Richard Purdie [Fri, 5 Jun 2026 09:11:53 +0000 (10:11 +0100)]
base: Allow zstd and xz decompression from the host tools
We have xz and zstd in HOSTTOOLS but not ASSUME_PROVIDED. In most cases where
we have a dependency on xz-native or zstd-native, we need to compress files
or we need the libraries those recipes provide. In some cases we just need
decompression though.
Compression output is version dependent so we need those dependencies. Libraries
need headers so those are needed too. THe main place we could benefit are the
unpack dependencies from base.bbclass.
Therefore add "XXX-decompress" PROVIDES to those recipes, then change the
dependency in base.bbclass to indicate decompression only. We can then
ASSUME_PROVIDED the decompression side of things to optimise builds whilst
still keeping the other dependencies functional.
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>
Richard Purdie [Fri, 5 Jun 2026 09:12:46 +0000 (10:12 +0100)]
sysstat: Add missing xz-native DEPENDS
sysstat uses xz to compress man pages. The version from the host may vary
and hence make the output non-reproducible so we need xz-native in
DEPENDS to make the output deterministic.
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>
- QEMU 11.0.0 has officially dropped all support for 32-bit host systems.
This includes the ability to build the system emulator (softmmu) on
any 32-bit host architecture. Add COMPATIBLE_HOST for class-target and
class-nativesdk to restrict builds to 64-bit architectures. Fix SDK
build failure when SDKMACHINE is 32-bit.
Note: QEMU 11.0.0 still supports emulating 32-bit guest architectures
(such as qemu-system-i386 or qemu-system-arm) as long as they are
running on a 64-bit host.
- qemu-targets.inc: skip linux-user targets when HOST_ARCH is not in the
set of architectures supported by QEMU 11.0.0's linux-user mode
(64-bit only: aarch64, loongarch64, mips64, ppc64, riscv64, s390x,
sparc64, x86_64).
- Add --cpu=${TUNE_ARCH} to EXTRA_OECONF for class-target, and add
--cpu=${SDK_ARCH} to nativesdk configure options to fix build
failure. QEMU 11's configure script falls back to 'uname -m' (x86_64)
when it cannot detect the host CPU.
- Add 0012-meson-fix-close_range-detection-on-older-glibc.patch:
the meson has_function('close_range') check succeeds at link time on
hosts with kernel >= 5.9 even when glibc < 2.34 does not declare the
function, causing implicit declaration errors. Add a prefix include
so the check only succeeds when the header declares close_range().
- Add 0013-hw-scsi-vhost-scsi-include-standard-headers-for-vhost-worker-structs.patch:
on build hosts with kernel headers < 6.0, vhost-scsi.c fails to
compile due to missing struct vhost_vring_worker/vhost_worker_state.
Include QEMU's bundled standard-headers before the system header,
matching the pattern already used in hw/virtio/vhost-backend.c.
- Drop fix-strerrorname_np.patch (fixed upstream).
Drop 0001-linux-user-elfload.c-Correction-to-HWCAP2-accessor.patch
(fixed upstream).
Drop 0001-accel-tcg-Fix-iotlb_to_section-for-different-Address.patch
(fixed upstream).
AI-Generated: kiro-cli
Based on a patch from Quan Sun <Quan.Sun@windriver.com>
Signed-off-by: Quan Sun <Quan.Sun@windriver.com> Signed-off-by: Alexander Kanavin <alex@linutronix.de> Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>
Also, adjust qemu recipe to be using it, instead of its
own internal copy. With qemu 11.x the internal copy is removed
and so a separate qemu-qmp recipes becomes a requirement.
Split out from a patch from Quan Sun <Quan.Sun@windriver.com>
Signed-off-by: Alexander Kanavin <alex@linutronix.de> Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>
Bin Cao [Thu, 4 Jun 2026 10:10:28 +0000 (18:10 +0800)]
insane.bbclass: check for build host HOME directory in packaged files
Extend package_qa_check_buildpaths to also detect the build host's
HOME directory path in target packages. This catches cases where
\$HOME-derived paths (such as userbase in Python's sysconfig) leak
into the rootfs.
Signed-off-by: Bin Cao <bin.cao.cn@windriver.com> Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>
Bin Cao [Thu, 4 Jun 2026 10:10:27 +0000 (18:10 +0800)]
python3: sanitize userbase in _sysconfig_vars JSON to avoid host path leak
The _sysconfig_vars__linux_x86_64-linux-gnu.json file contains a
"userbase" field that is populated from the build host user's $HOME
at build time. This leaks the build host user's home directory path
into the target rootfs.
The existing py_package_preprocess() cleanup for this JSON file only
strips known OE build path prefixes (STAGING_DIR_TARGET, RECIPE_SYSROOT,
etc.), but the userbase value comes from the build user's $HOME
environment variable which doesn't match any of those patterns.
Set userbase to an empty string in the packaged JSON. This is safe
because at runtime, sysconfig.get_config_vars() always recomputes
userbase by calling _getuserbase(), which resolves the actual target
user's ~/.local path dynamically. The static value in the JSON is
never used for runtime path resolution.
Signed-off-by: Bin Cao <bin.cao.cn@windriver.com> Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>
Haixiao Yan [Mon, 1 Jun 2026 06:46:41 +0000 (14:46 +0800)]
subversion: fix svn-revision.txt conflict when building for qemux86-64
The revision-install target in Makefile.in generates svn-revision.txt
by first trying to run the cross-compiled svnversion binary on the host.
When the target architecture is x86-64 (same as the build host), the
cross-compiled binary can execute on the host. If the host has libapr1
installed, the 64-bit svnversion succeeds and outputs "Unversioned
directory", while the 32-bit (lib32) binary lacks the 32-bit libapr1
and fails, falling back to "unknown". The differing content causes an
RPM file conflict on /usr/include/subversion-1/svn-revision.txt when
lib32-subversion-dev and subversion-dev are installed together in
do_rootfs.
This does not affect non-x86-64 targets (e.g. ARM) because the
cross-compiled binary cannot execute on the x86-64 host, so both
64-bit and 32-bit builds fall through to the same fallback.
Since the source is a release tarball with no SVN metadata, the correct
value is always "unknown". Fix by patching Makefile.in to skip the
svnversion invocations entirely.
Signed-off-by: Haixiao Yan <haixiao.yan.cn@windriver.com> Signed-off-by: Mathieu Dubois-Briand <mathieu.dubois-briand@bootlin.com> Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>
populate_sdk_ext: keep SDK_TARGETS so SPDX/SBOM tasks remain in locked sigs
Fixes [YOCTO #15853]
The filtering in populate_sdk_ext.bbclass was removing all image targets from
the locked signatures, including the main SDK images (SDK_TARGETS). As a
result, their tasks including the image SPDX/SBOM tasks were not added to
locked-sigs.inc.
Without these entries, the eSDK lacked the necessary sstate, and the installer
attempted to run the missing tasks during installation, leading to unexpected
task execution errors.
This patch keeps SDK_TARGETS (and their multilib variants) in the locked
signatures so their SPDX/SBOM tasks are included. With those tasks in the
sstate cache, the eSDK installs cleanly without needing to re-run anything.
Signed-off-by: Jayasurya Maganuru <Maganuru.Jayasurya@windriver.com> Signed-off-by: Sunil Dora <sunilkumar.dora@windriver.com> Signed-off-by: Mathieu Dubois-Briand <mathieu.dubois-briand@bootlin.com> Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>
create-spdx-image-3.0: Image SPDX/SBOM tasks are retained for eSDK installation
Fixes [YOCTO #15853]
Image SPDX/SBOM tasks were running after do_sdk_depends, causing their
signatures to be excluded from locked-sigs.inc. As a result, the eSDK installer
attempted to re-run these tasks, leading to unexpected task execution errors.
Run do_create_image_sbom_spdx before do_sdk_depends to ensure all image
SPDX/SBOM tasks are completed and captured in the locked signatures.
Signed-off-by: Jayasurya Maganuru <Maganuru.Jayasurya@windriver.com> Suggested-by: Joshua Watt <jpewhacker@gmail.com> Signed-off-by: Sunil Dora <sunilkumar.dora@windriver.com> Signed-off-by: Mathieu Dubois-Briand <mathieu.dubois-briand@bootlin.com> Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>
Adam Blank [Sat, 18 Apr 2026 18:34:07 +0000 (20:34 +0200)]
bitbake.conf: remove 'extend_recipe_sysroot' from BB_HASHEXCLUDE_COMMON
'extend_recipe_sysroot' can and should be handled just like any other
function. Fine grained use of 'vardepsexclude' is capable of achieving
the same effect as ingoring 'extend_recipe_sysroot' on the base hash
level, but with a better visibility, and is a more uniform approach.
Signed-off-by: Adam Blank <adam.blank.g@gmail.com> Signed-off-by: Mathieu Dubois-Briand <mathieu.dubois-briand@bootlin.com> Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>
classes: image-buildinfo: allow deploying the buildinfo file
The `buildinfo` file generated by the `image-buildinfo` class
contains useful information to identify an image. Having it in the
rootfs of the image as well as deployed alongside the image artifact
can be useful. Allow to deploy it alongside the image by setting
'IMAGE_BUILDINFO_DEPLOY' to '1'.
I use this in an automated testing workflow with unreliable deploy step
to ensure the deployment succeeded. With
IMAGE_BUILDINFO_VARS += "IMAGE_NAME"
IMAGE_BUILDINFO_DEPLOY = "1"
the generated buidinfo file is unique and the deploy step can be
verified by comparing (the hash of) the file on the booted system and in
`IMAGE_DEPLOY_DIR`.
Signed-off-by: Daniel Wagenknecht <dwagenknecht@emlix.com> Signed-off-by: Mathieu Dubois-Briand <mathieu.dubois-briand@bootlin.com> Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>
Sam Kent [Thu, 30 Apr 2026 08:39:05 +0000 (09:39 +0100)]
image_types_wic: add selftests for do_rootfs_wicenv hash safety
Add WicEnvHashTests to the wic selftest suite to cover the fix for
bug 15662:
- test_bblayers_excluded_from_task_hash: static check via tinfoil that
BBLAYERS appears in do_rootfs_wicenv[vardepsexclude].
- test_hash_does_not_change_when_empty_layer_added_to_bblayers: runs
with -S none, adds a layer containing no wic plugins, and asserts the
sigdata hash is unchanged.
- test_hash_changes_when_wic_plugin_added_to_layer: the inverse — adds
a layer that contains a wic plugin and asserts the hash does change.
Signed-off-by: Sam Kent <sam.john.kent@gmail.com> Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>
Sam Kent [Tue, 5 May 2026 18:28:44 +0000 (19:28 +0100)]
oeqa/selftest: add oe-selftest for kernel module pre-filter
Add selftest-ko-filter recipe and a PackageTests.test_kmodule_prefilter
oe-selftest that builds it to verify process_split_and_strip_files()
handles .ko.xz/.ko.gz alongside a real .ko without corrupting them.
Regression test for YOCTO #2348.
AI-Generated: Uses Claude Sonnet 4.6
Signed-off-by: Sam Kent <sam.john.kent@gmail.com> Signed-off-by: Mathieu Dubois-Briand <mathieu.dubois-briand@bootlin.com> Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>
Sam Kent [Tue, 5 May 2026 18:28:43 +0000 (19:28 +0100)]
oelib: add unit tests for kernel module detection helpers
Add tests for the filename pre-filter logic, is_kernel_module(), and
is_kernel_module_signed() to meta/lib/oeqa/selftest/cases/oelib/
Signed-off-by: Sam Kent <sam.john.kent@gmail.com> Signed-off-by: Mathieu Dubois-Briand <mathieu.dubois-briand@bootlin.com> Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>
Changqing Li [Fri, 15 May 2026 08:37:50 +0000 (16:37 +0800)]
report-error.bbclass: add site.conf/toolcfg.conf into error report
We had add local.conf/auto.conf into error report to help more easy to
reproduce a failure through error report, recently oe add a new default
configuration file toolcfg.conf, this commit also add site.conf and
toolcfg.conf in error report.
With this commit, the error report file will have the below snip:
[snip]
"site_conf": "DL_DIR ?= \"/bitbake-setup/bitbake-builds/.bitbake-setup-downloads\"\nSSTATE_DIR ?= \"/bitbake-setup/bitbake-builds/.sstate-cache\"\nBB_HASHSERVE_DB_DIR ?= \"${SSTATE_DIR}\"\n",
"target_sys": "x86_64-poky-linux",
"toolcfg_conf": "OE_FRAGMENTS += \"distro/poky machine/qemux86-64\"\n"
[snip]
[YOCTO #13252]
Signed-off-by: Changqing Li <changqing.li@windriver.com> Signed-off-by: Mathieu Dubois-Briand <mathieu.dubois-briand@bootlin.com> Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>
Jaeyoon Jung [Mon, 11 May 2026 03:11:51 +0000 (12:11 +0900)]
oeqa/selftest/pkgdata: add a test for variable consistency in multilib
There was a case where a package's LICENSE in a multilib build is not
expanded properly and it ends up with fallback to the main LICENSE value
which is unexpected. This adds a selftest for that; verify the pkgdata
variable LICENSE is identical between the base and multilib packages.
Jaeyoon Jung [Mon, 11 May 2026 03:11:50 +0000 (12:11 +0900)]
multilib: extend variables after packages split
A new function do_rename_package_variables() that invokes
ClassExtender.rename_package_variables() is appended to
PACKAGESPLITFUNCS in order to extend per-package variables in
PACKAGEVARS for all packages including dynamically split ones.
Pratik Farkase [Fri, 8 May 2026 13:16:15 +0000 (15:16 +0200)]
liburcu: ptest: add regression test suite
Add the upstream regression tests to the existing ptest for full
coverage of both unit and regression suites:
- rcutorture stress/perf tests across all RCU flavors (membarrier,
mb, bp, qsbr) with global, per-cpu, and per-thread configurations
- test_urcu_fork and test_urcu_fork_cxx for fork+RCU interaction
Move liburcu from PTESTS_FAST to PTESTS_SLOW as the regression tests
add ~6 minutes of runtime (total 379s).
Tested on qemux86-64 with ptest-runner: 122 PASS, 0 FAIL (379s).
Pratik Farkase [Wed, 6 May 2026 18:51:28 +0000 (20:51 +0200)]
gmp: add ptest support
Add ptest support for gmp, running the full upstream unit test suite
(200 tests) covering mpz, mpn, mpf, mpq, rand, misc, and C++ bindings.
Tests are compiled with 'make check TESTS=' to build without executing,
then installed via libtool to the ptest directory preserving the upstream
subdirectory structure.
Tested on qemux86-64 with ptest-runner: 200 PASS, 0 FAIL (268s).
Pratik Farkase [Tue, 5 May 2026 10:45:12 +0000 (12:45 +0200)]
liburcu: add ptest support
Add ptest support for liburcu, running the upstream unit test suite
(674 TAP assertions across 24 tests). This covers architecture checks,
atomic operations, multiflavor builds, sysfs helpers, and build
verification tests.
Tested on qemux86-64 with ptest-runner: 24 PASS, 0 FAIL.
Pratik Farkase [Tue, 19 May 2026 09:08:16 +0000 (11:08 +0200)]
zstd: add ptest support
Add comprehensive ptest support for zstd running the full upstream test
suite covering compression correctness, streaming, dictionaries, CLI
behavior, and fuzz testing.
Tests included:
- fullbench: validates all internal compression/decompression functions
- poolTests: thread pool ordering and synchronization
- invalidDictionaries: dictionary error handling
- legacy: legacy format (v0.4+) compatibility
- fuzzer: randomized compression/decompression (200s, ~700+ rounds)
- zstreamtest: streaming API fuzzer (90s, ~250+ rounds)
- decodecorpus: random frame generation and decode (30s, ~450+ rounds)
- playTests.sh: comprehensive CLI validation including compression,
decompression, multithreading, dictionaries, sparse files, frame
concatenation, long distance matching, benchmarks, and round-trips
- cli-tests: 36 Python-based CLI integration tests covering
compression, decompression, progress, file-stat, dictionaries,
symlinks, and file handling
Tests excluded from cli-tests (environment limitations):
- compression/levels.sh: requires >8GB for --ultra/--max levels
- compression/window-resize.sh: requires >4GB for 1GB window
- file-stat/compress-file-to-dir-without-write-perm.sh: ptest runs
as root which bypasses permission checks
- cltools/: requires GNU grep (--label), unavailable with BusyBox
Source .c/.h files from tests/ and programs/ are installed as training
corpus for the dictionary builder tests.
zstreamtest requires 4096MB RAM for (hashLog=29 allocates ~2.8GB).
Set QB_MEM:virtclass-mcextend-zstd = "-m 4096" in core-image-ptest.
Tested on qemux86-64 (4GB) with ptest-runner: 9 PASS, 0 FAIL. Runtime is ~13 min.
Chen Qi [Mon, 18 May 2026 05:55:56 +0000 (13:55 +0800)]
kmod: inherit upstream-stable-release-point
Upgrades like 34 -> 34.1 or 34.1 -> 34.2 for kmod are stable upgrades.
* Evidence
Its repo contains a branch called:
origin/kmod-34.y
Its tags are like:
v32
v33
v34
v34.1
So we can infer that the maintenance policy for kmod is that
upgrades such as 33 -> 34 contains features and upgrades such
as 34 -> 34.1 is a stable upgrade. And such maintenance policy
is likely to continue.
mark.yang [Tue, 24 Mar 2026 23:27:34 +0000 (08:27 +0900)]
llvm: add binutils-plugin PACKAGECONFIG for LLVMgold.so
Add a 'binutils-plugin' PACKAGECONFIG option to selectively build
LLVMgold.so by passing -DLLVM_BINUTILS_INCDIR and adding binutils
to DEPENDS only when enabled.
LLVMgold.so is the LLVM LTO plugin for GNU ld BFD.
When using lld as the linker, this plugin is not needed since lld has
built-in LTO support. However, for toolchains that rely on BFD
this plugin is required to enable LTO.
binutils-plugin is named after Gentoo's USE flag convention for similar
functionality.
Zhixiong Chi [Tue, 19 May 2026 12:49:42 +0000 (20:49 +0800)]
python3-cython: make generated _cyutility.c be reproducible
While python3 module use cython to build shared moudle utility library,
the generated source file _cyutility.c is not stable at each build and
made the generated library not be reproducible.
This commit use the fixed relative buildpath instead of the original
unpredictable tmpdir in the generated source file to assure the file
_cyutility.c should be reproducible.
Signed-off-by: Zhixiong Chi <zhixiong.chi@windriver.com> Signed-off-by: Antonin Godard <antonin.godard@bootlin.com> Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>
Chen Qi [Mon, 18 May 2026 05:55:53 +0000 (13:55 +0800)]
upstream-stable-release-point.bbclass: add bbclass for stable point upgrade
If a recipe can do stable version upgrade and the stable parts of the version
is seperated by '.', then it can inherit this bbclass.
By default, the stable parts number is 2, which means the following upgrades
are stable version upgrades:
x.y.z -> x.y.z+1
x.y.z+1 -> x.y.z+1.zz
x.y.z+1.zz -> x.y.z+2
Recipes that have different stable version parts can also inherit this bbclass
and set STABLE_VERSION_PARTS. For example, systemd sets this variable to "1".
For recipes whose stable version part is not separated by '.', they should not
inherit this bbclass and intead set UPSTREAM_STABLE_RELEASE_REGEX themselves.
For example, openssh's stable part is separted by 'p' and should not inherit
this bbclass.
Mark Hatle [Wed, 3 Jun 2026 21:57:39 +0000 (16:57 -0500)]
pseudo: Update to version 1.9.8
Changelog:
Makefile.in: Bump to 1.9.8
pseudo_client.h: Fix typo in the comment
client: permissions drop setuid and setgid
tests: Add setuid permission check
pseudo_client.h: Add +s to PSEUDO_DB_MODE for mkdir
tests: Add test that returned stat is correct
pseudo_client.h: Make it clear both macros must be updated together
Makefile.in: Add pseudo_client.h as a dependency
Signed-off-by: Mark Hatle <mark.hatle@kernel.crashing.org> Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>
The package_tar class was removed in 90ce19122802
("meta/classes-global: remove package_tar.bbclass"), remove the deltask
for the nopackages class as well.
The package_tar.bbclass was removed in 90ce19122802
("meta/classes-global: remove package_tar.bbclass"), remove the
documentation variable associated with it as well.
Remove a [True, False] list evaluation that itself relied on evaluating
a boolean condition. Instead pass the boolean condition directly.
The previous version works because int(True)=1 and int(False)=0.
So:
=> pkg_type = Manifest.PKG_TYPE_ATTEMPT_ONLY
=> [False, True][pkg_type == Manifest.PKG_TYPE_ATTEMPT_ONLY]
=> [False, True][True]
=> [False, True][1]
=> True
The install_order instance variable was only ever set to
Manifest.INSTALL_ORDER, modified nowhere and used as is. Instead of
using an instance variable use Manifest.INSTALL_ORDER directly to
simplify the code for those reading it.
The install_order instance variable was only ever set to
Manifest.INSTALL_ORDER, modified nowhere and used as is. Instead of
using an instance variable use Manifest.INSTALL_ORDER directly to
simplify the code for those reading it.
Ross Burton [Mon, 1 Jun 2026 14:07:04 +0000 (15:07 +0100)]
curl: fix mbedtls detection
The mbedtls detection logic in curl is broken and resulted in build
paths leaking into curl-config and libcurl.pc. Backport a patch to fix
the detection by looking for a symbol that wasn't removed in mbedtls 3.0
five years ago, and remove the explicit sysroot reference as it is no
longer needed.
Signed-off-by: Ross Burton <ross.burton@arm.com> Signed-off-by: Mathieu Dubois-Briand <mathieu.dubois-briand@bootlin.com> Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>
oeqa/selftest: test_testimage_virgl_gtk_sdl: run only if gtk or sdl are enabled in base config
The test was forcibly enabling both gtk and sdl in qemu and then running the test.
This changes the logic to skip the test if neither is enabled,
and run gtk and/or sdl tests only if either is already enabled
in the base config that was in place before running the selftest.
Together with disabling sdl support in qemu by default this effectively
skips the test on the yocto autobuilder, resolving the selftest
failures seen with qemu 11.0.
Signed-off-by: Alexander Kanavin <alex@linutronix.de> Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>
Brings following fixes: a255c1ed36a1 [WebAssembly] Avoid crash in LateEHPrepare with empty cleanup pads (#200322) a2b779045093 [SystemZ] Fix off-by-one error in backend (#200141) 2615295c15ba [libc++] Fix multi{map,set}::extract not returning the first matching element (#199703) a760d7c07024 [LLD] [COFF] Fix handling of immediates in ARM64_SECREL_HIGH12A (#200060) 294ae8d1e62a [PowerPC] Drop invalid range metadata when lowering i64 load to fp in INT_TO_FP (#198705) 6db0725886bb [X86] lowerV64I8Shuffle - avoid lowerShuffleAsRepeatedMaskAndLanePermute call on VBMI targets (#183109) c43c9b426055 [X86] lowerV64I8Shuffle - prefer VPERMV3 byte shuffles to OR(PSHUFB,PSHUFB) on VBMI targets (#182852) 6ac68778a76e [X86] Add test coverage for #137422 (#182832) ad9524f38657 [PowerPC] Fix i128 vcmpequb optimization for loads with range metadata and small constants (#196801) 48967cc1a0d4 build: adjust LLDB and clang library naming on Windows (#185084) 8798085803f1 [libc] Demote compiler check error to a warning (#198033) 6e5effc4990b [LoongArch] Revert "Add patterns to support vector type average instructions generation" (#198306) 4d5dffb59420 [ELF] Initialize Symbol fields in the constructor instead of via memset (#198129) 823afe046164 Reland [C++20] [Modules] Don't profiling the callee of CXXFoldExpr (#190732) (#195983) e624f1204282 [CoroSplit] Never collect allocas used by catchpad into frame (#186728) 81c69e140401 Bump version to 22.1.7
Signed-off-by: Sunil Dora <sunilkumar.dora@windriver.com> Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>
rust-target-config:
limit llvm-floatabi to ARM targets
Generate `llvm-floatabi` only for ARM/ARMv7 targets and set the
corresponding Rust target ABI (`eabi` or `eabihf`) according to
`TARGET_FPU`.
`llvm-floatabi` is an ARM-specific LLVM setting (Rust currently
uses it only for ARM targets) and should not be emitted for
other architectures.
https://github.com/rust-lang/rust/blob/main/compiler/rustc_target/src/spec/mod.rs#L2619
Following upstream Rust changes that tightened validation of
target ABI fields, ARM targets must provide an explicit ABI in
addition to the float ABI configuration. Match the ABI settings
used by Rust's built-in ARM target specifications when
generating target JSON files.
Upstream-Reference:
https://github.com/rust-lang/rust/commit/40ebcc031da647d6a76deaf05c471b7a1d4aa228
("Rename target.abi to target.cfg_abi and enum-ify llvm_abiname")
Drop merged patches:
0001-Update-call-llvm-intrinsics-test.patch
0001-Fix-multiple-option-or-permutations-test-for-big-end.patch
Updated patch:
rust-oe-selftest.patch
New patch:
0003-explicit-tail-calls-disable-two-tests.patch
Backport two commits that disable two failing tests on LoongArch.
Selftest Changes:
Ignore one failing test for arm
tests/ui/c-variadic/same-program-multiple-abis-arm.rs
The test is marked ignore-thumb and contains ARM-mode assembly (stmib sp, {r2, r3}).
Our target enables +thumb-mode by default, causing LLVM to assemble the naked_asm!
block in Thumb state, where the instruction is invalid. The resulting failure is due
to ARM-vs-Thumb assembly mode, not the c-variadic ABI behavior being tested. Therefore
the test is not applicable to our target configuration.
Ignore one failing test for riscv64
tests/assembly-llvm/riscv-redundant-memory-stores.rs
https://github.com/rust-lang/rust/issues/157074
Excluded tests:
+--------+----------+---------+
| Before | After | Skipped |
+---------+---------+---------+
| 39 | 41 | +2 |
+--------+----------+---------+
Ross Burton [Thu, 14 May 2026 13:37:47 +0000 (14:37 +0100)]
avahi: upgrade 0.8 -> 0.9-rc4
As per upstream[1], 0.9-rc4 is pretty much what is going to be in the
0.9 release, and upgrading to this now will make upgrading to the final
release a lot simplier.
As there are no tarballs for the release candidates, switch to git. This
means we need to disable the manpages as they need xmltoman to build.
Delete all of the backported CVE patches as they're included in this
release. The cvelistv5 database has version information on all of these
CVEs, so they don't reappear in reports.
Delete invalid-service.patch, fixed upstream differently in [2].
Upstream has deleted the in-tree init scripts, so copy them from before
deletion and add alongside the recipe.
Consolidate the systemd options, and depend on libsystemd as avahi links
to that instead of duplicating the code.
Pass --runstatedir so that /run is used as expected.
[1] https://github.com/avahi/avahi/issues/503#issuecomment-4443016376
[2] avahi 93b1436 ("core: no longer supply bogus services to callbacks")
Signed-off-by: Ross Burton <ross.burton@arm.com> Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>
Richard Purdie [Wed, 3 Jun 2026 17:27:34 +0000 (18:27 +0100)]
ppp: Remove PD license
The code referred to as having a "public domain" license (chat) has a
MIT SPDX license indentifier and matching license text. We can therefore
drop the PD license field entry.
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>
Joshua Watt [Mon, 1 Jun 2026 18:11:38 +0000 (12:11 -0600)]
glib-2.0: Change from Public Domain to Gnome GCR Documentation License
The file docs/reference/COPYING is an exact match for the text of the
Gnome GCR Documentation License (SPDX ID GCR-docs), so use that license
instead of Public Domain. This licenses actually requires that the
copyright notice be preserved, unlike Public Domain, so its likely that
using Public Domain is actually an error due to a misreading of the
license text.
Signed-off-by: Joshua Watt <JPEWhacker@gmail.com> Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>
Joshua Watt [Mon, 1 Jun 2026 18:11:36 +0000 (12:11 -0600)]
mobile-broadband-provider-info: Fix license to be CC-PDDC
Careful reading of the license text reveals that it is actually the
exact wording of the Creative Commons Public Domain Dedication and
Certification license, which has the SPDX identifier CC-PDDC
Signed-off-by: Joshua Watt <JPEWhacker@gmail.com> Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>
Pratik Farkase [Thu, 21 May 2026 07:49:10 +0000 (09:49 +0200)]
libxslt: add ptest support
Add ptest support for libxslt, running the upstream runtest binary which
validates XSLT transformations across REC conformance, general, encoding,
documents, numbers, keys, namespaces, extensions, reports, and EXSLT
function tests.
Tested on qemux86-64 with ptest-runner: 748 tests, 0 errors (9s).
Pratik Farkase [Thu, 21 May 2026 08:15:05 +0000 (10:15 +0200)]
libatomic-ops: add ptest support
Add ptest support for libatomic-ops, running the upstream test suite
which validates atomic operations, generalized emulation, pthread-based
fallbacks, lock-free stack, and memory allocation.
Tested on qemux86-64 with ptest-runner: 5 PASS, 0 FAIL.
Chen Qi [Mon, 18 May 2026 05:55:52 +0000 (13:55 +0800)]
recipeutils: add optional stable_upgrade parameter to get_recipe_upgrade_status
We want the ability to do stable version upgrades for recipes.
To this end, add an optional stable_upgrade parameter to the
get_recipe_upgrade_status function, which defaults to False and
when enabled will try to get the latest stable version of the recipe.
The UPSTREAM_STABLE_RELEASE_REGEX is respected. If a recipe sets
it, it will be used as the filter_regex. If it's not set explicitly,
it means that there's no stable updates or the recipe hasn't been
checked yet.
Daniel Turull [Mon, 25 May 2026 08:01:05 +0000 (10:01 +0200)]
devtool: upgrade: extract changelog between versions
Automatically extract changelog information when upgrading a recipe.
Uses the devtool-base tags created during upgrade to diff known
changelog files (NEWS, ChangeLog, CHANGES, etc.) between the old and
new versions. For git-based sources, falls back to git log --oneline
if no changelog file changed.
Prioritize per-version release notes files containing the new version
number (e.g. v2.42.1-ReleaseNotes, changelog-9.20.23.rst) over generic
changelog files. When multiple per-version files match, all are included.
For standard changelog files, picks the one with the most new content.
Resolve RST .. include:: directives by searching the source tree for
the included file by basename. Strip RST comment blocks (copyright
headers) from .rst files.
Collapse runs of multiple blank lines and strip leading/trailing
whitespace for cleaner output suitable for commit messages.
Output is written to workspace/changelogs/<pn>.txt and cleaned up on
devtool reset. This allows AUH and other tools to pick up the changelog
without implementing their own extraction logic.
Add test cases for existing test recipes.
Tested with AUH with changelog support (See yocto-patches)
master 2026-05-22:
commits in https://git.openembedded.org/openembedded-core-contrib/log/?h=dturull/devtool-changelog-test-master
- Total recipes attempted: 239
- With meaningful changelog: 191 (80%)
- Without changelog: 48 (20%)
Recipes without changelogs:
- acpica: devtool upgrade failed
- autoconf: devtool upgrade failed
- busybox: devtool upgrade failed
- cargo-c: devtool upgrade failed
- debugedit: devtool upgrade failed
- diffoscope: only PKG-INFO changed
- gcc-source-15.2.0: devtool upgrade failed
- gdb, gdb-cross-x86_64, gdb-cross-canadian-x86-64: devtool upgrade failed
- go-cross-x86-64-v3: devtool upgrade failed
- groff: devtool upgrade failed
- gtk-doc: devtool upgrade failed
- icu: no changelog files in source
- iproute2: no changelog files in source
- libcap: no changelog files in source
- libdrm: no changelog files in source
- libmpc: devtool upgrade failed
- libpciaccess: no changelog files in source
- librsvg: devtool upgrade failed
- linux-firmware: binary blobs, no changelog in tarball
- llvm-project-source: devtool upgrade failed
- nfs-utils: no changelog file changed
- python3-certifi: only CA bundle + PKG-INFO changed
- python3-dtschema: no changelog in PyPI sdist
- python3-editables: no changelog in PyPI sdist
- python3-hypothesis: no changelog in PyPI sdist
- python3-jsonpointer: no changelog in PyPI sdist
- python3-maturin: devtool upgrade failed
- python3-pdm-backend: no changelog in PyPI sdist
- python3-poetry-core: no changelog in PyPI sdist
- python3-pytz: only timezone data changed
- python3-shacl2code: no changelog in PyPI sdist
- python3-spdx-python-model: devtool upgrade failed
- python3-sphinxcontrib-svg2pdfconverter: no changelog in PyPI sdist
- python3-trove-classifiers: only classifiers list changed
- python3-uv-build: devtool upgrade failed
- python3-wcwidth: only unicode tables changed
- python3-xmltodict: no changelog in PyPI sdist
- rpm-sequoia: devtool upgrade failed
- spirv-llvm-translator: devtool upgrade failed
- sqlite3: amalgamation tarball, no changelog
- time: devtool upgrade failed
- vte: devtool upgrade failed
- wayland: devtool upgrade failed
- wayland-protocols: no changelog files in source
- wireless-regdb: no changelog files in source
- xwayland: no changelog in tarball
AI-generated: kiro with claude-opus-4.6 model Signed-off-by: Daniel Turull <daniel.turull@ericsson.com> Signed-off-by: Mathieu Dubois-Briand <mathieu.dubois-briand@bootlin.com> Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>
Mingli Yu [Mon, 1 Jun 2026 01:58:45 +0000 (09:58 +0800)]
valgrind: Upgrade 3.27.0 -> 3.27.1
This release contains only bug fixes and thhe following bugs have
been fixed or resolved in this release.
519574 valgrind 3.27 "--fair-sched=yes" does not work
519613 Valgrind incorrectly unpacks the result of sys_port (port_getn)
on error, leading to a ~60s wallclock time delay on every call
n-i-bz Update vg-lifespan (copyright) years
n-i-bz Use SSizeT for VG_(readlink) result in VG_(realpath)
To see details of a given bug, visit
https://bugs.kde.org/show_bug.cgi?id=XXXXXX
where XXXXXX is the bug number as listed above.
Signed-off-by: Mingli Yu <mingli.yu@windriver.com> Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>
Tim Orling [Sun, 31 May 2026 20:09:44 +0000 (13:09 -0700)]
python3-packaging: upgrade 26.0 -> 26.2
* Add python3-hypothesis to ptest REDEPENDS
26.2 - 2026-04-24
Fixes:
* Fix incorrect sysconfig var name for pyemscripten in (PR #1160)
* Make Version, Specifier, SpecifierSet, Tag, Marker, and Requirement
pickle-safe and backward-compatible with pickles created in 25.0-26.1
(including references to the removed packaging._structures module)
(PR #1163, PR #1168, PR #1170, PR #1171)
* Re-export ExceptionGroup in metatadata for now in (PR #1164)
26.1 - 2026-04-14
Fixes:
* Fix > comparison for versions with dev+local segments in (PR #1097)
* Fix incorrect self-comparison for InfinityType and NegativeInfinityType
in (PR #1093)
* Canonicalize when deduplicating specifiers in SpecifierSet in
(PR #1109)
* Fix charset error message formatting in (PR #1121)
* Handle the key parameter in SpecifierSet.filter when specifiers are
empty and prerelease is False in (PR #1096)
* Standardize inner components of repr output in (PR #1090)
* Specifier’s === uses original string, not normalized, when available
in (PR #1124)
* Propagate int-max-str-digits ValueError in (PR #1155)
For additional changes, see [1].
For full comparison, see [2].
Tan Siewert [Wed, 27 May 2026 11:06:20 +0000 (13:06 +0200)]
tune-cortexa9: add tunes without VFP enabled
VFP is optional on armv7a and almost always included, however the HPE
GXP SoC omits it. This SoC is used on HPE ProLiant server motherboards
as the out-of-band management controller (iLO).
Signed-off-by: Tan Siewert <tan.siewert@9elements.com> Signed-off-by: Mathieu Dubois-Briand <mathieu.dubois-briand@bootlin.com>
Jesse Van Gavere [Wed, 27 May 2026 08:31:30 +0000 (10:31 +0200)]
rust: export CARGO_BUILD_TARGET
To cross compile a rust project with the SDK one currently has to know
the associated target sys and run cargo build with it, by setting
CARGO_BUILD_TARGET to the rust target sys, running cargo build after
sourcing the SDK will immediately have the correct target.
See https://doc.rust-lang.org/cargo/reference/config.html#buildtarget
Signed-off-by: Adam Duskett <adam.duskett@amarulasolutions.com> Signed-off-by: Jesse Van Gavere <jesse.vangavere@teledyne.com> Signed-off-by: Mathieu Dubois-Briand <mathieu.dubois-briand@bootlin.com>
Trevor Gamblin [Fri, 29 May 2026 14:46:52 +0000 (10:46 -0400)]
python3: use SKIPPED_TESTS instead of test skip patches
Drop all 13 patches carried for skipping various known ptest failures
and replace them with usage of the SKIPPED_TESTS variable, which we have
started doing in recent commits (e.g. 40b0815dc9). Most but not all of
the test skips are due to load variability, distinguish between them by
using multiple SKIPPED_TESTS:append blocks with comments taken from the
patch files indicating their purpose. This adds 62 lines to the Python 3
recipe, but it also means that we no longer have to maintain the patch
files across version upgrades, where they often break due to changes in
the test modules upstream.
Results look the same as those from my last upgrade run:
Jörg Sommer [Sun, 31 May 2026 16:04:59 +0000 (18:04 +0200)]
python3-idna: Upgrade 3.13 -> 3.17
* 3.17 (2026-05-28)
- Substantial 75% reduction in memory usage through new data
structures and some optimization in processing speed.
- Added a general 1024-character input length cap to the public
validation, conversion, and codec entry points. This is well above
any legitimate domain or label and guards against pathological
inputs.
* 3.16 (2026-05-22)
- Add a command-line interface (`python -m idna`, also available as
the `idna` script). Encodes or decodes one or more domains supplied
as arguments or on standard input, with options to select A-label
or U-label output and control error handling.
- Raise the minimum supported Python version to 3.9
- Various code quality improvements
* 3.15 (2026-05-12)
- Enforce DNS-length cap on individual labels early in `check_label`,
short-circuiting contextual-rule processing for oversized input
while staying compatible with UTS 46 usage.
- Tidy core helpers: hoist bidi category sets to module-level
frozensets (avoiding per-codepoint list construction), simplify
length checks, and reuse the shared `_unicode_dots_re` from
`idna.core` in the codec module.
- Use `raise ... from err` for proper exception chaining and
switch internal string formatting to f-strings.
- Allow `flit_core` 4.x in the build backend.
- Expand the ruff lint set (flake8-bugbear, flake8-simplify,
pyupgrade, perflint) and apply the surfaced fixes; pin lint CI
to Python 3.14.
- Add Dependabot configuration for GitHub Actions.
- Convert README and HISTORY from reStructuredText to Markdown.
- Reference CVE-2026-45409 for the 3.14 advisory in place of the
initial GHSA identifier.
Thanks to Felix Yan, Stan Ulbrych, and metsw24-max for
contributions to this release.
* 3.14 (2026-05-10)
- Removed opportunity to process long inputs into quadratic
time by rejecting oversize inputs up-front. Closes a bypass
of the CVE-2024-3651 mitigation. [CVE-2026-45409]
Thanks to Stan Ulbrych for reporting the issue.
Signed-off-by: Jörg Sommer <joerg.sommer@navimatix.de> Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>
Tim Orling [Sun, 31 May 2026 00:17:55 +0000 (17:17 -0700)]
python3-wcwidth: upgrade 0.6.0 -> 0.7.0
Changes [1]:
* New support for kitty text sizing protocol (OSC 66) in width() and
clip().
* New clip() parameter control_codes='parse', 'ignore', and 'strict'.
clip() is now able to clip OSC 8 hyperlinks and OSC 66 text sizing
sequences.
* Improved clip() and width() to support horizontal cursor sequences
(cub, cuf, hpa). Cursor-left (cub) or backspace (\b) now overwrites
text. column_address (hpa) and carriage return (\r) are now parsed,
and more values conditionally raise ValueError when
control_codes='strict'.
PR's:
* Remove docs, add utils by @jquast in #209
* Bump requests from 2.32.5 to 2.33.0 in /docs by @dependabot[bot] in
#210
* Bump pygments from 2.19.2 to 2.20.0 in /docs by @dependabot[bot] in
#212
* dependabot nonsense by @jquast in #215
* Expand terminal escape sequence for three more ECMA-48 "families" by
@jquast in #214
* Improve clip() and width() with hyperlinks and overtyping by @jquast
in #216
* Improve width() and clip() with kitty Text Sizing Protocol by @jquast
in #213
Tim Orling [Sun, 31 May 2026 00:17:54 +0000 (17:17 -0700)]
python3-snowballstemmer: upgrade 3.0.1 -> 3.1.0
git shortlog:
Dmitry Shachnev (5):
python: Drop Python 2 support, require Python ≥ 3.3
python: Simplify code after dropping Python 2 support
python: Add a minimal pyproject.toml file
Stop excluding classifiers for Armenian and Yiddish
stemwords.py: Replace deprecated codecs.open() with built-in open()
Henry Schreiner (1):
Protect empty languages dict
Olly Betts (11):
python: Fix algorithms() when wrapping PyStemmer
python: Reduce overhead of forwarding to PyStemmer
stemwords.py: Make -i and -o optional
Python: Optimise string test
Avoid unnecessary cursor update in among helpers
Python: Change slice_check() to assert conditions
Python: Remove parentheses from assert
Update slice end after `delete` and `<-`
Python: Remove deprecated licence classifier
stemwords.py: Remove now-unused `import codecs`
Update for 3.1.0
Tim Orling [Sun, 31 May 2026 00:17:53 +0000 (17:17 -0700)]
python3-rpds-py: upgrade 0.30.0 -> 2026.5.1
git shortlog (ignoring merge, dependabot and pre-commit-ci commits):
Edgar Ramírez Mondragón (1):
Update to PyO3 0.28.0
Julian Berman (10):
Update pre-commit hooks.
Use regular gh CLI for release notes.
Run zizmor in pedantic mode and address findings.
Drop support for 3.10. Add 3.15.
Use cargo-release for releases, without the chore: prefix on commit messages.
Run each nox session in its own job with a single Python installed.
Reserve Mach-O header padding on macOS so install_name_tool can relocate.
Bump url-py to a version that supports Python 3.15.
Regenerate uv.lock without user-config artifacts.
Release rpds-py version 2026.5.1
Tim Orling [Sun, 31 May 2026 00:17:52 +0000 (17:17 -0700)]
python3-pytz: upgrade 2026.1 -> 2026.2
git shortlog:
Fullchee Zhang (1):
fix typo
Shaun Walbridge (1):
try to access resource using `importlib.resources`
Stuart Bishop (6):
Bump version number to 2026.1.post1
Update test runners for new Pythons and github actions
Squashed 'tz/' changes from dd6be6d155..8be0d5483d
IANA 2026b
Bump version numbers to 2026.2 (IANA 2026b)
Bump github actions/checkout to @v6
Stub (2):
Try to access resource using `importlib.resources`
Fix typo in README
Tim Orling [Sun, 31 May 2026 00:17:48 +0000 (17:17 -0700)]
at-spi2-core: upgrade 2.60.3 -> 2.60.4
What's new in at-spi2-core 2.60.4 [1]:
* AtspiDeviceA11yManager: Make use of pid when returned by queryPointer
* Fix deregistering application-specific event listeners
* Drop 0001-bump-uv_build-to-0.11.0.patch; change is included in upgrade.
* Refresh python3-crytpography-crates.inc
48.0.0 - 2026-05-04 [1]
* BACKWARDS INCOMPATIBLE: Support for Python 3.8 has been removed.
cryptography now requires Python 3.9 or later.
* BACKWARDS INCOMPATIBLE: Loading an X.509 CRL whose inner
TBSCertList.signature algorithm does not match the outer
signatureAlgorithm now raises ValueError. Previously, such CRLs
were parsed successfully and only rejected during signature validation.
* Added support for ML-KEM key encapsulation and ML-DSA signing when
using OpenSSL 3.5.0 or later, in addition to the existing AWS-LC and
BoringSSL support. This means post-quantum algorithms are now available
to users of our wheels.
- Note: Going forward, we do not guarantee that all functionality in
cryptography will be available when building against OpenSSL. See
The State of OpenSSL for pyca/cryptography for more information.
47.0.0 - 2026-04-24 [2]
* Support for Python 3.8 is deprecated and will be removed in the next
cryptography release.
* BACKWARDS INCOMPATIBLE: Support for binary elliptic curves
(SECT* classes) has been removed. These curves are rarely used and
have additional security considerations that make them undesirable.
* BACKWARDS INCOMPATIBLE: Support for OpenSSL 1.1.x has been removed.
OpenSSL 3.0.0 or later is now required. LibreSSL, BoringSSL, and
AWS-LC continue to be supported.
* BACKWARDS INCOMPATIBLE: Dropped support for LibreSSL < 4.1.
* BACKWARDS INCOMPATIBLE: Loading keys with unsupported algorithms or
keys with unsupported explicit curve encodings now raises
UnsupportedAlgorithm instead of ValueError. This change affects
load_pem_private_key(), load_der_private_key(), load_pem_public_key(),
load_der_public_key(), and public_key() when called on certificates
with unsupported public key algorithms.
* BACKWARDS INCOMPATIBLE: When parsing elliptic curve private keys, we
now reject keys that incorrectly encode a private key of the wrong
length because such keys are impossible to process in a constant-time
manner. We do not believe keys with this problem are in wide use,
however we may revert this change based on the feedback we receive.
* Deprecated passing 64-bit (8-byte) and 128-bit (16-byte) keys to
TripleDES. In a future release, only 192-bit (24-byte) keys will be
accepted. Users should expand shorter keys themselves (e.g., for
single DES: key + key + key, for two-key: key + key[:8]).
Tim Orling [Sat, 30 May 2026 22:42:32 +0000 (15:42 -0700)]
python3-uv-build: upgrade 0.11.14 -> 0.11.17
0.11.17 [1]
Released on 2026-05-28.
Bug fixes
---------
* Improve the performance of large entries in tool.uv.conflicts
(#19538)
* Avoid modifying the parent process' env with --env-file in uv
run (#19567)
* Fix script environment creation for scripts with long filenames
(#19539)
* Fix transitive Git archive dependencies in lockfiles (#19589)
* Preserve Git repository URLs in direct URL metadata (#19590)
* Support redirects in --check-url (#19594)
* Accept case-insensitive HTML tags in --find-links parsing (#19537)
* Reject duplicate script metadata blocks (#19544)
* Ban names like "python3" as script entry points (#19535, #19536)
* Validate Git LFS artifacts for Git archives (#19592)
* Use a relative path when creating symlinks in cache to improve
relocatability (#19033)
See [1] for the rest of 0.11.17 changes.
0.11.16 [2]
Released on 2026-05-21.
Bug fixes
---------
* Allow environment variables that take a list to be empty (#19503)
* Ensure that incompatible wheel hints do not leak secrets (#19504)
* Reject unsafe entry points in uv-build (#19495)
* Restrict delimiters in entry point parsing (#19471)
* uv-netrc: fix multi-word no-space comment lines causing parse errors
(#19494)
See [2] for the rest of 0.11.16 changes.
0.11.15 [3]
Released on 2026-05-18.
Security
--------
* Fix a TAR parser differential, see GHSA-3cv2-h65g-fgmm (#19463)
* Enforce that entry points cannot escape in the scripts directory, see
GHSA-4gg8-gxpx-9rph (#19464)
Bug fixes
---------
* Apply workspace-member [tool.uv.sources] credentials under uv sync
--frozen (#19423)
* Skip empty directories in uv build outputs (#19437)
* Fix Git submodule handling when using relative paths (#12156)
* Fix line number reporting in netrc parsing (#19452)
Tim Orling [Sat, 30 May 2026 22:42:31 +0000 (15:42 -0700)]
python_uv_build: add PEP-517 backend class
uv_build is a slimmed down version of uv containing only the build
backend. See https://pypi.org/project/uv/ and https://docs.astral.sh/uv/
for the main project package and documentation.
python3-cryptography-vectors uses the 'uv_build' backend.
Signed-off-by: Tim Orling <tim.orling@konsulko.com> Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>
Peter Tatrai [Fri, 29 May 2026 17:01:17 +0000 (19:01 +0200)]
time64: enable 64-bit time/file-offset flags for 32-bit nativesdk
When SDKMACHINE is set to i686 or i586, nativesdk binaries are compiled
as 32-bit. Without -D_TIME_BITS=64 and -D_FILE_OFFSET_BITS=64, stat()
and time-related syscalls use 32-bit types, causing EOVERFLOW on
filesystems with large inode numbers (e.g. container overlay filesystems)
and Y2038 issues.
Add SDK_CC_ARCH appends for class-nativesdk:i686 and class-nativesdk:i586
using GLIBC_64BIT_TIME_FLAGS, mirroring how target architectures are
handled.
Signed-off-by: Peter Tatrai <peter.tatrai.ext@siemens.com>" Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>
Ross Burton [Mon, 17 Jun 2024 13:42:40 +0000 (13:42 +0000)]
oeqa/core/runner: stub addDuration in OETestResult
We have a custom TestResult implementation, and Python 3.12 added a new
method addDuration() to the TestResult interface. This would be useful
to implement correctly, but for now stub it out to silence the warning
when running under Python 3.12:
/usr/lib64/python3.12/unittest/case.py:580: RuntimeWarning: TestResult has no addDuration method
warnings.warn("TestResult has no addDuration method",
Signed-off-by: Ross Burton <ross.burton@arm.com> Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>
projects / thirdparty / openembedded / openembedded-core.git / log
