Le 04/11/2024 à 12:25, Yoann Congal a écrit : > Hi, > > > Le 29/10/2024 à 11:09, Yoann Congal a écrit : >> From: Yoann Congal <yoann.con...@smile.fr> >> >> The current reproducibility test stops all build tasks when a single >> task fails (default BitBake behavior). This means that a single build >> failure prevents the reproducibility of other packages from being >> tested, which is not ideal. >> >> To address this, we now use the --continue option of BitBake during test >> builds, allowing the build process to proceed even when some tasks fail. >> The failure cases are handled as gracefully as possible. >> >> In the event of a build failure: >> - The entire reproducibility test will be considered a failure. >> - The complete BitBake log will be saved in the "saved output" directory >> to facilitate debugging. On the autobuilder, this log should be >> available at https://autobuilder.yocto.io/pub/repro-fail/. >> - The last 20 lines of the log, which should show the failing recipes, >> will be displayed in the oe-selftest console. >> - If BitBake fails to create the deployment directory, it will be >> manually created to allow the comparison process to proceed. >> >> Here is what the output looks like when testing reproducibility of bash, >> a failing recipe (hello-fail) and a non-reproducible recipe >> (hello-norepro): >> >> <snip> >> >> => "reproducible.ReproducibleTests.test_reproducible_builds".status is >> correctly "FAILED" but the reproducibility of bash and hello-norepro is >> tested. >> >> Signed-off-by: Yoann Congal <yoann.con...@smile.fr> >> --- >> v2->v3: >> * Added a missing end of line in config concatenation >> >> v1->v2: >> * Upgrade some log from INFO to ERROR in the failing case to make it >> more visible in the logs. >> Richard Purdie review: >> * Simplify log handling >> --- >> meta/lib/oeqa/selftest/cases/reproducible.py | 56 ++++++++++++++++---- >> 1 file changed, 47 insertions(+), 9 deletions(-) > > That did not work in the https://valkyrie.yoctoproject.org/#/builders/87 > builder : > https://valkyrie.yoctoproject.org/#/builders/87/builds/17/steps/40/logs/stdio > Sorry about that :( > > I'll look into this as soon as possible but I would totally understand > if this patch is reverted in the meantime.
For what it's worth, with the fix [0], this worked as intended : in [1] (reproducibility test for meta-gnome), geary fails to build but reproducibility was still computed for other recipes. But, this is not over since geary is displayed in the "unused_exclusions" list (which is not wrong but not useful), a better list would be "excluded package that were identical between build A and build B". This list should serve as an argument to remove a package from the known non-reproducible lists. [0]: https://git.openembedded.org/openembedded-core/commit/?id=e89bbc00ba16574d719b199c01ffbf37646f4f54 [1]: https://valkyrie.yoctoproject.org/#/builders/87/builds/18/steps/16/logs/stdio -- Yoann Congal Smile ECS - Tech Expert
-=-=-=-=-=-=-=-=-=-=-=- Links: You receive all messages sent to this group. View/Reply Online (#206907): https://lists.openembedded.org/g/openembedded-core/message/206907 Mute This Topic: https://lists.openembedded.org/mt/109274319/21656 Group Owner: openembedded-core+ow...@lists.openembedded.org Unsubscribe: https://lists.openembedded.org/g/openembedded-core/unsub [arch...@mail-archive.com] -=-=-=-=-=-=-=-=-=-=-=-
