Pull in:
ports/rename/renameat: Avoid race when renaming files
ports/unix: Add faccessat and faccessat2
ports/access.c: Use EACCES, not EPERM
which includes a fix for rename race issues causing pseudo aborts.
(From OE-Core rev: 79aba3fdce70baa2456de90ba407ce86d0f5e0f0)
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>
(cherry picked from commit 330c232e4f756296331f9026e91ac26fd45f0315)
Signed-off-by: Anuj Mittal <anuj.mittal@intel.com>
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>
Update to a pseudo version which contains some heqader fixes for
glibc 2.33.
(From OE-Core rev: 4a28f22f9d8a0a5329030c68d33c43683c3e7cf3)
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>
(cherry picked from commit c897ac317926b132547578b1f6bd347fe5677dfc)
Signed-off-by: Anuj Mittal <anuj.mittal@intel.com>
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>
There are several thousand files in the test directory which we don't need.
Adding these for the native and target sysroots is a crazy amount of files
to be throwing around needlessly. Delete the files from the sysroot side
of things to tidy up the sysroots and improve performance.
(From OE-Core rev: 653f8b1a8adf59d924028bfc69ead1c7437de11e)
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>
(cherry picked from commit f6bced03011ad1663d68b0322a2f8aeb4d836646)
Signed-off-by: Anuj Mittal <anuj.mittal@intel.com>
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>
The externalsrc class deletes do_patch task which results with:
| ERROR: Task do_create_manifest in <PATH>/python3_3.8.2.bb depends upon
| non-existent task do_patch in <PATH>/python3_3.8.2.bb
Use addtask to define correct order to prevent this error, since addtask
mechanism accepts deleted tasks.
[YOCTO #14151]
(From OE-Core rev: a746d034fa7eaad4f4876fa61c5a8c3c15e211c8)
(From OE-Core rev: 782f7f4f73d7b1e1bd649db1984a63d94d2410fe)
Signed-off-by: Tomasz Dziendzielski <tomasz.dziendzielski@gmail.com>
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>
(cherry picked from commit 8ed8b81af6)
Signed-off-by: Anuj Mittal <anuj.mittal@intel.com>
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>
Pulls in:
pseudo_client: Ensure renames update open fd file paths
pseudo_client.c: Rebuild passwd paths after chroot
which should fix issues seen in apt package index creation, new
binutils and other autobuilder race issues in pseudo amongst other
issues.
(From OE-Core rev: 44d11b56001f40622c055069b0901cc4ae15c76c)
(From OE-Core rev: 0fc140e6aeba27afc5bd7c4e8d486d73fa6e3f2e)
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>
(cherry picked from commit e1caf4f305)
Signed-off-by: Anuj Mittal <anuj.mittal@intel.com>
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>
Option --noline or -L does not handled properly.
So generated code contains #line directives with
file absolute path and prevents to create
reproducible builds.
(From OE-Core rev: e02c02459b35129402b2a77b592896c787e590ff)
Signed-off-by: Oleksiy Obitotskyy <oobitots@cisco.com>
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>
(cherry picked from commit 0bf2cb7dc2123f220accf1542c2ae4c4b4b8275a)
Signed-off-by: Anuj Mittal <anuj.mittal@intel.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: ef87c3609ecf93158685f4d4cdc53754a6102330)
Signed-off-by: Nathan Rossi <nathan@nathanrossi.com>
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>
(cherry picked from commit d430f425e1a5725e93833a652e5c6988576f7347)
Signed-off-by: Anuj Mittal <anuj.mittal@intel.com>
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>
New versions of glibc have an lchmod function so we need to wrap it.
Identified through a reproducibility issue in initramfs-base where
/dev/console created by mknod from coreutils changed permissions
depending on the host distro (mknod used the gnulib wrapper on most
hosts but newer ones used the libc call).
[YOCTO #14162]
(From OE-Core rev: cac0ad6f82c0f10b9ad6a656ef728e894096049a)
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>
(cherry picked from commit 20a645664977530e602e1ac97e8dc0962e730e6c)
Signed-off-by: Anuj Mittal <anuj.mittal@intel.com>
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>
Previously install-sh was used, which is installed by autoconf so isn't
a relevant license.
Also remove S assignment, as that's the default.
(From OE-Core rev: cec30244d9c7051dbd727fddeb4b617c783c6fd5)
Signed-off-by: Ross Burton <ross.burton@arm.com>
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>
(cherry picked from commit 23cb39a5fa2a55681e7bc2605f435135cec9173b)
Signed-off-by: Anuj Mittal <anuj.mittal@intel.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: 5ce0102ca75754f0a7c777505678677cdf2a62bc)
Signed-off-by: Nathan Rossi <nathan@nathanrossi.com>
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>
(cherry picked from commit d39b686cc5fcc89405ffeb560ea64b5d305b0608)
Signed-off-by: Anuj Mittal <anuj.mittal@intel.com>
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>
Define CXX_FOR_TARGET and CC_FOR_TARGET for target go
since we are cross building it, it helps in defining the compiler
for cgo on target
backport a patch to fix CGO_LDFLAGS regression like below
Fixes
| Building std for target, linux/arm.
| go build runtime/cgo: invalid flag in go:cgo_ldflag: -Wl,-O1
(From OE-Core rev: 4b923dc9e77dfcfec7c9004483896d347bd88bdb)
Signed-off-by: Khem Raj <raj.khem@gmail.com>
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>
(cherry picked from commit a71d1b58e8cf0976a614c49aa2dfe1e0605dfe95)
Signed-off-by: Anuj Mittal <anuj.mittal@intel.com>
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>
Set PV directly to allow automated version updates between point releases.
(From OE-Core rev: 7a0f56c2c92c21f5da6635d5f3089e554a076655)
Signed-off-by: Alexander Kanavin <alex.kanavin@gmail.com>
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>
(cherry picked from commit 84034ea18a9fb1a012a1ecdc3a84296b08d3cc23)
Signed-off-by: Anuj Mittal <anuj.mittal@intel.com>
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>
On one of my buildservers I noticed perl do_install failing with the
following message:
| rm: cannot remove '<OEROOT>/tmp-rpb-glibc/work/armv8-2a-linaro-linux/perl/5.32.0-r0/image//usr/lib/perl5/5.32.0/*/CORE/libperl.so': No such file or directory
I tracked this down to shell being dash rather than bash not being able
to expand this glob in the middle of the filename. So replace the glob
expansion with the simpler one which works in all cases.
(From OE-Core rev: 784ca95f7e3a816be9c9782944fa49b896d58690)
Signed-off-by: Dmitry Baryshkov <dmitry.baryshkov@linaro.org>
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>
(cherry picked from commit d1ea1b5c12120abdd085dc4eb69120af9258a99b)
Signed-off-by: Anuj Mittal <anuj.mittal@intel.com>
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>
ffmpeg in qtwebengine/chromium fails to build on aarch64
ffmpeg/ffmpeg_internal/videodsp.o: in function `ff_prefetch_aarch64':
(.text+0x10): relocation truncated to fit: R_AARCH64_CONDBR19 against symbol `ff_prefetch_aarch64' defined in .text section in obj/third_party/ffmpeg/ffmpeg_internal/videodsp.o
Backport an upstream fix to handle this error which is a regrression in
binutils 2.35
(From OE-Core rev: 658024f47b5f96d3f4e1813b4716e8981fbf2e47)
Signed-off-by: Khem Raj <raj.khem@gmail.com>
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>
(cherry picked from commit 0a68def6b1f69b61096e58ae7778b61412dec4a2)
Signed-off-by: Anuj Mittal <anuj.mittal@intel.com>
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>
When comparing builds built with different host umasks, this file jumped out.
The umask from do_compile was influencing ${D} and as cp was used to add the
file it wasn't deterministic. Fix the file mode to ensure determinism.
(From OE-Core rev: b99796ec9436b63e4fc7cb7d12c0c9bcceef5d4b)
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>
(cherry picked from commit 37f37f4a52de3711973b372160f23672b61ff6ad)
Signed-off-by: Anuj Mittal <anuj.mittal@intel.com>
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>
PTH_FUNC definition needs to be modified in order to
intercept posix thread functions in both libc and libpthread.
In order to handle this in helgrind, weak alias the pthread functions in glibc.
Include a special case for musl.
See https://bugs.kde.org/show_bug.cgi?id=428909 for additional
discussion.
Upstream-Status: Submitted
(From OE-Core rev: 4c33ce1b1eca9aff0009bf71ce50f6398f7cd281)
Signed-off-by: Paul Floyd <paulf@free.fr>
Signed-off-by: Stacy Gaikovaia <Stacy.Gaikovaia@windriver.com>
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>
(cherry picked from commit 5da46a552d54de34a5243e1d90dcc6f52b7af746)
Signed-off-by: Anuj Mittal <anuj.mittal@intel.com>
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>
This issue describes expected behaviour, do not use tarfile with
untrusted data.
(From OE-Core rev: 391ed53928db0df325798a0bce18ec6947e09ddd)
Signed-off-by: Ross Burton <ross.burton@arm.com>
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>
(cherry picked from commit f4c22e83f2e68ff157da5ea1303acc2931d63f5f)
Signed-off-by: Anuj Mittal <anuj.mittal@intel.com>
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>
The diffstat tool is part of HOSTTOOLS. To support hosts that do not
have it installed with buildtools-tarball it must be enabled for
nativesdk.
(From OE-Core rev: 3a4ac9d028e6d7840660bb9640614d92fd89246f)
Signed-off-by: Nathan Rossi <nathan@nathanrossi.com>
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>
(cherry picked from commit 0ed002422bc46539f1d71ed19ee17358b6691bf0)
Signed-off-by: Anuj Mittal <anuj.mittal@intel.com>
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>
powerpc 32bit Linux Kernel widely uses .stabs pseudo-op to
produce debugging information in stabs format. Faced an issue
that during Linux Kernel build with Yocto build system for 32bit
powerpc platform resulting vmlinux contains absolute path in
.stabstr section that cannot be remapped with -fdebug-prefix-map
option.
Yocto uses scripts/mkmakefile Linux Kernel build approach that
allows to store all generated files outside of kernel source
tree. With this approach each compilier invocation is performed
with an absolute path to a file that will be compiled and this
absolute path is recorded in init stab. There is no way to remap
this path.
Reuse remap_debug_filename api to make -fdebug-prefix-map flag
aplicable for init stab.
(From OE-Core rev: b4b79870d7946e58692adb68d1329955500d3c56)
Signed-off-by: Denys Zagorui <dzagorui@cisco.com>
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>
(cherry picked from commit 4dce4e01cfa153fb12cfd1684d36e0432bef6741)
Signed-off-by: Anuj Mittal <anuj.mittal@intel.com>
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>
The license headers are clear that the code is "or later", fix LICENSE
to match.
(From OE-Core rev: 01fd8b51074a91053f632b2932238e35c926045c)
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>
(cherry picked from commit e565e0b908c71ad5106d1c6c73d269b819787e55)
Signed-off-by: Anuj Mittal <anuj.mittal@intel.com>
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>
After glibc uprev 2.31 -> 2.32, the following drd ptests
fail with the error "condition variable has not been initialized".
drd/tests/annotate_hb_err
drd/tests/annotate_sem
drd/tests/annotate_rwlock
drd/tests/annotate_order_2
drd/tests/annotate_smart_pointer
drd/tests/annotate_spinlock
drd/tests/monitor_example
drd/tests/pth_cond_race
drd/tests/pth_inconsistent_cond_wait
In glibc 2.32, the POSIX thread functions are in both
libc and libpthread, causing valgrind to misinterpert
test behaviour. This patch tells valgrind to intercept
both the libc and the libpthread functions, resolving
these ptest failures.
Here are the latest test results on qemux86-64:
=== Test Summary ===
TOTAL: 728
PASSED: 688
FAILED: 1
SKIPPED: 39
The failed test is in helgrind:
FAIL: helgrind/tests/pth_destroy_cond
Patch can be removed once we uprev valgrind.
See valgrind commit 15330adf7c2471fbaa6a0818db07078d81dbff97.
(From OE-Core rev: e8f265e957e187b4f84f566aaeec8f514e3044d6)
Signed-off-by: Stacy Gaikovaia <Stacy.Gaikovaia@windriver.com>
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>
It failed as below when rerun do_install for tcl:
$ bitbake tcl
$ bitbake tcl -f -cinstall
[snip]
| NOTE: make -j 72 DESTDIR=/build/tmp/work/core2-64-wrs-linux/tcl/8.6.10-r0/image install
| abort()ing pseudo client by server request. See https://wiki.yoctoproject.org/wiki/Pseudo_Abort for more details on this.
| /build/tmp/work/core2-64-wrs-linux/tcl/8.6.10-r0/temp/run.do_install.3490694: line 184: 3499214 Aborted (core dumped) make -j 72 "$@"
By default the S is ${WORKDIR}/${BPN}-${PV}, but after unpack,
the tcl source [1] unpacked to ${WORKDIR}/${BPN}${PV} and all the
files under ${WORKDIR}/${BPN}${PV} are acutally the source files.
But the the main Makefile.in is under ${WORKDIR}/${BPN}${PV}/unix
for tcl, so there is below logic in tcl recipe:
S = "${WORKDIR}/${BPN}${PV}/unix"
To adapt the potential pseudo changes, there is a general logic to
exclude ${S} from pseudo database in base.bbclass in [2]. That's to
say, just the dir ${WORKDIR}/${BPN}${PV}/unix is excluded from the
pseudo database.
But it's not enough for tcl, we need to exclude the actual source dir
${WORKDIR}/${BPN}${PV} from pseudo database specifically to fix the
above do_install failure.
[1] https://downloads.sourceforge.net/tcl/tcl8.6.0-src.tar.gz
[2] https://git.openembedded.org/openembedded-core/tree/meta/classes/base.bbclass#n396
(From OE-Core rev: 0d835f8069a336fd6bab1058d4e605a81c844212)
Signed-off-by: Mingli Yu <mingli.yu@windriver.com>
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>
Replace OE private qemu patch with one that got upstreamed
and solves the same problem: increase qemumips CI performance
by increasing number of TLBs in CPU model and reduce need to
run software TLB refill code.
(From OE-Core rev: a99dace7463d310688f4098a51316dc0743651e2)
Signed-off-by: Victor Kamensky <kamensky@cisco.com>
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>
This reverts commit 4470a04943352224955f17e004962f0f9e1c9b0c.
Will be replaced with upstreamed solution instead, that just
bumped number of TLBs to 64 in original 34Kf CPU model.
(From OE-Core rev: 894f1d58d93073d290f35d1090b03717bc7b4dc0)
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>
