Upstream changes (NEWS):
* dump.exfat: prevent a heap buffer overflow in name parsing.
* Install lsdosattr and chdosattr in /bin instead of /sbin (covered by
the default FILES:${PN}, so packaging is unaffected).
* libexfat: use __builtin_popcountl() throughout and add an internal
exfat_count_used_clusters() helper (static lib, no ABI impact).
No dependency or build-system changes; configure.ac is unchanged, so
DEPENDS and the ptest install steps stay as-is.
While enabling the ptest image (meta-filesystems-image-ptest-exfatprogs)
on qemux86-64, the upcase_table test cases (present since 1.4.0, so this
is a pre-existing failure now exercised) failed because they used
`truncate` and `xxd -r - <file>`, which are unavailable in a BusyBox
core-image-minimal rootfs:
truncate: command not found
FAIL: to create corrupted image
Add 0001-tests-upcase_table-use-dd-printf-instead-of-truncate.patch,
which rewrites those scripts to size the image with dd and poke the
corrupted bytes with printf|dd (tools already present in the base
image). ptest result on qemux86-64 is now 23 of 23 passing.
Signed-off-by: Khem Raj <khem.raj@oss.qualcomm.com>
387 commits across 21 fortnightly releases (739 files changed,
+16397/-2042). 114 new tests added (xfs x44, generic x32, btrfs x16,
f2fs x14, ext4 x3, overlay x2, selftest x2, ceph x1) and one removed
(xfs/539).
Major themes:
* Multi-block atomic writes: new common/atomicwrites, fio atomic-write
helpers, fsx atomic-write support, and O_[D]SYNC / sudden-shutdown /
bigalloc / hardware-AWU atomic tests.
* XFS zoned / RT-device work: common/zoned with zloop helpers, zoned
growfs, zone-alignment mkfs checks, zone GC low-space, external and
internal realtime device support.
* New xfs_healer test family (background repair service).
* file_getattr(2)/file_setattr(2) support (new src/file_attr.c,
AC_HAVE_FILE_GETATTR probe), VFS delegation tests, and a new
fsnotify/fanotify error-reporting helper (src/fs-monitor.c).
* Infrastructure: treewide $MOUNT_PROG -> _mount conversion, temp
files moved to $TMPDIR, systemd-coredump collection in check.
Build-system: configure.ac gains AC_HAVE_FILE_GETATTR and a
btrfs_ioctl_received_subvol_args type check; src/Makefile adds
unlink-fsync, truncate, rw_hint, fs-monitor, btrfs_ioctl and a
conditional file_attr target. No new link libraries, so no DEPENDS
change is needed. LICENSE checksum unchanged.
Recipe patch changes:
* Drop 0002-Drop-detached_mounts_propagation-and-remove-sys-moun.patch:
its rationale (glibc sys/mount.h vs linux/mount.h clash reached via
src/vfs/utils.c) no longer applies because utils.c no longer includes
sys/mount.h, and the src/Makefile hunk no longer applied anyway.
* Add 0001-locktest-do-not-tie-struct-delegation-to-the-F_GETDE.patch:
glibc 2.43 defines F_GETDELEG/F_SETDELEG but neither glibc nor the
kernel UAPI declares struct delegation, so tying the struct to the
"#ifndef F_GETDELEG" guard leaves it undefined and locktest fails to
build with "variable has incomplete type 'struct delegation'".
Also bump the bundled unionmount-testsuite by one commit
(dir-weird-open-dir.py adjusted to the upstream kernel).
Signed-off-by: Khem Raj <khem.raj@oss.qualcomm.com>
Move from the v2.2.y stable branch to the v2.3.y line (tag v2.3.1,
2026-06-19). v2.3.x carries all v2.2.6..v2.2.16 backports plus the
2.3.0 rework.
Security:
* CVE-2026-55392: validate s_log_block_size in nilfs_sb_is_valid().
A crafted NILFS2 image with s_log_block_size > 6 caused oversized
shifts / huge allocations in nilfs-tune, dumpseg and any tool
reading a device.
2.3.0 highlights:
* Utilities (lscp, lssu, nilfs-clean, ...) accept file/directory
arguments and auto-resolve the backing device.
* cleanerd: cap segments per cleaning step by free segments (avoids
stalling near ENOSPC) and against 32-bit size_t overflow; clean
SIGINT exit, SIGUSR1 state dump.
* Library refactor with an API/ABI break: obsolete APIs and internal
structs removed from public headers, new segment I/O / resize /
freeze-thaw APIs, types unified to uint32_t/uint64_t/bool.
* mkfs.nilfs2 sets the block-count read-only-compat flag by default.
Packaging-relevant build-system changes:
* Soname bump libnilfs/libnilfsgc 2 -> 3; libnilfscleaner is no longer
a shared library (now an internal convenience lib), so
libnilfscleaner.so.* correctly disappears from the runtime package.
* New pkg-config producer files (nilfs.pc, nilfsgc.pc) land in
${libdir}/pkgconfig -> require inherit pkgconfig.
* LT_INIT([disable-static]) means static libs are no longer installed.
* The old --with-base-sbindir knob was replaced upstream by an
--enable-usrmerge=(no|sbin|bin|auto) mechanism. Drop the local
0001-configure.ac-Add-knob-to-define-base-sbindir.patch and select
the hierarchy explicitly from DISTRO_FEATURES, because the "auto"
default probes the build host and is wrong when cross compiling.
LDCONFIG=true is still required.
Signed-off-by: Khem Raj <khem.raj@oss.qualcomm.com>
Advance the SRCREV by 14 commits to the current master tip
(474b3acb, committed 2026-06-29) and bump PV to that commit date per
the recipe's date-based versioning convention.
Notable changes in this range (Aleph One / Charles Manning):
* yaffs_guts.c: fix the size of the allocated buffer for unmanaged
buffers (real bugfix).
* Fully deinitialise the cache when unmounting (real bugfix).
* yaffs_guts.c: use AND instead of remainder for hash calculation.
* Move Linux attribute handling out of the core into
yaffs_linux_attribs.c (affects only the out-of-tree kernel module,
which this recipe does not build).
* Remove K&R function declarations, silence gcc warnings on the dtest
filename array, several benign typo/whitespace cleanups, and
power-fail stress-test reporting improvements.
Signed-off-by: Khem Raj <khem.raj@oss.qualcomm.com>
Two upstream stable point releases. The "2.4.99" that the automated
version checker reports is the OpenZFS development/META tag on master,
not a real release; 2.4.3 is the newest 2.4.x stable.
2.4.2:
* Linux mount API rewrite: a new fs_parse-based mount-parameter parser
replaces the legacy options parser and zpl_parse_monolithic; support
for the old mount API is removed and the build now refuses to build
without fs_context. vfs_t lifetime is tied to fs_context.
* Linux 7.0/7.1 compatibility: fs_context shims, explicit setlease
handler (also for directories), posix_acl_to_xattr() allocation,
blk_queue_nonrot()->blk_queue_rot(), removal of the minimum kernel
version check.
* dRAID fixes: checksum errors after rebuild with degraded disks,
import failure after disk replacement, data corruption after disk
clear, sequential-resilver reads from degraded vdevs.
* Data-integrity fixes: read corruption after block-clone-after-
truncate, range-tree corruption race in dnode_sync(), a
use-after-free in dmu_write_direct_done(), "kernel BUG at
mm/usercopy.c".
* New feature: POSIX_FADV_DONTNEED support.
* Build: SPDX tags + spdxcheck enforcement, "make dist" uses pax tar,
objtool taken from $kernelbuild.
2.4.3:
* Security/robustness hardening: enforce exact decompressed length for
lz4/gzip/zstd, validate size fields/strings and reject un-terminated
strings in packed nvlists, sharenfs invalid-character check, extended
zfs_ioc_set_prop() secpolicy, key lookup failures always return
EACCES, ZVOL encryption-key check for block cloning, and removal of
/etc/sudoers.d/zfs.
* Correctness: double free for blocks cloned after DDT prune, off-by-
one in the redaction handler dropping the last block, several
lock-ordering/lockdep fixes, FreeBSD cache_vop_rename panic fix.
* Linux: properly apply ro/rw mount option to the superblock, Linux 5.6
fs_parse API mismatch fix, spl_kvmalloc strips __GFP_COMP before
vmalloc().
* aarch64 build failure fixed by removing earlyclobber.
Supported kernel range is now 4.18 - 7.0.
Signed-off-by: Khem Raj <khem.raj@oss.qualcomm.com>
python3-black RDEPENDS on python3-pytokens, so extending black to native and
nativesdk only moves the "Nothing RPROVIDES" warning one level down unless
pytokens is extended as well.
pytokens has no runtime dependencies of its own and only needs
python3-mypy-native to build, so it is safe to extend.
Signed-off-by: Khem Raj <khem.raj@oss.qualcomm.com>
python3-pytest-examples is built for native and nativesdk and RDEPENDS on
python3-black, but black itself was never extended, so parsing universe warns:
WARNING: Nothing RPROVIDES 'nativesdk-python3-black' (but
virtual:nativesdk:...python3-pytest-examples_0.0.18.bb RDEPENDS on or
otherwise requires it)
WARNING: Nothing RPROVIDES 'python3-black-native' (but
virtual:native:...python3-pytest-examples_0.0.18.bb RDEPENDS on or
otherwise requires it)
and both python3-pytest-examples-native and nativesdk-python3-pytest-examples
end up unbuildable and get dropped.
Signed-off-by: Khem Raj <khem.raj@oss.qualcomm.com>
tests/cc passes -DCMAKE_CURRENT_BINARY_DIR="${CMAKE_CURRENT_BINARY_DIR}" to
every test object so the tests can find archive.zip, the debuginfo libraries
and dummy_proc_map.txt at runtime. That put 13 copies of the absolute build
directory into the installed test binary:
QA Issue: File /usr/lib/bcc/ptest/tests/cc/test_libbcc_no_libbpf in package
bcc-ptest contains reference to TMPDIR [buildpaths]
and, on autobuilders whose TMPDIR sits below $HOME, additionally:
QA Issue: ... contains a reference to the build host HOME directory.
The recipe worked around this by installing the assets to ${D}${B}/tests/cc,
i.e. by reproducing the build path inside the image, and by demoting
buildpaths to a warning.
Add a patch introducing a TEST_ASSET_DIR cache variable, defaulting to
CMAKE_CURRENT_BINARY_DIR so the upstream default is unchanged, and point it at
${PTEST_PATH}/tests/cc. The assets are now installed next to the test binaries
and no build path is embedded, so the buildpaths overrides can go away.
Also install dummy_proc_map.txt, which the tests reference but which was never
shipped.
Signed-off-by: Khem Raj <khem.raj@oss.qualcomm.com>
wxPython's pyproject.toml declares setuptools.build_meta, so do_check_backend
warns:
QA Issue: inherits setuptools3 but has pyproject.toml with
setuptools.build_meta, use the correct class [pep517-backend]
Switching to python_setuptools_build_meta is not possible though. The sdist
ships a build.py in its top level directory which shadows the "build" module
that pyproject-build imports itself, so do_compile dies before it even looks
at the recipe:
File ".../recipe-sysroot-native/usr/bin/pyproject-build", line 5, in <module>
from build.__main__ import entrypoint
ModuleNotFoundError: No module named 'build.__main__'; 'build' is not a package
Signed-off-by: Khem Raj <khem.raj@oss.qualcomm.com>
uefi-firmware-parser's pyproject.toml declares setuptools.build_meta as its
build backend, so setuptools3 is the wrong class and do_check_backend warns:
QA Issue: inherits setuptools3 but has pyproject.toml with
setuptools.build_meta, use the correct class [pep517-backend]
Unlike setup.py, pyproject-build honours the declared build requirements, so
the setuptools_scm dependency has to be provided explicitly, otherwise
do_compile fails with:
ERROR Unmet dependencies ...
setuptools-scm>=8.0
wanted: >=8.0
found: not installed
setup.py passes use_scm_version to setup(), which was silently discarded
before ("Unknown distribution option: 'use_scm_version'"), so the package used
to ship its metadata as uefi_firmware-0.0.0.dist-info. Pin the version to
${PV} so the directory is now correctly named uefi_firmware-1.16.dist-info.
Signed-off-by: Khem Raj <khem.raj@oss.qualcomm.com>
canopen's pyproject.toml declares setuptools.build_meta as its build backend,
so setuptools3 is the wrong class and do_check_backend warns:
QA Issue: inherits setuptools3 but has pyproject.toml with
setuptools.build_meta, use the correct class [pep517-backend]
Unlike setup.py, pyproject-build honours the declared build requirements, so
the setuptools_scm dependency has to be provided explicitly, otherwise
do_compile fails with:
ERROR Unmet dependencies ...
setuptools_scm>=8
wanted: >=8
found: not installed
setuptools_scm was silently ignored before, so the package used to ship its
metadata as canopen-0.0.0.dist-info. Pin the version to ${PV} so the
directory is now correctly named canopen-2.4.1.dist-info.
Signed-off-by: Khem Raj <khem.raj@oss.qualcomm.com>
thrift's pyproject.toml declares setuptools.build_meta as its build backend,
so setuptools3 is the wrong class and do_check_backend warns:
QA Issue: inherits setuptools3 but has pyproject.toml with
setuptools.build_meta, use the correct class [pep517-backend]
Signed-off-by: Khem Raj <khem.raj@oss.qualcomm.com>
Changes:
- Add floating panes: non-modal panes that sit above the tiled layout,
created with the new new-pane command (bound to * by default). Early,
currently limited release (move/resize via mouse only).
- Allow run-shell arguments after the shell command to be expanded as
#{1}, #{2}, and so on.
- Add -g to kill-session to kill all sessions in a session group.
- Add line numbers in copy mode via the new copy-mode-line-numbers option
(off/default/absolute/relative/hybrid) plus matching styles.
- Tighten read-only checks on attach-session, detach-client and
switch-client; various control-mode and paste-detection fixes.
Signed-off-by: Khem Raj <khem.raj@oss.qualcomm.com>
Changes across the 6.24.13 - 6.24.16 maintenance releases:
- Handle redirections in bracketed expressions, e.g. if { foo > /dev/null }.
- Recognize history searches in history expansion (!?foo<tab>).
- Fix !^:h and !^:t failing when no / is found.
- Add cross-build support for the gethost helper; add support for $?<.
- Fix issues with S-JIS handling; improve nls/Makefile.in compatibility hacks.
Makefile changed the gethost host-helper rule to use
CPPFLAGS_FOR_BUILD / CFLAGS_FOR_BUILD, which lack the -I. -I$(srcdir) include
paths, so the configure-generated config.h was invisible and do_compile failed.
Set EXTRA_OEMAKE CC_FOR_GETHOST and {CFLAGS,CPPFLAGS,LDFLAGS}_FOR_BUILD from
the BUILD_* vars and simplified do_compile:prepend to 'oe_runmake gethost'.
The *_FOR_BUILD vars are only consumed by the gethost rule, so the target
build is unaffected.
Signed-off-by: Khem Raj <khem.raj@oss.qualcomm.com>
Version-locked to oe-core's gstreamer1.0 1.28.4; this jump spans the 1.24,
1.26 and 1.28 stable cycles.
Notable GES changes by 1.28:
- GESPipeline manages a shared task pool (auto-sized to the number of
processors), with apps able to supply their own via context negotiation.
- New thread-safe controller API for control-point access, fixing
use-after-free issues in the old API.
- The OpenTimelineIO formatter moved to a standalone Python plugin in
gst-python.
- Framepositioner z-order is now exposed/controllable; new APIs for detecting
clip layer movement, a GES error domain, and improved timeline error
reporting.
License-Update: COPYING and COPYING.LIB
(6762ed442b3822387a51c92d928ead0d -> 69333daa044cb77e486cc36129f7a770).
Upstream replaced the stale "GNU LIBRARY General Public License, Version 2"
text with the correct "GNU LESSER General Public License, Version 2.1" file;
the declared LICENSE (GPL-2.0-or-later AND LGPL-2.1-or-later) is unchanged and
is now more accurately matched.
Signed-off-by: Khem Raj <khem.raj@oss.qualcomm.com>
Changes (2.10; 2.09 shipped no separate release notes):
- Recognize new languages and file types: Beluga, C3, CodeQL, DAML, Quxlang,
Rhai and Slang.
- Skip the contents of .venv directories.
- Add --by-percent support with the --fmt formatting styles.
- Support process substitution as input to --sum-reports (Unix-like systems).
- Treat wscript as a filename rather than an extension; assorted bug fixes.
Signed-off-by: Khem Raj <khem.raj@oss.qualcomm.com>
Changes within the 7.0 LTS branch:
- 7.0.28: raise the maximum supported MariaDB (12.3), MySQL (9.7) and
TimescaleDB (2.27) versions; update VMware alarm discovery; fix excessive
history upload delay for Zabbix agents (ZBX-27495); assorted frontend/agent
fixes.
- 7.0.29: raise the maximum supported TimescaleDB version to 2.28; fix a
potential VMware collector crash on vCenter connectivity loss (ZBX-27975);
fix RabbitMQ template duplicate discovery keys; fix trend-average producing
inf values (ZBX-27453); fix remote command execution for trigger actions
when the target host is monitored by a proxy group (ZBX-27665).
Signed-off-by: Khem Raj <khem.raj@oss.qualcomm.com>
Deliberately switch from the conservative 2.0.x branch to upstream's
recommended 2.1.x line (upstream describes 2.0.x as "provided for the more
conservative distros" and recommends 2.1.x for most users).
Notable changes across the 2.1.x arc:
- Remove the flashrom plugin (2.1.0).
- Add native CBOR parsing and drop the libcbor2 dependency (2.1.2).
- No longer depend on json-glib, libarchive or protobuf-c (2.1.2).
- Add a systemd-pcrlock plugin wired into UEFI updates; add MTD lock and TCG
disk-encryption security attributes and support for externally-managed EFI
signature lists; enable more plugins when compiling for Android.
- Security/robustness: require trusted metadata for device updates, disallow
force installs over D-Bus, and numerous segfault / out-of-bounds-read fixes.
2.1.x removed the cbor, libarchive, plugin_flashrom and protobuf
meson options. Removed plugin_flashrom and protobuf from the default
PACKAGECONFIG, deleted the now-dead PACKAGECONFIG for libarchive,
plugin_flashrom, protobuf
Signed-off-by: Khem Raj <khem.raj@oss.qualcomm.com>
Changes:
- imtcp/omfwd: add experimental TCP stream compression (zlib/zstd), distinct
from TLS-native compression.
- core: fix negated exact-priority filters such as local4.!=debug so they
match all documented priorities except the negated one.
- rainerscript: preserve embedded NULs at C-string boundaries; warn on
constant boolean AND/OR operands.
- imfile: deliver same-file monitors independently; RFC3164 parser honors
parser.parseHostnameAndTag at runtime; invalidate the cached programname
when a message tag changes.
- General hardening and testbench de-flaking across many commits.
8.2606.0 makes libyaml an explicit configure feature (previously
auto-detected) that is enabled by default.
Added PACKAGECONFIG for libyaml and keep it disabled
Signed-off-by: Khem Raj <khem.raj@oss.qualcomm.com>
Changes across the 0.7.x series:
- 0.7.0: build fonts reproducibly (source restructuring, hence the major
version bump); generate OTF by default instead of TTF; provide LaTeX
OpenType fonts for XeTeX.
- 0.7.1 - 0.7.3: further reproducibility fixes; make Purisa/Sawasdee
embeddable-as-installable in PDF; glyph-width and PANOSE fixes; normalize
OS/2 Typo metrics for consistent line spacing.
- 0.7.4: fix parallel-build failure and a Type1 reproducibility issue; add a
Stylistic Set 1 (ss01) lookup for descenderless YO YING and THO THAN;
balance Kinnari/Norasi cubic splines.
- The 'ttf' tarball moved upstream from .../software/ttf-tlwg/ to
.../software/fonts-tlwg/fonts/ (old path 404s for this version); SRC_URI
updated accordingly.
- 0.7.4 relocated the fontconfig .conf files into the source tarball at
fonts-tlwg-${PV}/fontconfig/, so do_install:append now iterates that path
(58 ttf + 14 conf files installed).
- Updated LIC_FILES_CHKSUM md5 for COPYING: upstream additively added a
"LaTeX Supports" LPPL section (merged fonts-tlwg LaTeX package); no existing
license text removed, declared LICENSE unchanged.
Signed-off-by: Khem Raj <khem.raj@oss.qualcomm.com>
Changes:
- Add dnf and dnf_group package modules.
- Refactor cfe_autorun_inventory_packages; use a Python script to check the
Linux distribution version.
- Add cryptsetup to the stdlib paths.
- Add full-hosts-delete and 2FA support to the distributed cleanup script.
- Numerous lint/format cleanups, a fixed pip packages-promise regex, and
CHANGELOG formatting improvements.
Release announcement:
https://cfengine.com/blog/2026/cfengine-3-28-released-introspection/
Signed-off-by: Khem Raj <khem.raj@oss.qualcomm.com>
Changes (spanning the 1.19.0 and 1.19.1 releases):
- Add a Path iterator (Cairo::Path::Element / const_iterator) and matching
path-iter example.
- Add XCB device and XCB surface APIs, and an Xlib device API.
- Surface gains create_similar_image(), supports_mime_type(), map_to_image()
and the MappedImageSurface class.
- Add the mesh pattern API.
- 1.19.1: mark Cairo::Path::Element/const_iterator for export in the Windows
build, fix a memory leak in Context::pop_group(), plus build/doc fixes.
Signed-off-by: Khem Raj <khem.raj@oss.qualcomm.com>
Changes:
- New --monitor option to run a command and watch its input, output, or both.
- Allow decimal values for --rate-limit; extend the %T / %{buffer-percent}
format to indicate the transfer method when the buffer is not being used.
- Performance: use splice() in more circumstances, use copy_file_range(2)
when input and output are both regular files, and speed up --discard by
splicing to /dev/null internally.
- Fixes: report signal interrupts on a new line (no "^C" over the progress),
prevent an endless wait on unread output-pipe data, avoid double-EOF on
spliced terminal input, and error out when the --pidfile cannot be replaced.
Signed-off-by: Khem Raj <khem.raj@oss.qualcomm.com>
Changes:
- core: remove the deprecated DCT and JPX decoders; GPG-based signature
improvements; fix a signature regression (KDE bug #520496); fix added
annotations getting lost in some cases (#1732); fix crashes in malformed
documents.
- utils: pdfinfo now sanitizes its output.
- cpp: add render_hint::ignore_paper_color to allow a transparent paper color.
- glib: construct PopplerPage::mutex, fixing a crash on macOS (#1743).
- qt5/qt6: fix a crash in some signature operations (#1737).
Signed-off-by: Khem Raj <khem.raj@oss.qualcomm.com>
Changes:
- Fix a bug (introduced in 2.5.0) with multi-slots readers.
- Add support for the Haiku operating system.
Signed-off-by: Khem Raj <khem.raj@oss.qualcomm.com>
Changes:
- Add support for the --unreachable-script callback to filter unreachable
branch and MC/DC expressions.
- Update llvm2lcov to support llvm/22.
- Generate man and HTML documentation from RST sources.
- Various bug fixes and test-coverage improvements.
lcov 2.5's install target now generates its man pages (and HTML)
from RST via sphinx-build instead of shipping pre-built man pages as 2.4 did.
Added DEPENDS += python3-sphinx-native python3-sphinx-rtd-theme-native to
preserve the man-page packaging behavior (the install target hard-codes
"man html" and docs/conf.py uses the sphinx_rtd_theme).
Signed-off-by: Khem Raj <khem.raj@oss.qualcomm.com>
Changes:
- 1.8.1: correctly close the slots of a multi-slots reader; fix three minor
issues found by AISLE in partnership with Red Hat; other minor improvements.
- 1.8.2: fix initialisation of composite devices (such as Yubico tokens) and
correctly close the slots of a serial multi-slots reader.
Signed-off-by: Khem Raj <khem.raj@oss.qualcomm.com>
Changes:
- Allow sorting the process list on any column via the new 'O' key
(case-insensitive column name, ascending/descending); the 'A' automatic
sort on the most-utilized resource is dropped.
- Show oom_score and oom_score_adj per process, and a system-wide available
(reclaimable) memory breakdown per category in bar-graph mode.
- Improved offline/online and hotpluggable CPU detection; maximum number of
CPUs raised from 2048 to 4096.
- Show full user and group names with the -i flag; support longer pod and
container names in parsable/json output.
- atopgpud now also gathers graphical-use NVIDIA GPU stats, can run without
root privileges, and uses UNIX domain sockets instead of TCP sockets.
- Raw file format is incompatible with previous versions (convert with
atopconvert).
Signed-off-by: Khem Raj <khem.raj@oss.qualcomm.com>
oe-core carries a backport of libarchive PR #2846 (see
0003-Convert-RPM-reader-into-a-proper-format-supporting-b.patch, added
for RPM 6 support) which converts the RPM reader from a filter into a
proper format. That drops archive_read_support_filter_rpm() from the
public API in favour of archive_read_support_format_rpm(), and
fuse-archive fails to compile:
lib/reader.cc:509:9: error: use of undeclared identifier
archive_read_support_filter_rpm
Add a patch selecting the available entry point at compile time.
ARCHIVE_FORMAT_RPM is defined by archive.h only in versions providing
the new format, so it is an exact feature test and the recipe keeps
building against both stock and patched libarchive.
Signed-off-by: Khem Raj <khem.raj@oss.qualcomm.com>
We now have Rust 1.71+ version in oe-core so we can build ruff,
add it to the ptest runtime dependencies, install docs and fix
paths in test_docs.py.
There is a known jemallocator issue on aarch64 that affects ruff
and some of the pydantic tests from test_docs.py:
E <jemalloc>: Unsupported system page size E memory allocation of 4 bytes failed
More details about JEMALLOC_SYS_WITH_LG_PAGE:
https://github.com/gnzlbg/jemallocator/issues/170#issuecomment-1503228963
This work was sponsored by GOVCERT.LU.
Signed-off-by: Leon Anavi <leon.anavi@konsulko.com>
Signed-off-by: Khem Raj <khem.raj@oss.qualcomm.com>
Add a recipe for devtools, Python's missing debug print command and
more. Release 0.12.2:
- install debug into builtins via DebugProxy
From release 0.12.1:
- Fix docs publish.
From release 0.12.0:
- build docs on CI
- Update usage to reflect the recent addition of the pytest plugin
- Support dataclasses with slots
This work was sponsored by GOVCERT.LU.
Signed-off-by: Leon Anavi <leon.anavi@konsulko.com>
Signed-off-by: Khem Raj <khem.raj@oss.qualcomm.com>
Add a recipe for pytest-examples, Pytest plugin for testing examples
in docstrings and markdown files. Release v0.0.18:
- Adapts tests to pytest 8.3.4
- Increases triple quote search
This work was sponsored by GOVCERT.LU.
Signed-off-by: Leon Anavi <leon.anavi@konsulko.com>
Signed-off-by: Khem Raj <khem.raj@oss.qualcomm.com>
Ruff is an extremely fast Python linter and code formatter, written
in Rust. Add a recipe for release 0.16.0:
- Ruff now enables a much larger set of rules by default (413, up
from 59). See the blog post for more details and the new Default
Rules page for a full listing of the enabled rules.
- Ruff can now format Python code blocks in Markdown files and will
do this by default. See the documentation for more details.
- Ruff now supports ruff: ignore comments at the ends of lines,
like noqa comments, or on the line preceding a diagnostic. For
example, these both suppress an unused-import (F401) diagnostic.
- Fixes are now shown in check and format --check output.
- format --check now supports the same output formats as the linter,
including the github and gitlab outputs for rendering annotations
in CI
- The filename, location, end_location, fix.edits[].location, and
fix.edits[].end_location fields in the JSON output format may now
be null rather than defaulting to the empty string and row 1,
column 1, respectively.
This work was sponsored by GOVCERT.LU.
Signed-off-by: Leon Anavi <leon.anavi@konsulko.com>
Signed-off-by: Khem Raj <khem.raj@oss.qualcomm.com>
uutils-coreutils sets PROVIDES/RPROVIDES = "coreutils" and is a drop-in
replacement that cannot coexist with GNU coreutils; both ship the same
update-alternatives links. meta-oe/conf/layer.conf already selects between
them with PREFERRED_PROVIDER_coreutils{,-native,nativesdk-}, defaulting to
GNU coreutils.
PREFERRED_PROVIDER only disambiguates providers reached through a
dependency, not recipes built as explicit targets. Since uutils-coreutils
was in world, 'bitbake world' and 'bitbake universe' pulled both recipes
into the same task graph as targets, so bitbake reported:
NOTE: Multiple .bb files are due to be built which each provide coreutils:
NOTE: Multiple .bb files are due to be built which each provide coreutils-native:
NOTE: Multiple .bb files are due to be built which each provide nativesdk-coreutils:
Key EXCLUDE_FROM_WORLD off the selected provider so only the chosen
implementation is built in world/universe. Building uutils-coreutils
explicitly, or selecting it via PREFERRED_PROVIDER_coreutils, still works.
This follows the existing conditional EXCLUDE_FROM_WORLD idiom used by
mpv, sox and vlc.
Signed-off-by: Khem Raj <khem.raj@oss.qualcomm.com>
python3-py-serializable is extended to nativesdk and RDEPENDS on
python3-defusedxml, but defusedxml itself was only extended to native, so
the nativesdk variant of the dependency did not exist:
WARNING: Nothing RPROVIDES 'nativesdk-python3-defusedxml' (but virtual:nativesdk:...python3-py-serializable_2.1.0.bb RDEPENDS on or otherwise requires it)
This also made nativesdk-python3-cyclonedx-python-lib unbuildable, since it
RDEPENDS on python3-py-serializable.
nativesdk-python3-xml, the recipe's own runtime dependency, is provided by
nativesdk-python3, so the nativesdk variant resolves cleanly.
Signed-off-by: Khem Raj <khem.raj@oss.qualcomm.com>
hwloc is only a build dependency for class-target, but libhwloc was an
unconditional RDEPENDS. native.bbclass rewrites runtime dependencies via
native_filter(), which suffixes any unsuffixed dependency, so libhwloc
became libhwloc-native for tbb-native and nativesdk-libhwloc for
nativesdk-tbb. The hwloc recipe sets no BBCLASSEXTEND, so neither variant
exists and nothing can provide them:
WARNING: Nothing RPROVIDES 'libhwloc-native' (but virtual:native:...tbb_2022.3.0.bb RDEPENDS on or otherwise requires it)
WARNING: Nothing RPROVIDES 'nativesdk-libhwloc' (but virtual:nativesdk:...tbb_2022.3.0.bb RDEPENDS on or otherwise requires it)
This cascaded to tbb-dev-native/tbb-native via the generic
RDEPENDS:${PN}-staticdev in bitbake.conf, and from there to opencv-native
through opencv's PACKAGECONFIG[tbb].
Gate the runtime dependency on class-target to match the existing DEPENDS
gating. Target builds keep libhwloc unchanged.
Signed-off-by: Khem Raj <khem.raj@oss.qualcomm.com>
Upstream is dead. The newest tools tarball on SourceForge is
uml_utilities_20040406 from 2004-04-08 [1], which is what this recipe
tracks, and the project's last artifact of any kind is a 2.4 kernel
patch from Nov 2004. The tarball still ships CVS metadata directories.
Keeping it building against a current toolchain has meant carrying six
portability patches (~450 lines) for missing headers, stat64 -> stat and
LDFLAGS handling, with a new one needed every couple of years.
Most of what it installs is no longer relevant to UML. The current
kernel UML HOWTO [2] documents only the vector transports and does not
mention uml_net or uml_switch at all. uml_net is also installed setuid
root and execvp()s bare "ifconfig"/"route" off $PATH from that setuid
context, which is not something we should be shipping.
tunctl is not lost: meta-networking has a standalone tunctl recipe, and
uml-utilities only registered it through update-alternatives at the
default priority anyway.
Nothing in any layer depends on it - the only reference was the
meta-oe-extended packagegroup, for build coverage.
Debian has since moved to a later 20070815.4 snapshot from a different
upstream tree, and even that was removed from testing in Nov 2025 with
open RC bugs [3]. Should anyone want uml_mconsole, uml_moo/uml_mkcow or
port-helper back, the right starting point is that snapshot rather than
this one.
[1] https://sourceforge.net/projects/user-mode-linux/files/tools/1/
[2] https://docs.kernel.org/virt/uml/user_mode_linux_howto_v2.html
[3] https://tracker.debian.org/pkg/uml-utilities
Signed-off-by: Khem Raj <khem.raj@oss.qualcomm.com>
Khem Raj
Wang Mingyu
Jason Schonberg
Khem Raj
Gianfranco Costamagna
Leon Anavi