It takes about a second to execute.
The test suite is covered by a different license (lgpl 2.1+) than the
main recipe (3-clause bsd).
Sample output:
root@qemux86-64:~# ptest-runner
START: ptest-runner
2025-12-22T19:34
BEGIN: /usr/lib/uriparser/ptest
[==========] Running 407 tests from 35 test suites.
[----------] Global test environment set-up.
[----------] 14 tests from CopyUriSuite
PASS: CopyUriSuite.ErrorSourceUriNull
PASS: CopyUriSuite.ErrorDestUriNull
[...many lines...]
PASS: SetUserInfo.UriWithoutHostNonNullRejected
[----------] 2 tests from VersionSuite
PASS: VersionSuite.EnsureVersionDefinesInSync
PASS: VersionSuite.EnsureRuntimeVersionAsExpected
[----------] Global test environment tear-down
[==========] 407 tests from 35 test suites ran.
[ PASSED ] 407 tests.
DURATION: 1
END: /usr/lib/uriparser/ptest
2025-12-22T19:34
STOP: ptest-runner
TOTAL: 1 FAIL: 0
Signed-off-by: Gyorgy Sarvari <skandigraun@gmail.com>
Signed-off-by: Khem Raj <raj.khem@gmail.com>
There are two different types of tags in glslang git repo. One is the
release tag of the project itself: 15.2.0, 14.3.0, etc. The other tag is
for Vulkan SDK: vulkan-sdk-1.4.309, vulkan-sdk-1.4.304.1, etc.
The vulkan sdk tag is used for glslang in openembedded-core because it
needs to update in locksetup with vulkan, which leads to a mismatch
between the runtime version and the build version. Set CHECK_VERSION_PV
for it to skip the version check.
Signed-off-by: Yi Zhao <yi.zhao@windriver.com>
Signed-off-by: Khem Raj <raj.khem@gmail.com>
bpftrace set the version by "git describe --dirty", since we have local
patch for bpftrace, '-dirty' will be added into the version, set
CHECK_VERSION_PV to mute the version mismatch warning
Signed-off-by: Changqing Li <changqing.li@windriver.com>
Signed-off-by: Khem Raj <raj.khem@gmail.com>
* flite --version return 1 block version output for
check-version-mismatch.bbclass
* even with version output flite-2.2-current, regular version match
regexp cannot match the version
so mute version mismatch warning for flite
Signed-off-by: Changqing Li <changqing.li@windriver.com>
Signed-off-by: Khem Raj <raj.khem@gmail.com>
It takes under 10 seconds to run the suite.
Executed succesfully on x86-64, with musl and glibc.
The recipe requires pam DISTRO_FEATURE to be present.
Signed-off-by: Gyorgy Sarvari <skandigraun@gmail.com>
Signed-off-by: Khem Raj <raj.khem@gmail.com>
It takes less than 30 seconds to execute.
Add a backported patch that fixes a broken test.
Signed-off-by: Gyorgy Sarvari <skandigraun@gmail.com>
Signed-off-by: Khem Raj <raj.khem@gmail.com>
The project has extensive runtime- and unittest suite - make use of it.
It takes around 10 seconds to run both suites.
Added two patches:
One is needed to convert python unittest output into automake output.
The other is required to be able to configure test data folder at
compile time - otherwise the tests are looking for the test-data
with absolute paths from the build machine.
Signed-off-by: Gyorgy Sarvari <skandigraun@gmail.com>
Signed-off-by: Khem Raj <raj.khem@gmail.com>
There is a timezone related ptest that fails using musl-libc.
This has been reported to the mariadb developers[1], who came up with
the backported patch that's the subject of this change.
This patch skips the timezone related tests with musl, in case the
testcase uses a timezone that behaves differently with musl than on
other platforms.
[1]: https://jira.mariadb.org/browse/MDEV-38029
Signed-off-by: Gyorgy Sarvari <skandigraun@gmail.com>
Signed-off-by: Khem Raj <raj.khem@gmail.com>
According to [1][2], generate phar.php during cross-compile can't be
done, but upstream test res of $(TEST_PHP_EXECUTABLE) is not suitable
for Yocto.
Explicitly set TEST_PHP_EXECUTABLE_RES = "1" to not generate phar.php
for target recipe
Drop 0005-sapi-cli-config.m4-fix-build-directory.patch which is obsolete
for generating phar.php
After apply this commit
...log.do_compile...
Generating phar.php
Skipping phar.php generating during cross compilation
Generating phar.phar
Skipping phar.phar generating during cross compilation
...log.do_compile...
Then php supports reproducible build
[1] https://github.com/php/php-src/issues/11099
[2] 93fa9613e1
Signed-off-by: Hongxu Jia <hongxu.jia@windriver.com>
Signed-off-by: Khem Raj <raj.khem@gmail.com>
When compressing docs, do not save the original file name and
timestamp by default (gzip -n). Make archives be reproducible
at each build
Signed-off-by: Hongxu Jia <hongxu.jia@windriver.com>
Signed-off-by: Khem Raj <raj.khem@gmail.com>
The tests are executing and passing fine using glibc.
It takes a bit more than 5 minutes to run them.
When using musl, the timezone tests fail due to the different ways
glibc and musl handle posix timezones. It have been reported
upstream: https://jira.mariadb.org/browse/MDEV-38029
Signed-off-by: Gyorgy Sarvari <skandigraun@gmail.com>
Signed-off-by: Khem Raj <raj.khem@gmail.com>
The logic used is :
- We check if the required tools are present or not
- We used magick to create an raw RGB file
- The created RGB is then converted to PNG using convert
- We re-gerenate RGB from PNG and compare the original and re-generated RGB
- Enabled the ptest in ptest-packagelists-meta-oe.inc as
suggested by Gyorgy Sarvari and incorporated logging suggestion
- This was done as standard imagemagick test like drawtest requires manual
internetion to verify the file
Signed-off-by: AshishKumar Mishra <ashishkumar.mishra@bmwtechworks.in>
Signed-off-by: Khem Raj <raj.khem@gmail.com>
The project is getting more mature, and it's now doable to do a yocto build using uutils-coreutils-native.
Signed-off-by: Etienne Cordonnier <ecordonnier@snap.com>
Signed-off-by: Khem Raj <raj.khem@gmail.com>
Contains fixes for CVE-2025-54764 and CVE-2025-59438
Also, add the recipe to the ptest image list, because it was missing.
Ptests passed successfully.
Changelog: https://github.com/Mbed-TLS/mbedtls/releases/tag/mbedtls-3.6.5
Signed-off-by: Gyorgy Sarvari <skandigraun@gmail.com>
Signed-off-by: Khem Raj <raj.khem@gmail.com>
The tests are supposed to work now - they were not moved to the correct
ptest list with the last update.
Signed-off-by: Gyorgy Sarvari <skandigraun@gmail.com>
Signed-off-by: Khem Raj <raj.khem@gmail.com>
Here is the setting in bpftool recipe:
COMPATIBLE_HOST = "(x86_64|aarch64|riscv64).*-linux"
bpftool only support 64 bit arch, so add it into NON_MULTILIB_RECIPES
to avoid expanding lib32- variant
Signed-off-by: Changqing Li <changqing.li@windriver.com>
Signed-off-by: Khem Raj <raj.khem@gmail.com>
[ snip of CMakeLists.txt ]
BISON_TARGET(constexp
${CMAKE_CURRENT_LIST_DIR}/constexp.y
${GENERATED_SRC}/ce_parse.cpp
COMPILE_FLAGS "${YACC_FLAGS}")
[ end of snip ]
bison embeds full paths in its generated headers, eg:
bison -o /full/build/path/generated/example.cpp ...
The header will include:
#ifndef YY_FULL_BUILD_PATH_GENERATED_MSCGEN_LANGUAGE_HPP_INCLUDED
#define YY_FULL_BUILD_PATH_GENERATED_MSCGEN_LANGUAGE_HPP_INCLUDED
This make doxygen-src not reproducible, fix by replacing full build path
Signed-off-by: Changqing Li <changqing.li@windriver.com>
Signed-off-by: Khem Raj <raj.khem@gmail.com>
The ocl-icd and opencl-icd-loader provide libOpenCL.so.1 rather than the
OpenCL Installable Client Driver (ICD). Rename the virtual package
accordingly.
Signed-off-by: Dmitry Baryshkov <dmitry.baryshkov@oss.qualcomm.com>
Signed-off-by: Khem Raj <raj.khem@gmail.com>
These two packages don't provide ICD drivers at runtime. They provide
the libOpenCL.so.1. Corresponding package is selected at the build time
via the PROVIDES_virtual/opencl-icd. Drop the useless and incorrect
RPROVIDES for virtual-opencl-icd.
Signed-off-by: Dmitry Baryshkov <dmitry.baryshkov@oss.qualcomm.com>
Signed-off-by: Khem Raj <raj.khem@gmail.com>
bpftrace need line up with kernel architecture type,
so add it to NON_MULTILIB_RECIPES.
bpftrace must be built with the same architecture (32-bit/64-bit) as the kernel
to ensure compatibility when tracing kernel functions and using eBPF features.
Signed-off-by: Sunil Dora <sunilkumar.dora@windriver.com>
Signed-off-by: Khem Raj <raj.khem@gmail.com>
It is needed for poco 1.14+
Ptest runs for less than a second.
Signed-off-by: Peter Marko <peter.marko@siemens.com>
Signed-off-by: Khem Raj <raj.khem@gmail.com>
The current include file that stores the known non-reproducible packages
is layer dependent and that forces the user of the layers to maintain
the list of the files (for example, see AB config[0]).
By moving the exclude list to each layer.conf and extending the common
OEQA_REPRODUCIBLE_EXCLUDED_PACKAGES variable, the known non-reproducible
packages will be automatically excluded for each layer used in the
reproducibility test without any special knowledge in the test
environment.
NB: the empty list for meta-initramfs was just removed not moved.
[0]: https://git.yoctoproject.org/yocto-autobuilder-helper/tree/config.json?id=7d8933e75bdf7fb821a25617cb2dcabf1f3f8700#n322
Suggested-by: Quentin Schulz <quentin.schulz@cherry.de>
Co-Developed-by: Guillaume Swaenepoel <guillaume.swaenepoel@smile.fr>
Signed-off-by: Guillaume Swaenepoel <guillaume.swaenepoel@smile.fr>
Signed-off-by: Yoann Congal <yoann.congal@smile.fr>
Signed-off-by: Khem Raj <raj.khem@gmail.com>
Refer [1], dialog's tarballs released with date, for convenient, PV also
end with date. Change to check without date suffix.
[1] https://invisible-mirror.net/archives/dialog/
Signed-off-by: Changqing Li <changqing.li@windriver.com>
Signed-off-by: Khem Raj <raj.khem@gmail.com>
jemalloc set the version by pass --with-version=${PV_LONG}, PV_LONG :=
"${PV}-171-g${SRCREV}", set match pattern for jemalloc to mute the
version mismatch warning
Signed-off-by: Changqing Li <changqing.li@windriver.com>
Signed-off-by: Khem Raj <raj.khem@gmail.com>
Refer [1], upstream set this version specially, so mute by skipp
checking
[1] 49057f6ce5
Signed-off-by: Changqing Li <changqing.li@windriver.com>
Signed-off-by: Khem Raj <raj.khem@gmail.com>
The version get by the version check tool is wrong, the output
include version of grep. Skip checking to mute the warning
Signed-off-by: Changqing Li <changqing.li@windriver.com>
Signed-off-by: Khem Raj <raj.khem@gmail.com>
ld.libhugetlbfs --version return the ld version on the host, and cause
version mismatch warning, mute the warning by skip checking for this
package
Signed-off-by: Changqing Li <changqing.li@windriver.com>
Signed-off-by: Khem Raj <raj.khem@gmail.com>
bpftool is compiled from kernel source, we hope it has version
aligned with kernel version, but the sources will printout some
other versions, skip this checking.
Signed-off-by: Changqing Li <changqing.li@windriver.com>
Signed-off-by: Khem Raj <raj.khem@gmail.com>
Fix reproducibility issue by retrieving the compiler version from the CC
env variable, which define the compiler used in the build and not from
the native gcc compiler install in the host machine.
Signed-off-by: Kéléfa Sané <kelefa.sane@smile.fr>
Signed-off-by: Khem Raj <raj.khem@gmail.com>
The reproducibility issue has been fixed by the commit "cjson: fix
buildpath warnings (0cc7315408)"
which replace the build path present in files:
- CTestTestfile.cmake
- tests/CTestTestfile.cmake
- fuzzing/CTestTestfile.cmake
by PTEST_PATH (/usr/lib/cjson/ptest).
Signed-off-by: Kéléfa Sané <kelefa.sane@smile.fr>
Reviewed-by: Yoann Congal <yoann.congal@smile.fr>
Signed-off-by: Khem Raj <raj.khem@gmail.com>
Last modification on boinc-client recipe (see commit "boinc-client: Fix
contains reference to TMPDIR [buildpaths] warning"
3e38af7e8b) fix the repoducibility issue,
with a do_install:append task which remove the build path present in the
header source file svn_version.h.
Signed-off-by: Kéléfa Sané <kelefa.sane@smile.fr>
Signed-off-by: Yoann Congal <yoann.congal@smile.fr>
Signed-off-by: Khem Raj <raj.khem@gmail.com>
Now check-version-mismatch.bbclass can check scripts' output
beside elf binaries output. We need to skip a few recipe.
These recipes' scripts give some version information, but they
are version information of scripts themselves instead of the package.
There's no executable in these packages that can give out useful
version information.
Signed-off-by: Chen Qi <Qi.Chen@windriver.com>
Signed-off-by: Khem Raj <raj.khem@gmail.com>
piglit is an open-source test suite for OpenGL implementations. It
contains a whole lot of programs which all do not give out any version
information. Running version check for piglit is a waste of time (~1h),
so setting a '%' to skip it.
Signed-off-by: Chen Qi <Qi.Chen@windriver.com>
Signed-off-by: Khem Raj <raj.khem@gmail.com>
* Current PV 3.4 is the kernel version when the recipe is newly added,
set PKGV to KERNEL_VERSION to make it always align with kernel version
* turbostat has its runtime version, set CHECK_VERSION_PV to this version to mute
the version mismatch warning
[snip of turbostat.c]
void print_version()
{
fprintf(outf, "turbostat version 2024.07.26 - Len Brown <lenb@kernel.org>\n");
}
[snip]
Signed-off-by: Changqing Li <changqing.li@windriver.com>
Signed-off-by: Khem Raj <raj.khem@gmail.com>
Runtime version of brctl is 1.7.1, but the configure.ac set version as
1.7 (AC_INIT([bridge-utils],[1.7])). And refer [1], this utility is
DEPRECATED, only critical bug fixes only, so it is nearly not possible
to correct from upstream, so mute this warning since it has no critical
influence.
[1] https://git.kernel.org/pub/scm/network/bridge/bridge-utils.git/commit/?h=main&id=ab8a2cc330253321be7bc69dea88bfaa3d48415e
Signed-off-by: Changqing Li <changqing.li@windriver.com>
Signed-off-by: Khem Raj <raj.khem@gmail.com>
For netperf, current PV is 2.7.0+git, and the SRCREV is newer than 2.7.0,
and there is no tag 2.7.1, the latest tag is in 2015, refer [1], so it
is not possile for us to make PV align with the runtime version below
by update SRCREV now, add CHECK_VERSION_PV for pn-netperf to mute the
version mismatch waring
$netserver -V
Netperf version 2.7.1
[1] https://github.com/HewlettPackard/netperf
Signed-off-by: Changqing Li <changqing.li@windriver.com>
Signed-off-by: Khem Raj <raj.khem@gmail.com>
