GNU coreutils installs these in ${base_bindir}, so match it. This only
matters when usrmerge is not enabled, as otherwise ${base_bindir} and
${bindir} are the same directory and the generated links are unchanged.
Only the path of the update-alternatives symlinks changes (from ${bindir}
to ${base_bindir}); they still point to ${bindir}/coreutils.
Checked that the generated postinstall now links all six in /bin.
AI-Generated: Uses Claude Sonnet 5.5
Signed-off-by: Etienne Cordonnier <ecordonnier@snap.com>
Signed-off-by: Khem Raj <khem.raj@oss.qualcomm.com>
The previous SUMMARY could be read as suggesting that this recipe is the
official GNU coreutils project. Reword it to state that uutils is a
separate, Rust reimplementation.
Signed-off-by: Etienne Cordonnier <ecordonnier@snap.com>
Signed-off-by: Khem Raj <khem.raj@oss.qualcomm.com>
Same conflict as in the 2.66 series; 2.90.0 carries the fix. The API
version stays 2.68, so the pkg-config names are unchanged.
Depends on the glib-2.0 2.90.0 upgrade in oe-core.
AI-Generated: Uses Claude Code (Claude Opus 5)
Signed-off-by: Markus Volk <f_l_k@t-online.de>
Signed-off-by: Khem Raj <khem.raj@oss.qualcomm.com>
GEmblem and GDBus::ActionGroup are final types as of GLib 2.89.2, which
makes the generated headers of 2.66.9 conflict with the class structs
GLib declares:
error: conflicting declaration 'using GEmblemClass = struct _GEmblemClass'
2.66.10 stops deriving those types.
Depends on the glib-2.0 2.90.0 upgrade in oe-core.
AI-Generated: Uses Claude Code (Claude Opus 5)
Signed-off-by: Markus Volk <f_l_k@t-online.de>
Signed-off-by: Khem Raj <khem.raj@oss.qualcomm.com>
Without -Dlibc set, systemd's in-tree musl support was disabled and the
static libsystemd.a build failed on musl (missing printf.h).
Signed-off-by: Greg Breen <gregbree@amazon.com>
Signed-off-by: Khem Raj <khem.raj@oss.qualcomm.com>
coredump:coredump_socket_test streams cores of a 128-thread process
(~1GB of thread stacks) into a tmpfs sized at half of RAM and fails
with ENOSPC with 1GB of RAM; it passes with 4GB.
run_kselftest.sh -s prints nothing while a test runs, and some
kselftests run for a long time (rseq:run_param_test.sh ~700s,
net/mptcp allows up to 1800s), so ptest-runner's default 450s
no-output timeout kills the whole suite. Raise PTEST_RUNNER_TIMEOUT
to 2400s for the kernel-selftest image.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
Signed-off-by: Khem Raj <khem.raj@oss.qualcomm.com>
do_compile fails with OpenSSL 4.0:
lm-ssl-openssl.c:281:92: error: invalid use of incomplete typedef
'ASN1_IA5STRING' {aka 'struct asn1_string_st'}
This break is a three line change to the ASN1_STRING accessors, but
upstream is dead. The last release is 1.5.4 from January 2021 and the
last commit in the mcabber fork is from June 2022, so every future
toolchain or dependency change is ours to carry.
Nothing in any layer depends on it, the only reference was the meta-oe
packagegroup.
Signed-off-by: Alper Ak <alperyasinak1@gmail.com>
Signed-off-by: Khem Raj <khem.raj@oss.qualcomm.com>
hackbench is a part of the rt-tests suite which has the recipe in
OE-Core since prehistoric times. Revert the commit 1e0d296bda
("hackbench: add Linux scheduler benchmark recipe") which duplicated a
part of OE-Core recipe.
Cc: Vaibhav Jindal <vaibjind@qti.qualcomm.com>
Signed-off-by: Dmitry Baryshkov <dmitry.baryshkov@oss.qualcomm.com>
Signed-off-by: Khem Raj <khem.raj@oss.qualcomm.com>
cyclictest is a part of the rt-tests suite which has the recipe in
OE-Core since prehistoric times. Revert the commit 1253fb8431
("cyclictest: add real-time latency measurement recipe") which
duplicated a part of OE-Core recipe.
Cc: Vaibhav Jindal <vaibjind@qti.qualcomm.com>
Signed-off-by: Dmitry Baryshkov <dmitry.baryshkov@oss.qualcomm.com>
Signed-off-by: Khem Raj <khem.raj@oss.qualcomm.com>
Minor release. MZ_PPMD (new default-on feature since 4.2.x) git-clones
ip7z/7zip over the network at configure time via CMake FetchContent;
disabled it, there's no network access in the build sandbox and nothing in
this recipe needs it. do_install:append's post-install sed also needed
retargeting: 4.2.x renamed the generated cmake config file from
minizip.cmake to minizip-config.cmake.
AI-Generated: Uses Claude Code
Signed-off-by: Khem Raj <khem.raj@oss.qualcomm.com>
Point release. LIC_FILES_CHKSUM md5 updated for COPYING and COPYING.tools (upstream re-encoding, license text unchanged).
AI-Generated: Uses Claude Code
Signed-off-by: Khem Raj <khem.raj@oss.qualcomm.com>
Per upstream NEWS, 0.2.0 is a small additive release on the GtDBusQueue
mock-D-Bus harness: new functions to emit a signal from the mocked service
and to mock activatable bus names (!14), a switch to POSIX file functions
instead of GIO when writing activatable service files (!15), and
gt_dbus_queue_assert_pop_message() now accepts empty varargs (!16). The
remainder is build/CI work - gtkdoc-check now runs as part of the test suite
(!7) and the docs CI build and deployment were fixed (!13). The public API is
additive only, so no consumer changes were needed; the recipe just needed the
new tag SRCREV and PV.
AI-Generated: Uses Claude Code
Signed-off-by: Khem Raj <khem.raj@oss.qualcomm.com>
Upstream reworked the build plumbing in this release: most of the kconfig
machinery moved out of kconfig/Makefile into scripts/genconfig.sh and
scripts/kconfig.c, and the defconfig/allyesconfig/allnoconfig/randconfig
targets are now thin wrappers around genconfig.sh. As part of that
cleanup (commit 228cb983) the rules that build the host-side kconfig
frontends lost their $(HOSTCC) and were hardcoded to "cc", which breaks
cross builds where only an explicit HOSTCC is exported and no "cc" is on
PATH. Add a patch restoring $(HOSTCC) there; the rest of the recipe needs
no changes.
AI-Generated: Uses Claude Code
Signed-off-by: Khem Raj <khem.raj@oss.qualcomm.com>
Add a recipe for cyclictest from the rt-tests project, a
real-time latency measurement tool used to evaluate Linux
kernel scheduler latency and jitter under various workloads.
The recipe builds and installs the cyclictest executable with
NUMA support enabled for use in real-time performance
evaluation, latency analysis, and regression testing.
Signed-off-by: Vaibhav Jindal <vaibjind@qti.qualcomm.com>
Signed-off-by: Khem Raj <khem.raj@oss.qualcomm.com>
Add a recipe for hackbench, a Linux scheduler benchmark used
to measure scheduling and inter-process communication
performance under various workloads.
The recipe builds and installs the hackbench executable with
NUMA support enabled for use in performance evaluation and
regression testing.
Signed-off-by: Vaibhav Jindal <vaibjind@qti.qualcomm.com>
Signed-off-by: Khem Raj <khem.raj@oss.qualcomm.com>
Add a recipe for UnixBench, a benchmark suite used to measure
CPU, memory, and overall system performance.
The recipe has been successfully built and validated on
aarch64 targets. Functional testing confirmed that the
benchmark suite executes correctly and generates results.
Signed-off-by: Vaibhav Jindal <vaibjind@qti.qualcomm.com>
Signed-off-by: Khem Raj <khem.raj@oss.qualcomm.com>
Add recipes for a set of benchmark and performance evaluation
tools commonly used to measure CPU, memory, network, and web
server performance.
- coremark-pro provides a comprehensive benchmark suite from
EEMBC that evaluates processor performance using integer,
floating-point, and real-world workloads.
- ramspeed-smp measures memory bandwidth and latency across
multiple CPU cores.
- sockperf provides network latency and throughput benchmarking
for TCP and UDP workloads.
- wrk2 generates HTTP workloads with constant throughput and is
useful for web server and service performance testing.
The recipes have been successfully built and validated on
aarch64 targets. Functional testing confirmed that the
benchmark binaries execute correctly and produce expected
results.
Signed-off-by: Vaibhav Jindal <vaibjind@qti.qualcomm.com>
Signed-off-by: Khem Raj <khem.raj@oss.qualcomm.com>
This will align it to oe-core systemd recipe.
Download mirror will contain only one repository copy, not two.
Signed-off-by: Peter Marko <peter.marko@siemens.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>
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>
Replace the long-dead anthy 9100h (last released 2009, fetched from an old
Ubuntu source tarball) with the maintained fork anthy-unicode, which is UTF-8
native and actively released on GitHub.
anthy-unicode renames its libraries (libanthy-unicode, libanthydic-unicode,
libanthyinput-unicode), pkg-config module file (anthy-unicode.pc), headers
(includedir/anthy-unicode-1.0/anthy), config file (anthy-unicode.conf) and
binaries (anthy-*-unicode). The recipe and packaging are updated accordingly:
libanthy-unicode0 for the runtime libraries and the standard PN-dev for the
development files.
The cross-compilation strategy is carried over from the old recipe: dictionary
generation runs helper programs (mkfiledic, mkworddic, mkdepgraph, calctrans,
proccorpus) that cannot execute when cross compiling. native-helpers.patch
promotes them to bin_PROGRAMS so anthy-unicode-native stages them into the
native sysroot; target-helpers.patch makes the target dictionary rules call
them from PATH instead of ./. not_build_elc.patch avoids byte-compiling the
Emacs lisp (no target Emacs under cross compilation). The patches are
regenerated against anthy-unicode's -unicode-suffixed libtool archives. The
legacy 2ch_t.patch is dropped; anthy-unicode has no mkanthydic/2ch.t.
uim is switched from the legacy EUC-JP anthy backend to the UTF-8
anthy-unicode backend it already supports: configure with --without-anthy
--with-anthy-utf8, DEPENDS on anthy-unicode, and the uim-anthy package is
renamed uim-anthy-utf8 (plugin libuim-anthy-utf8, anthy-utf8*.scm, module
registered as anthy-utf8) with RDEPENDS on libanthy-unicode0.
packagegroup-meta-oe: anthy -> anthy-unicode.
Verified by building anthy-unicode-native and anthy-unicode for
x86-64 (do_package_qa clean): the native helpers stage into the native
sysroot, the target build generates anthy.wdic/anthy.dep/anthy.dic via those
native helpers, and packaging splits into anthy-unicode, libanthy-unicode0 and
anthy-unicode-dev. uim variables verified by parse (it requires the x11
DISTRO_FEATURE, which the qemux86-64 yoe distro strips, so it is not buildable
in this configuration).
Signed-off-by: Khem Raj <khem.raj@oss.qualcomm.com>
proxy-libintl provides a Win32 static libintl stub sourced from an
ancient (2010) GNOME win32 dependency zip. It is not used by any target
build and its lone consumer was the meta-oe packagegroup; drop the
recipe and its packagegroup entry.
Signed-off-by: Khem Raj <khem.raj@oss.qualcomm.com>
dbus-broker's meson build installs its test and benchmark binaries to a
hardcoded path derived from the prefix:
conf.set('testdir', get_option('prefix') / 'lib/dbus-broker/tests')
i.e. always ${prefix}/lib/dbus-broker/tests, regardless of ${baselib}.
In OE terms that is ${nonarch_libdir}, not ${libdir}.
The recipe however collects them with ${libdir}, which only matches on
targets where baselib is "lib". On lib64 targets and for multilib
variants (e.g. lib32-dbus-broker) ${libdir} differs from the hardcoded
install path, so the installed tests are not captured by any package and
do_package fails with:
QA Issue: dbus-broker: Files/directories were installed but not
shipped in any package:
/usr/lib/dbus-broker/tests/unit/test-...
/usr/lib/dbus-broker/tests/dbus/test-...
[installed-vs-shipped]
Use ${nonarch_libdir} so the FILES glob always matches meson's install
location.
Signed-off-by: Khem Raj <raj.khem@gmail.com>
Suppress warnings from -Wfree-nonheap-object check. The warning occurs during
std::vector cleanup when deallocating memory with offset pointers for
test "compare_pointers" only on 32 bit platforms.
/recipe-sysroot/usr/include/c++/16.1.0/bits/new_allocator.h:183:66: error: 'void operator delete(void*, std::size_t)' called on pointer '__result' with nonzero offset 24 [-Werror=free-nonheap-object]
183 | _GLIBCXX_OPERATOR_DELETE(_GLIBCXX_SIZED_DEALLOC(__p, __n));
| ^
/test_conformance/generic_address_space/basic_tests.cpp: In function 'int test_compare_pointers(cl_device_id, cl_context, cl_command_queue, int)':
/test_common/harness/testHarness.h:151:9: note: declared here
151 | int test_##name(cl_device_id device, cl_context context, \
| ^~~~~
/test_common/harness/testHarness.h:154:29: note: in expansion of macro 'REGISTER_TEST_VERSION'
154 | #define REGISTER_TEST(name) REGISTER_TEST_VERSION(name, Version(1, 2))
| ^~~~~~~~~~~~~~~~~~~~~
/test_conformance/generic_address_space/basic_tests.cpp:776:1: note: in expansion of macro 'REGISTER_TEST'
776 | REGISTER_TEST(compare_pointers)
| ^~~~~~~~~~~~~
Signed-off-by: Antonios Christidis <a-christidis@ti.com>
Signed-off-by: Khem Raj <khem.raj@oss.qualcomm.com>
This file hardcodes a specific compiler for aarch64-linux builds, which
is not the compiler that we provide. As it's otherwise useless, we can
just delete the file.
Signed-off-by: Ross Burton <ross.burton@arm.com>
Signed-off-by: Khem Raj <khem.raj@oss.qualcomm.com>
The recipe was suppressing errors by including
'--compile-no-warning-as-error' and removing '-Werror=format-security'
from SECURITY_STRINGFORMAT. Remove both workarounds and carry upstream
patches that fix the underlying warnings instead.
Signed-off-by: Antonios Christidis <a-christidis@ti.com>
Signed-off-by: Khem Raj <khem.raj@oss.qualcomm.com>
Update OpenCL CTS to the latest release. With the move of the
opencl-headers and other opencl-* packages to a newer version, this one
also needs to update to fix compile issues related to the header
changes.
Signed-off-by: Ryan Eatmon <reatmon@ti.com>
Signed-off-by: Khem Raj <khem.raj@oss.qualcomm.com>
0001-test-Fix-error-factory-test-for-musl-libc-compatibil.patch
refreshed for 2.3.1
Changelog:
=============
- Add coroutine support for client-side async calls
- Add support for dumping variant to string
- Introduce clang-tidy for static analysis and fix issues
- Add deduction guides for Struct from std::tuple
- Other fixes and improvements
- Fix ABI compat issue by moving new virtual functions to the end of the class
Signed-off-by: Wang Mingyu <wangmy@fujitsu.com>
Signed-off-by: Khem Raj <khem.raj@oss.qualcomm.com>
Etienne Cordonnier
Igor Opaniuk
Markus Volk
Jaipaul Cheernam
Greg Breen
Khem RajandClaude Opus 5
Alper Ak
Dmitry Baryshkov
Vaibhav Jindal
Peter Marko
Khem Raj
Antonios Christidis
Wang Mingyu
Ross Burton
Zheng Ruoqin
Alexander Kanavin
Thomas Noack
Ryan Eatmon
Adam Duskett
Frede Hoey Braendstrup