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

Issue 675712 link

Starred by 1 user

Issue metadata

Status: WontFix
Owner: ----
Closed: Jan 2017
Cc:
Components:
EstimatedDays: ----
NextAction: ----
OS: ----
Pri: 3
Type: Bug

Blocking:
issue 611579
issue 672278
issue 672280



Sign in to add a comment

Bisect - android_perf_bisect_builder failing with unknown target chrome_modern_public_apk

Project Member Reported by simonhatch@chromium.org, Dec 19 2016

Issue description

https://build.chromium.org/p/tryserver.chromium.perf/builders/android_perf_bisect_builder?numbuilds=200

Last 100 or so builds have failed with this error:

ninja: Entering directory `/b/build/slave/android_perf_bisect_builder/build/src/out/Release'
ninja: error: unknown target 'chrome_modern_public_apk'
step returned non-zero exit code: 1
 
Cc: jbudorick@chromium.org
+jbudorick

I checked a few of the failing builds, and they're from src@(~Oct/Sept) while the passing ones were from src@(~last week). My guess is that maybe chrome_modern_public_apk didn't exist as a target back then? We had a recent change to force the android perf builders to build specific targets, maybe we should just have them build everything? wdyt?
Blocking: 672278

Comment 3 by benhenry@google.com, Dec 19 2016

This is probably due to N SDK update. Anything older than that will not work anymore. 

Simon - have you seen any builds/bisects that failed with this same error that were testing over revision ranges during that time? We're going to have to wontfix them if so.
#1: Correct, chrome_modern_public_apk was added Nov 9: https://chromium.googlesource.com/chromium/src/+/be18d7c8226d8949d9b840a99b365e032697becf

#3: per the above, this is not due to the N SDK.
So should we just go ahead and build some generic "all" target and re-introduce these targets at a later date?
I guess so, though all will be significantly slower.
I don't know the exact number, but there seem to be a handful of regressions from pre-Nov 9 still being investigated, so maybe what we can do is:

- Fix this for now by just building all (that's what it used to do anyway)
- At some point in next few months switch back to building specific targets (or even get chromium_builder_perf working for android?), let perf sheriffs know this will happen and we won't be able to bisect older regressions, and WontFix the older regressions.

What does everyone think?
Blocking: 672280
benhenry: Can you help us figure out which regressions from before November 9 are most important to bisect?
Looks like we have 149 alerts that are still open. I'm going to narrow this down more to see which don't need bisect and report back.

https://bugs.chromium.org/p/chromium/issues/list?can=2&q=opened%3C2016-11-09+Performance%3DSheriff+Android&colspec=ID+Pri+M+Stars+ReleaseBlock+Component+Status+Owner+Summary+OS+Modified&x=m&y=releaseblock&cells=ids
Blocking: 611579
Status: WontFix (was: Untriaged)
Chatting offline w/ Annie/Dave, feeling here is that we can leave as is. If anybody pursues those we can revisit the "all" fix.
Components: Speed>Bisection

Sign in to add a comment