Eric Anholt <e...@anholt.net> writes: > [ Unknown signature status ] > I pushed the branch enabling freedreno CI today. Now all the MRs will > get pre-merge testing with GLES3.1 on A630, and GLES2 on A307. > > See .gitlab-ci/README.md for more info on how it works. Also, as a > reminder, if you want to get something tested before generating your MR, > you can either edit .gitlab-ci.yml to change .ci-run-policy, or push the > branch to your repo with a name starting with "ci-". Secondary note: If > you make an MR from a branch starting with "ci-", it will have two > pipelines run every time you update it :( > > Today's surprises: > > - layout_binding.ssbo.fragment_binding_array is unstable > > I did one test to see if a patch I had on hand would fix it, and it > didn't, so I pushed through a patch disabling that test from CI so you > all wouldn't see those failures. > > - EXT4 FS corruption > > A few unfortunate events came together for this one: I had the kernel > boot args mounting / rw out of habit from back in my NFS-root days, and > there's no initrd to fsck /, and also I have harshly power-cycled them > all in the process of setting things up. This adds up to "there's minor > corruption on / that may impact test runs." Watch out for mysterious > docker errors. If this occurs at a bothersome rate, feel free to push a > disable of freedreno CI until I can fix it, though it doesn't look too > bad yet. I'll be fixing those boot args and making sure that we've got > clean filesystems on all boards tomorrow.
I've now enabled fsck on all the boards and we should be back in business. Looks like I can do this kind of change (bringing each board down in turn) without interrupting overall CI availability, which seems like a pretty good test.
signature.asc
Description: PGP signature
_______________________________________________ mesa-dev mailing list mesa-dev@lists.freedesktop.org https://lists.freedesktop.org/mailman/listinfo/mesa-dev