Ross Burton [Tue, 23 Jul 2024 14:51:01 +0000 (15:51 +0100)]
python3: add dependency on -compression to -core
importlib.metadata is part of -core, but that will import zipfile which
is part of -compression.
Obviously this shows that our packaging of the Python modules is not
optimal. I plan to follow up with a redesign of the splitting which
focuses on simply pulling out the larger or esoteric modules and
having a more featureful core.
Signed-off-by: Ross Burton <ross.burton@arm.com> Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>
Richard Purdie [Tue, 23 Jul 2024 14:22:40 +0000 (15:22 +0100)]
pseudo: Fix to work with glibc 2.40
glibc 2.40 renames some internal header variables. Update our hack to
work with the new version. These kinds of problems illustrate we need to
address the issue properly.
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>
rootfs: Ensure run-postinsts is not uninstalled for read-only-rootfs-delayed-postinsts
This patch ensures that pkg_postinst_ontarget task is executed for read only rootfs when
read-only-rootfs-delayed-postinsts is set as IMAGE_FEATURES. The issue was that run-postinsts
could be uninstalled at the end of rootfs construction and that shouldn't happen for
the delayed usecase.
In addition to the fix, a test in meta/lib/oeqa/selftest/cases/overlayfs.py testing
the fix has been implemented.
Niko Mauno [Thu, 4 Jul 2024 07:15:18 +0000 (07:15 +0000)]
systemd: Mitigate /var/tmp type mismatch issue
The base-files recipe provides /var/tmp -> /var/volatile/tmp symlink
which is in conflict with systemd upstream tmpfiles.d/tmp.conf which
defines it as a directory (or subvolume on btrfs).
This generates following error in journal:
Jul 03 15:37:21 qemux86-64 systemd-tmpfiles[158]: "/var/tmp" already exists and is not a directory.
Mitigate the issue by defining /var/tmp as symlink corresponding to
the one created by base-files.
Signed-off-by: Niko Mauno <niko.mauno@vaisala.com> Signed-off-by: Alexandre Belloni <alexandre.belloni@bootlin.com>
Jose Quaresma [Wed, 17 Jul 2024 12:34:22 +0000 (13:34 +0100)]
oeqa/runtime/scp: requires openssh-sftp-server
The SCP protocol was deprecated in favor of the SFTP.
For the legacy SCP protocol scp should be run with "-O".
Instead of adding "-O" on the scp_options ssh oeqa we can
require the openssh-sftp-server to be instaled on the target.
This way the test will work more deterministic regardless of
the host machine client used.
For the old fashion clients still using legacy SCP protocol
the openssh-sshd server will be used, for the new ones using
the SFTP the openssh-sftp-server will be picked.
Signed-off-by: Jose Quaresma <jose.quaresma@foundries.io> Signed-off-by: Alexandre Belloni <alexandre.belloni@bootlin.com>
Richard Purdie [Mon, 22 Jul 2024 14:27:50 +0000 (15:27 +0100)]
time64.inc: Add warnings exclusion for known toolchain problems for now
Fixing our infrastructure to show warnings for nativesdk packages resulted
in some issues with 32 bit toolchains and 32 bit time becoming clear.
For now, add INSANE_SKIP entries for these which effectively documents
the issue and allows us to at least start running the tests in the
infrastructure.
Over time we will need to try and fix these if upstream doesn't.
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>
systemd started to warn about used but unset environment variables.
Let us set watchdog_module=none which is used by the watchdog.service to get
rid of the following warning:
watchdog.service: Referenced but unset environment variable evaluates to an empty string: watchdog_module
Signed-off-by: Wadim Egorov <w.egorov@phytec.de> Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>
Martin Jansa [Mon, 22 Jul 2024 08:15:11 +0000 (10:15 +0200)]
kernel.bbclass: add original package name to RPROVIDES for -image and -base
* -image and -base change PKG to:
PKG:${KERNEL_PACKAGE_NAME}-image = "${KERNEL_PACKAGE_NAME}-image-${@legitimize_package_name(d.getVar(KERNEL_VERSION))}"
PKG:${KERNEL_PACKAGE_NAME}-base = "${KERNEL_PACKAGE_NAME}-${@legitimize_package_name(d.getVar(KERNEL_VERSION))}"
* but only when debian.bbclass is inheritted they add the original package name into RPROVIDES by:
https://git.openembedded.org/openembedded-core/commit/?id=3409c4379559afbb1d1d29045582995147a33bbc
* fixes the build if some packagegroup or something RDEPENDS on kernel-image or kernel-base
and the DISTRO doesn't inherit debian.bbclass
* as shown in pkgdata:
linux-raspberrypi $ egrep "^(PKG:)|(RPRO)" 6.6.36+git-*/pkgdata/runtime/kernel-image
6.6.36+git-debian/pkgdata/runtime/kernel-image:RPROVIDES:kernel-image: kernel-image (=6.6.36+git@PRSERV_PV_AUTOINC@+733366844f_769634f344)
6.6.36+git-debian/pkgdata/runtime/kernel-image:PKG:kernel-image: kernel-image-6.6.36-v8
6.6.36+git-without-debian/pkgdata/runtime/kernel-image:PKG:kernel-image: kernel-image-6.6.36-v8
Below commits on binutils-2.42 stable branch are updated. 29ae8b8ea71 x86-64: Skip -z mark-plt tests on MUSL 92cc764e58f hppa: Fix handling of relocations that apply to data c439c1e1f56 elf: Add glibc version dependency only if needed 68ae8e2a849 ld: pass -g for ld-elf tests a1e3cb45c67 aarch64: Enable +cssc for armv8.9-a
Signed-off-by: Deepthi Hemraj <Deepthi.Hemraj@windriver.com> Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>
gobject-introspection: split tools and giscanner into a separate package
The tools and giscanner require python3-setuptools at runtime. This increases
the image size quite a bit (on my end ~1.7MB compressed). Most users don't need
these tools in the final image so make them optional.
This commit introduces a new packcage gobject-introspection-tools
Signed-off-by: Guðni Már Gilbert <gudni.m.g@gmail.com> Signed-off-by: Alexandre Belloni <alexandre.belloni@bootlin.com> Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>
* The <stdbit.h> header type-generic macros have been changed when using
GCC 14.1 or later to use __builtin_stdc_bit_ceil etc. built-in functions
in order to support unsigned __int128 and/or unsigned _BitInt(N) operands
with arbitrary precisions when supported by the target.
* The GNU C Library now supports a feature test macro _ISOC23_SOURCE to
enable features from the ISO C23 standard. Only some features from
this standard are supported by the GNU C Library. The older name
_ISOC2X_SOURCE is still supported. Features from C23 are also enabled
by _GNU_SOURCE, or by compiling with the GCC options -std=c23,
-std=gnu23, -std=c2x or -std=gnu2x.
* The following ISO C23 function families (introduced in TS
18661-4:2015) are now supported in <math.h>. Each family includes
functions for float, double, long double, _FloatN and _FloatNx, and a
type-generic macro in <tgmath.h>.
- Exponential functions: exp2m1, exp10m1.
- Logarithmic functions: log2p1, log10p1, logp1.
* A new tunable, glibc.rtld.enable_secure, can be used to run a program
as if it were a setuid process. This is currently a testing tool to allow
more extensive verification tests for AT_SECURE programs and not meant to
be a security feature.
* On Linux, the epoll header was updated to include epoll ioctl definitions
and the related structure added in Linux kernel 6.9.
* The fortify functionality has been significantly enhanced for building
programs with clang against the GNU C Library.
* Many functions have been added to the vector library for aarch64:
acosh, asinh, atanh, cbrt, cosh, erf, erfc, hypot, pow, sinh, tanh
* On x86, memset can now use non-temporal stores to improve the performance
of large writes. This behaviour is controlled by a new tunable
x86_memset_non_temporal_threshold.
Deprecated and removed features, and other changes affecting compatibility:
* Architectures which use a 32-bit seconds-since-epoch field in struct
lastlog, struct utmp, struct utmpx (such as i386, powerpc64le, rv32,
rv64, x86-64) switched from a signed to an unsigned type for that
field. This allows these fields to store timestamps beyond the year
2038, until the year 2106. Please note that applications are still
expected to migrate off the interfaces declared in <utmp.h> and
<utmpx.h> (except for login_tty) due to locking and session management
problems.
* __rseq_size now denotes the size of the active rseq area (20 bytes
initially), not the size of struct rseq (32 bytes initially).
Signed-off-by: Khem Raj <raj.khem@gmail.com> Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>
Mark Hatle [Fri, 19 Jul 2024 18:58:18 +0000 (13:58 -0500)]
selftest-hardlink: Add additional test cases
Additional test cases for debug symlink generation both binaries
and static libraries.
This also has the side effect of testing for race conditions in the
hardlink debug generation and stripping.
Signed-off-by: Mark Hatle <mark.hatle@amd.com> Signed-off-by: Mark Hatle <mark.hatle@kernel.crashing.org> Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>
Mark Hatle [Fri, 19 Jul 2024 18:58:17 +0000 (13:58 -0500)]
package.py: Fix static library processing
When PACKAGE_STRIP_STATIC is enabled the system did not pay attention to
hardlinks. This could trigger a race condition during stripping of static
libraries where multiple strips (through hardlinks) could run at the same
time triggering a truncated or modified file error.
The hardlink breaking code is based on the existing code for elf files, but
due to the nature of the symlinks needed to be done in a separate block of
code.
Add support for static-library debugfs hardlinking through the existing
inode processing code.
Print a note to the logs if the link target can't be found. This isn't
strictly an error, but may be useful for debugging an issue where a file
isn't present.
Signed-off-by: Mark Hatle <mark.hatle@amd.com> Signed-off-by: Mark Hatle <mark.hatle@kernel.crashing.org> Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>
recipeutils/get_recipe_upgrade_status: group recipes when they need to be upgraded together
This will allow 'lockstep upgrades' of such recipes, improving success
rates in automated version updating process.
devtool check-upgrade-status now prints:
These recipes need to be upgraded together {
glib-2.0 2.80.2 2.80.4 Anuj Mittal <anuj.mittal@intel.com>
glib-2.0-initial 2.80.2 2.80.4 Anuj Mittal <anuj.mittal@intel.com>
}
These recipes need to be upgraded together {
util-linux 2.39.3 2.40.2 Chen Qi <Qi.Chen@windriver.com>
util-linux-libuuid 2.39.3 2.40.2 Chen Qi <Qi.Chen@windriver.com>
}
These recipes need to be upgraded together {
cmake 3.29.3 3.30.0 Unassigned <unassigned@yoctoproject.org>
cmake-native 3.29.3 3.30.0 Unassigned <unassigned@yoctoproject.org>
}
etc.
Signed-off-by: Alexander Kanavin <alex@linutronix.de> Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>
lib/recipeutils: add a function to determine recipes with shared include files
This functionality is needed for 'lockstep version upgrades' where several
recipes need to be upgraded at the same time to produce a buildable
outcome.
The function itself obtains BBINCLUDED for each recipe and then massages
the data until it takes the form of a list of sets:
There's also a selftest that checks for the above.
Unfortunately this won't detect mutually exclusive recipes like mesa and mesa-gl
as they're chosen with PREFERRED_PROVIDER and can't be enabled in the same build
at the same time. ('devtool upgrade' will also accept just one of them but not the other)
Signed-off-by: Alexander Kanavin <alex@linutronix.de> Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>
lib/oe/recipeutils: return a dict in get_recipe_upgrade_status() instead of a tuple
Putting various things in a tuple is an anti-pattern of sorts, as the consumers
have to unpack it into local variables for readability, or access items directly
with indexes, which makes code pretty much unreadable.
Signed-off-by: Alexander Kanavin <alex@linutronix.de> Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>
Mark Hatle [Fri, 19 Jul 2024 18:58:16 +0000 (13:58 -0500)]
package.py: Fix static debuginfo split
Fix:
NameError: name 'shutil' is not defined
Signed-off-by: Mark Hatle <mark.hatle@amd.com> Signed-off-by: Mark Hatle <mark.hatle@kernel.crashing.org> Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>
Changelog:
** libgnutls: PBMAC1 is now supported as a MAC mechanism for PKCS#12
To be compliant with FIPS 140-3, PKCS#12 files with MAC based on
PBKDF2 (PBMAC1) is now supported, according to the specification
proposed in draft-ietf-lamps-pkcs12-pbmac1.
** libgnutls: SHA3 extendable output functions (XOF) are now supported
SHA3 XOF, SHAKE128 and SHAKE256, are now usable through a new
public API gnutls_hash_squeeze.
** API and ABI modifications:
gnutls_pkcs12_generate_mac3: New function
gnutls_pkcs12_flags_t: New enum
gnutls_hash_squeeze: New function
Signed-off-by: Simone Weiß <simone.p.weiss@posteo.com> Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>
Improve 'unknown status' error message by including the CVE id and by
quoting status field. Otherwise it might be troublesome to understand
messages like 'do_create_spdx: Unknown CVE status ' (the status was
empty).
Signed-off-by: Dmitry Baryshkov <dmitry.baryshkov@linaro.org> Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>
Below commits on glibc-2.39 stable branch are updated. e8f5217097 Linux: Make __rseq_size useful for feature detection (bug 31965) 143a7a0623 elf: Make dl-rseq-symbols Linux only 1062ebbd19 nptl: fix potential merge of __rseq_* relro symbols 5c46e6b666 s390x: Fix segfault in wcsncmp [BZ #31934] a036311246 stdlib: fix arc4random fallback to /dev/urandom (BZ 31612) d2cbfcf1d9 math: Provide missing math symbols on libc.a (BZ 31781) d473c9bb3b math: Fix isnanf128 static build (BZ 31774) 3950cbd7a1 math: Fix i386 and m68k exp10 on static build (BZ 31775) 6cb25aff85 math: Fix i386 and m68k fmod/fmodf on static build (BZ 31488) 74630b1bb7 posix: Fix pidfd_spawn/pidfd_spawnp leak if execve fails (BZ 31695) 7f9f25f255 Linux: Include <dl-symbol-redir-ifunc.h> in dl-sysdep.c 198632a05f NEWS: update list of fixed CVEs in 2.39 77bb3c7154 NEWS: update list of fixed bugs in 2.39 f05638731e x86: Properly set x86 minimum ISA level [BZ #31883] b7f5b0a711 x86: Properly set MINIMUM_X86_ISA_LEVEL for i386 [BZ #31867] e1d0040a6d localedata: ssy_ER: Fix syntax error 9de9cd17e7 malloc: New test to check malloc alternate path using memory obstruction a2da98aa2a malloc: Improve aligned_alloc and calloc test coverage. 305ee48826 malloc/Makefile: Split and sort tests 00899eba26 x86/cet: fix shadow stack test scripts 6ade91c211 elf: Avoid some free (NULL) calls in _dl_update_slotinfo 70f560fc22 misc: Add support for Linux uio.h RWF_NOAPPEND flag c9d8534406 i386: Disable Intel Xeon Phi tests for GCC 15 and above (BZ 31782) aee37de299 Reinstate generic features-time64.h dd535f4f19 Always define __USE_TIME_BITS64 when 64 bit time_t is used 26e7005728 socket: Use may_alias on sockaddr structs (bug 19622) 9f2b100d67 parse_fdinfo: Don't advance pointer twice [BZ #31798]
Signed-off-by: Deepthi Hemraj <Deepthi.Hemraj@windriver.com> Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>
Ross Burton [Fri, 19 Jul 2024 13:25:20 +0000 (14:25 +0100)]
ruby: upgrade 3.3.0 -> 3.3.4
3.3.4:
This release fixes a regression in Ruby 3.3.3 that dependencies are missing in
the gemspec of some bundled gems: net-pop, net-ftp, net-imap, and prime [Bug
like Heroku. If your bundle install runs correctly now, you may not have this
issue.
3.3.3:
This release includes:
RubyGems 3.5.11
Bundler 2.5.11
REXML 3.2.8
strscan 3.0.9
--dump=prism_parsetree is replaced by --parser=prism --dump=parsetree
Invalid encoding symbols raise SyntaxError instead of EncodingError
Memory leak fix in Ripper parsing
Bugfixes for YJIT, **{}, Ripper.tokenize, RubyVM::InstructionSequence#to_binary,
--with-gmp, and some build environments
3.3.2:
This release includes many bug-fixes. See the GitHub releases for further details.
3.3.1:
This release includes security fixes. Please check the topics below for details.
CVE-2024-27282: Arbitrary memory address read vulnerability with Regex search
CVE-2024-27281: RCE vulnerability with .rdoc_options in RDoc
CVE-2024-27280: Buffer overread vulnerability in StringIO
Signed-off-by: Ross Burton <ross.burton@arm.com> Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>
oeqa/selftest/reproducibile: rename of reproducible directories
We do see a few reproducible issues are depending on the path length of the build directory.
https://bugzilla.yoctoproject.org/show_bug.cgi?id=15554
The current implementation of reproducible tests having different names for directories but with same length.
The build directory names are changed to have different length.
Signed-off-by: Sundeep KOKKONDA <sundeep.kokkonda@windriver.com> Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>
Ross Burton [Thu, 18 Jul 2024 14:48:28 +0000 (15:48 +0100)]
openssl: strip the test suite
The test suite is huge because every test binary is statically linked to
libssl and/or libcrypto. This bloats the size of the -dbg package hugely,
so strip the test suite before packaging.
This reduces the size of openssl-dbg by 90% from ~1.2GB to ~120MB, and
reduces the size of the build tree from ~1.9GB to ~800MB.
Signed-off-by: Ross Burton <ross.burton@arm.com> Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>
Richard Purdie [Wed, 17 Jul 2024 21:27:00 +0000 (22:27 +0100)]
qemu: Drop mips workaround
In local testing this appears to not be needed any more. Since the bash dependency
is painful in many different ways, particularly being injected at do_package time,
drop it all.
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>
To fulfill OverlayFS workdir requirements, the mount-copybind script
creates a workdir. But if the mount operation fails for any reason,
the workdir is left there.
Then, subsequent runs of mount-copybind will again try to
create the directory and pollute system logs with failed mkdir error
messages.
This commit mitigates the problem by unconditionally removing workdir
if the OverlayFS is not used or fails to run.
Signed-off-by: Ricardo Simoes <ricardo.simoes@pt.bosch.com> Signed-off-by: Mark Jonas <mark.jonas@de.bosch.com> Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>
volatile-binds: Do not create workdir if OverlayFS is disabled
When the mountpoint parameter is a directory, the mount-copybind will
first try to use OverlayFS. Because of that, it needs to create the
OverlayFS workdir (determined by the overlay_workdir).
But if the environment variable MOUNT_COPYBIND_AVOID_OVERLAYFS is set
to "1", the script uses bind mount. In that case, the overlay_workdir
is useless, leaving the spec parent directory in a dirty state.
This commit changes mount-copybind so that the overlay_workdir is only
created when MOUNT_COPYBIND_AVOID_OVERLAYFS is not set to 1.
Signed-off-by: Ricardo Simoes <ricardo.simoes@pt.bosch.com> Signed-off-by: Mark Jonas <mark.jonas@de.bosch.com> Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>
The systemd-mime package is added as a recommendation for systemd so
that it is installed by default, but can be excluded, e.g., via
BAD_RECOMMENDATIONS. Excluding it avoids pulling in the MIME database
for products that have no use for it.
Signed-off-by: Peter Kjellerstedt <peter.kjellerstedt@axis.com> Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>
Richard Purdie [Wed, 17 Jul 2024 13:29:20 +0000 (14:29 +0100)]
oeqa/postactions: Separate artefact collection from test result collection
Combining the test result collection and artefact collection hasn't worked out
well as the data has different life cycles, the artefacts can be large and
we need to be able to clean them up on a different timescale.
Separate them out to be controlled by a separate variable, OEQA_ARTEFACT_DIR.
Also rework the code to inject a directory with a date/time and random component
to allow builds to run in parallel. Pass function arguments to avoid re-reading
variables.
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>
MCNAME isn't defined outside our mcextend ptest images so use a wildcard in
TESTIMAGE_FAILED_QA_ARTIFACTS instead. This unbreaks the value in other images.
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>
Jose Quaresma [Tue, 16 Jul 2024 09:29:54 +0000 (10:29 +0100)]
openssh: systemd sd-notify patch was rejected upstream
Still side effects of the XZ backdoor.
Racional [1]:
License incompatibility and library bloatedness were the reasons.
Given recent events we're never going to take a dependency on libsystemd,
though we might implement the notification protocol ourselves if it isn't too much work.
File /usr/bin/afuc-asm in package mesa-tools contains reference to TMPDIR [buildpaths]
File /usr/bin/afuc-disasm in package mesa-tools contains reference to TMPDIR [buildpaths]
Signed-off-by: Dmitry Baryshkov <dmitry.baryshkov@linaro.org> Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>
Noticed in the installation logs a few paths have
double slashes '//'. Doesn't seem to do any harm, though
it is good to clean this up for consistency.
Signed-off-by: Guðni Már Gilbert <gudni.m.g@gmail.com> Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>
Ross Burton [Tue, 16 Jul 2024 16:11:11 +0000 (17:11 +0100)]
openssl: disable tests unless ptest is enabled
The upstream Makefile always builds the tests unless they're explicitly
disabled. Whilst this doesn't make a difference to the final package and
sysroot output, disabling the tests for openssl-native reduces the size
of the build tree from 659M to 78M and reduces the CPU time used by 30%.
Signed-off-by: Ross Burton <ross.burton@arm.com> Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>
Kai Kang [Wed, 17 Jul 2024 02:27:43 +0000 (10:27 +0800)]
systemd: fix VERSION_TAG related build error
The patch 0001-src-boot-efi-meson.build-ensure-VERSION_TAG-exists-i.patch for
systemd-boot is also required by systemd to resolve VERSION_TAG related build
error.
Signed-off-by: Kai Kang <kai.kang@windriver.com> Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>
Marta Rybczynska [Mon, 15 Jul 2024 10:20:02 +0000 (12:20 +0200)]
classes/kernel.bbclass: update CVE_PRODUCT
Add linux:linux to CVE_PRODUCT. linux:linux is used by the kernel CNA
in raw CVE entries. We can't use just linux, because of conflicts with
CPE entries of multiple distributions.
Signed-off-by: Marta Rybczynska <marta.rybczynska@syslinbit.com> Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>
Ross Burton [Fri, 12 Jul 2024 18:43:58 +0000 (19:43 +0100)]
oeqa/sdk: add out-of-tree kernel module building test
Validate that out-of-tree kernel module building using kernel-devsrc
works as expected.
This test uses cryptodev-linux as a idiomatic out of tree module. As the
latest release doesn't actually build with kernel 6.7+, use the same
commit that our recipe uses.
Signed-off-by: Ross Burton <ross.burton@arm.com> Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>
Mark Hatle [Mon, 15 Jul 2024 19:56:06 +0000 (14:56 -0500)]
create-sdpx-2.2.bbclass: Switch from exists to isfile checking debugsrc
While debugsrc is almost always a file (or link), there are apparently
cases where a directory could be returned from the dwarfsrcfiles
processing. When this happens, the hashing fails and an error results
when building the SPDX documents.
Signed-off-by: Mark Hatle <mark.hatle@amd.com> Signed-off-by: Mark Hatle <mark.hatle@kernel.crashing.org> Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>
Changqing Li [Tue, 16 Jul 2024 09:46:33 +0000 (17:46 +0800)]
pixman: update patch for fixing inline failure with -Og
Previous patch works on fedora40 with gcc-14, but not works
on ubuntu2004 with gcc-9. Update the patch to fix the do_compile
failure:
In function ‘combine_inner’,
inlined from ‘combine_soft_light_ca_float’ at ../pixman-0.42.2/pixman/pixman-combine-float.c:655:1:
../pixman-0.42.2/pixman/pixman-combine-float.c:370:5: error: inlining failed in call to ‘always_inline’ ‘combine_soft_light_c’: function not
considered for inlining
370 | combine_ ## name ## _c (float sa, float s, float da, float d)
Signed-off-by: Changqing Li <changqing.li@windriver.com> Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>
Jose Quaresma [Tue, 16 Jul 2024 14:16:37 +0000 (15:16 +0100)]
openssh: drop rejected patch fixed in 8.6p1 release
The rationale [1] is that C11 6.5.6.9 says:
"""
When two pointers are subtracted, both shall point to elements of the
same array object, or one past the last element of the array object; the
result is the difference of the subscripts of the two array elements.
"""
In these cases the objects are arrays of char so the result is defined,
and we believe that the compiler incorrectly trapping on defined behaviour.
I also found https://gcc.gnu.org/bugzilla/show_bug.cgi?id=63303
("Pointer subtraction is broken when using -fsanitize=undefined") which seems to support this position.
Richard Purdie [Wed, 10 Jul 2024 23:00:28 +0000 (00:00 +0100)]
distro/include: Add yocto-space-optimize, disabling debugging for large components
Add an include file to allow turning off the debug compiler options
for a small set of recipes to reduce build on disk footprint and
package/sstate sizes.
This is currently applied to llvm and qemu target recipes.
The llvm-staticdev package takes up around 1.3GB alone. These three
changes lead to a reduction in TMPDIR size for a world build from
240GB to 199GB, also removing some very large sstate objects.
There is more that could and should be done but this does illustrate
one way to speed up and reduce build size in a focused way whilst we
ideally look into other approaches.
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>
classes-recipe/multilib_script: Expand before splitting
multilib_script.bbclass was unable to work correctly in the case where
e.g. a PACKAGECONFIG removed the script that it was intended to rename
(as an example, the "trace" PACKAGECONFIG in cairo). The way to fix this
would be to do something like:
but this is not possible because the variable is not expanded before
being split.
To fix this, change the class to expand the variable before splitting.
There are two cases to be considered that could possibly break:
1) If the RHS of the ":" contains a ":", which is accounted for by
limiting the splitting to 1 split, which will leave the ":" in the
RHS in tact. Of note, this works because ":" isn't valid in a
package name
2) If the RHS of the ":" contained whitespace, however this would have
broken the mv command written to multilibscript_rename(), so this
isn't occurring in practice.
projects / thirdparty / openembedded / openembedded-core-contrib.git / log
