On Fri, 2026-03-06 at 19:42 +0100, Yoann Congal via lists.yoctoproject.org wrote: > Following recent discussions regarding merging ptests in stable branches, we > are clarifying the policy for backporting ptests and selftests to stable > branches. > > (I'm transcribing the Yocto project TSC decision) > > Moving forward, we will apply the following criteria: > * Backporting new tests is acceptable only when they accompany a > corresponding bugfix. > * No standalone tests: If a test case is submitted on its own without > fixing a specific regression, it will be rejected by default. > > The rationale is twofold: > 1. New tests on stable branches add limited value if they aren't targeting a > known, occurring issue. > 2. AB-INT: We cannot afford additional intermittent failures. We need to > minimize the risk of introducing noise into the build results.
Just to add, the TSC realises there is a balance here. We do want to encourage new additional tests and improve our testing over time. Allowing backporting of ptests would encourage that. However, it is also clear we as a project are struggling with the existing set of tests and "intermittent" issues on the autobuilder. It is near impossible to know which of the new tests are going to cause such issues. It is also unclear how much testing many of the submitted and then "instantly backported" tests have against master. For those reasons the TSC is being cautious. If the autobuilder were in a better state, we would perhaps think differently but it isn't. It is still possible to request exceptions if there is a clear justification for adding specific tests to the stable series. Cheers, Richard
-=-=-=-=-=-=-=-=-=-=-=- Links: You receive all messages sent to this group. View/Reply Online (#232610): https://lists.openembedded.org/g/openembedded-core/message/232610 Mute This Topic: https://lists.openembedded.org/mt/118186596/21656 Group Owner: [email protected] Unsubscribe: https://lists.openembedded.org/g/openembedded-core/unsub [[email protected]] -=-=-=-=-=-=-=-=-=-=-=-
