471]
(From OE-Core rev: bb05814335e7101bfd8df0a11dc18a044e867bed)
Signed-off-by: Dorinda Bassey <dorindabassey@gmail.com>
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>
Upstream has merged a change to handle detection of target overrides so
that they are only applied as required. Backport this patch from
upstream and remove the OE specific patch.
(From OE-Core rev: d430f425e1a5725e93833a652e5c6988576f7347)
Signed-off-by: Nathan Rossi <nathan@nathanrossi.com>
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>
When compiling gcc-runtime for i686 libatomic, libgomp and libitm are
compiled with additional '-march' and '-mtune' overrides. This typically
does not cause build errors (e.g. core2-32 tune), and results in less
optimally tuned outputs. However with specific TUNE_CCARGS (e.g. has
'-mfpmath=sse' but not '-msse*'), the build errors with conflicting
'-mfpmath' options.
This is caused by gcc-runtime specific configure scripts and Makefile
options that override by adding additional '-march'/'-mtune' args. OE
builds gcc-runtime differently to other build systems, and provides
target specific arch/tune options directly. As such, this change adds a
patch for gcc to remove the specific cases where the args are added.
(From OE-Core rev: d39b686cc5fcc89405ffeb560ea64b5d305b0608)
Signed-off-by: Nathan Rossi <nathan@nathanrossi.com>
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>
PREFIXVAR is defined in cppdefaults.c as well which shows up as error
when LTO is enabled to build gcc, therefore rename the one instance to PREFIXVAR1
(From OE-Core rev: 984da023a595a10ae62d4fc2a1f2fe96c4701223)
Signed-off-by: Khem Raj <raj.khem@gmail.com>
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>
These flags are redundant as LTO now enabled in GCC by default for
sometime, Moreover, dropping LTO variable here avoids a conflict with
global LTO variable being added for distrowide LTO support
(From OE-Core rev: 50cd8edc9849d6ab6249d4c49c7e41bb476dcb71)
Signed-off-by: Khem Raj <raj.khem@gmail.com>
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>
It has been almost 3 months since the upgrade to gcc 10. Switch back to
relying on gcc default configuration (-fno-common).
This reverts OE-Core commit 951e859b1e8297970278c539e989b8a6d06a9cb3.
(From OE-Core rev: 4ee733e1551d3960a23a600eb71a01cf7a51fa06)
Signed-off-by: Jacob Kroon <jacob.kroon@gmail.com>
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>
gcc needs -mmusl option to be passed in SDK since we ship crossdk compiler
configured for glibc by default, this helps in creating correct
compiler defaults for musl based SDK compilers
[YOCTO #13459]
(From OE-Core rev: e6fbac7aeabe00d8fb734992012dd629a8527b25)
Signed-off-by: Khem Raj <raj.khem@gmail.com>
Cc: Leon Woestenberg <leon@sidebranch.com>
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>
After gcc 10 this expression needed to be adjusted, its better to use a
bitbake variable which we do have readily available
(From OE-Core rev: ee2beaff701d02736c8244ebc76f6d6ca96a04f4)
Signed-off-by: Khem Raj <raj.khem@gmail.com>
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>
Drop 0036-Enable-CET-in-cross-compiler-if-possible.patch and
pr96130.patch they are already in 10.2
(From OE-Core rev: 35eda14fbdc38ae5e686c2d30a2c1d5569bd3088)
Signed-off-by: Khem Raj <raj.khem@gmail.com>
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>
To quote Zhuang <qiuguang.zqg@alibaba-inc.com>:
"""
A few days ago, I tried to compile a gcc plugin with the toolchain from poky sdk.
It failed with errors about missing header files such as backend.h etc.
After investigation, I found that the problem was brought by a gcc patch:
0012-gcc-Fix-argument-list-too-long-error.patch (which is considered derived from the original patch)
- headers=`echo $(PLUGIN_HEADERS) $$(cd $(srcdir); echo *.h *.def) | tr ' ' '\012' | sort -u`; \
+ headers="$(sort $(PLUGIN_HEADERS) $$(cd $(srcdir); echo *.h *.def))"; \
It changes the commands of install-plugin, making the sorting taken effect before the shell globs.
Thus results in the header files under gcc $(srcdir) being not installed.
By checking log.do_install, we can find that the `headers=' statement to run is incorrect and will not work as expected:
headers="$(cd *.def) *.h ../../../../../../../work-shared/gcc-10.1.0-r0/gcc-10.1.0/gcc/../include/ansidecl.h ...
As the patch says,
"The PLUGIN_HEADERS is too long before sort, so the "echo" can't handle it, ..."
my suggestion is that we can simply take care of PLUGIN_HEADERS
using the original proposed sort.
"""
This fixes the gcc patch as proposed as it does appear its been
broken over time.
(From OE-Core rev: dce28d8ac7fbae487cb6674b91fe2b574036b26d)
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>
mudflap was removed in gcc 4.9.
(From OE-Core rev: 6d649a07cfa0a89448caa67e4ca0a990973961b9)
Signed-off-by: Adrian Bunk <bunk@stusta.de>
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>
Issue has been fixed in valgrind itself
(From OE-Core rev: c4070f3d76e0170cf6ee672a8a9a38e4cdbbcad9)
Signed-off-by: Khem Raj <raj.khem@gmail.com>
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>
libatomic has mind of its own when it comes to setting -march for arm64
which conflicts with -mcpu option we pass from environment in some cases
since we always pass -march/-mcpu in OE, its safe to remove this option
mcpu removal from cortex-a55 is no longer needed since the option
conflict is now removed from libatomic instead
(From OE-Core rev: a5331c5a8bbe63c6c2e56ebec496b28968d4663d)
Signed-off-by: Khem Raj <raj.khem@gmail.com>
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>
This was added recently, but it seems be chewing more than what it
should and causes non glibc packages also depend on it.
(From OE-Core rev: 595d2df62b049e463568ab97cfe26d6df96a18a9)
Signed-off-by: Khem Raj <raj.khem@gmail.com>
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>
For time being change back to -fcommon as default, helps us iron
out other issues, eventually this should be removed as we fix the
packages to work with -fno-common
(From OE-Core rev: 951e859b1e8297970278c539e989b8a6d06a9cb3)
Signed-off-by: Khem Raj <raj.khem@gmail.com>
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>
gcc-runtime/libatomic explicitly add -march=armv8-a+lse for all arch64
but cortex-a55 is armv8.2-a, which essentially conflicts, so let gcc
override it by not forcing the -mcpu option from TUNE_CCARGS
(From OE-Core rev: 882df891e13ce5c64718c364efb9ef2bf189eabf)
Signed-off-by: Khem Raj <raj.khem@gmail.com>
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>
A nativesdk BBCLASSEXTEND was added to gcc-target without realising this
would pass arch=native through to it for x86-64. This heavily optimises
gcc output for the host its running on meaning it can't be reused via
sstate on other machines.
Add class-target overrides here to get the desired behaviour. All
targets have been covered for completeness.
(From OE-Core rev: d0d4853dd773f7bbe0f72e50144559b80e80d67a)
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>
OE uses libiberty from binutils, since its properly compiled as pic
archive and applications and other libraries needing libiberty can
properly link with it.
With this option applied, explicit delete of libiberty headers and
libraries is not required in install step, since they wont get installed
in first place.
(From OE-Core rev: fa8a205c69770d23323c9a06373db958af4b34d3)
Signed-off-by: Khem Raj <raj.khem@gmail.com>
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>
This should help gcc detect and use target ISA on x86_64 machines when
-march is not used on cmdline
[YOCTO #139]
(From OE-Core rev: f9e410521c92e2458ba7e2ca63d28434618b9f25)
Signed-off-by: Khem Raj <raj.khem@gmail.com>
Cc: Ross Burton <ross.burton@intel.com>
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>
The line numbers are influenced by the gcc version on the host used to generate
the code. Remove these to ensure the shipped source code is the same.
(From OE-Core rev: 5c3d3440809e9d76377af653ac8c5307bc1a01b4)
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>
The build host configuration isn't reproducible as it varies depending
on the gcc version of the build host. This information isn't useful on the
target anyway so remove it.
(From OE-Core rev: f9154b2c3eff8434914710ab453e13cf338597ec)
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>
Linux kernel compilation for aarch64 triggers ICE if
CONFIG_CC_OPTIMIZE_FOR_SIZE=y.
The rootcause is GCC bug #91102 'aarch64 ICE on Linux kernel with -Os'.
Apply the fix to 9.2.
(From OE-Core rev: 14f34d32bfdaa752f5043e62750d2e7b92c4b419)
Signed-off-by: Taras Kondratiuk <takondra@cisco.com>
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>
If the OS is not Linux, the code could end up generating a recursive symlink.
This can happen because the same symlink can be created twice in a row. If this
happenes, the second symlink becomes a link to itself within the directory
pointed to by the original link.
In order to prevent this, verify that the destination does not already exist.
(From OE-Core rev: ef28e5a74d939acc98d6e8e9a0efbeb143b8025f)
Signed-off-by: Mark Hatle <mark.hatle@xilinx.com>
Signed-off-by: Mark Hatle <mark.hatle@kernel.crashing.org>
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>
When configuring the cross-candian toolchain for a non-linux target system,
the resulting gcc configuration included:
--enable-initfini-array--without-headers
these should have been two separate options.
(From OE-Core rev: 7b52893632dae7bc9ac75dddc7ad625e19f41050)
Signed-off-by: Mark Hatle <mark.hatle@kernel.crashing.org>
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>
Some baremetal applications might require support from libstdc++
On newlib based toolchains, libstdc++ can be built as a static
library that applications can then link against it.
Pass libsdtc++-(static)dev to LIBC_DEPENDENCIES allowing the
library to be present for cross compilation as well as on
sdk builds.
(From OE-Core rev: 18af9ecef6e247519d8a1573e32208bb69cf81fe)
Signed-off-by: Alejandro Enedino Hernandez Samaniego <alejandro@enedino.org>
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>
While the gnu style for symbol versioning is the most usual,
--enable-symvers[=style] can be provided several values,
gnu, gnu-versioned-namespace, darwin, darwin-export, and sun,
depending on users needs.
Introduce the SYMVERS_CONF variable to allow the user to
configure the symbol versioning in shared libraries.
(From OE-Core rev: f850931173fc210ed25706fd8fbfe0a310f99dfc)
Signed-off-by: Alejandro Enedino Hernandez Samaniego <alejandro@enedino.org>
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>
Even though we do not expect any legacy ( 32bit ) for LE, linux-yocto
does enable the compat code, so enable 32bit support to get that going
(From OE-Core rev: 8e24fd8dcadc6dd13171a2c4cfec6bcff8db772a)
Signed-off-by: Khem Raj <raj.khem@gmail.com>
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>
Backport a patch from latest gcc-9-branch to fix an API change in glibc
2.31, this patch wont be needed when upgrading to 9.3 or 10.x
(From OE-Core rev: d089f84d64b90776623e0e43d237ac3fcc6a0d7d)
Signed-off-by: Khem Raj <raj.khem@gmail.com>
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>
