New issue
Advanced search Search tips
Note: Color blocks (like █ or █) mean that a user may not be available. Tooltip shows the reason.

Issue 865290 link

Starred by 1 user
Status: Assigned
Owner:
la...@chromium.org
Cc:
benchan@chromium.org
█ marc...@chromium.org
█ chadversary@chromium.org
chromeos-continuous-integration-fire-hose@google.com
Components:
EstimatedDays: ----
NextAction: ----
OS: Chrome
Pri: 2
Type: Bug



Sign in to add a comment

PreCQ isn't upreving arc-mesa, but CQ is

Project Member Reported by ihf@chromium.org, Jul 19 
https://chrome-internal-review.googlesource.com/c/chromeos/manifest-internal/+/642087

This change went through betty-pre-cq without failing
https://cros-goldeneye.corp.google.com/chromeos/healthmonitoring/buildDetails?buildbucketId=8940741373719917888

but surprisingly failed on betty-paladin
https://cros-goldeneye.corp.google.com/chromeos/healthmonitoring/buildDetails?builderName=betty-paladin&buildNumber=3484


It looks like the change was applied properly on paladin (leading to the 18.2 uprev + build failure)
[ebuild     U  ] media-libs/arc-mesa-18.2.0_pre-r129::chromiumos [17.3.0_rc5-r128::chromiumos] to /build/betty/ USE="android_gles30 cheets classic dri egl gallium gles1 gles2 llvm nptl shared-glapi -X -android_aep -android_gles2 -android_gles31 -android_gles32 -android_vulkan_compute_0 -cheets_user -cheets_user_64 -cros_host -debug -gbm -pic -profiling -selinux -vulkan -xlib-glx" ABI_X86="32 (-64) (-x32)" VIDEO_CARDS="llvmpipe virgl -amdgpu -freedreno -intel -mach64 -mga -nouveau -powervr -r128 -radeon -savage -sis -tdfx -via -vmware" 0 KiB


But there were problems with applying the patch on pre-cq that were not detected

08:43:54: INFO: RunCommand: git fetch -f https://chrome-internal-review.googlesource.com/chromeos/manifest-internal refs/changes/87/642087/4 in /b/swarming/w/ir/cache/cbuild/repository/manifest-internal
08:43:55: INFO: Attempting to create transaction for chadversary:*642087:*374373c2
08:43:56: INFO: Transaction for chadversary:*642087:*374373c2 is chadversary:*642087:*374373c2.
08:43:56: INFO: Attempting to apply change chadversary:*642087:*374373c2:"arc-mesa: Move to branch arc-18.2.0-pre1"
08:43:56: INFO: Applying via cherry-pick.
Trying simple merge.
Simple merge failed, trying Automatic merge.
Auto-merging external_full.xml
[patch_branch 1f02152] arc-mesa: Move to branch arc-18.2.0-pre1
 Author: Chad Versace <chadversary@chromium.org>
 Date: Tue Jun 19 00:44:26 2018 -0700
 1 file changed, 1 insertion(+), 1 deletion(-)


leading to building the old 17.3 package
[binary  N     ] media-libs/arc-mesa-17.3.0_rc5-r128::chromiumos to /build/betty/ USE="android_gles30 cheets classic dri egl gallium gles1 gles2 llvm nptl shared-glapi -X -android_aep -android_gles2 -android_gles31 -android_gles32 -android_vulkan_compute_0 -cheets_user -cheets_user_64 -cros_host -debug -gbm -pic -profiling -selinux -vulkan -xlib-glx" ABI_X86="32 (-64) (-x32)" VIDEO_CARDS="llvmpipe virgl -amdgpu -freedreno -intel -mach64 -mga -nouveau -powervr -r128 -radeon -savage -sis -tdfx -via -vmware" 4695 KiB

Comment 1 by ihf@chromium.org, Jul 22

Cc: benchan@chromium.org
Labels: -Pri-2 Pri-1
This breakage slipping through the pre-cq keeps repeatedly biting us (hitting the cq) in
https://chrome-internal-review.googlesource.com/c/chromeos/manifest-internal/+/642087

Comment 2 by jclinton@chromium.org, Jul 22

Owner: la...@chromium.org
Status: Assigned (was: Untriaged)
Assigning to next oncall.

Comment 3 by la...@chromium.org, Jul 23

Owner: dgarr...@chromium.org
Summary: PreCQ isn't upreving arc-mesa, but CQ is (was: builder did not detect that a patch was not applied correctly)
"Simple merge failed" doesn't mean the patch failed to apply, just that the "simple merge" strategy didn't work, which is common and fine. The "auto-merging" strategy succeeded immediately after.

Clearly though, something didn't work correctly. dgarrett@: Any idea why the ebuild uprev isn't being picked up in precq?

Comment 4 by dgarr...@chromium.org, Jul 23

Owner: la...@chromium.org
Testing manifest CLs currently requires very special support per-class of builder that only exists only for the CQ. Even there, it isn't very trustworthy.

It's ugly and painful, but it's a known issue.

Lann is currently putting together a plan for how to redo our sync code into a single consistent script that behaves the same everywhere. I forgot to mention manifest changes as a requirement for that plan, but they probably should be.

This is the best way to solve this issue going forward.

Comment 5 by pmalani@chromium.org, Jul 23

Cc: -pmalani@chromium.org

Comment 6 by jclinton@chromium.org, Sep 21

Labels: -Pri-1 Pri-2
►

Sign in to add a comment