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

Issue 674202 link

Starred by 2 users

Issue metadata

Status: Available
Owner: ----
Cc:
Components:
EstimatedDays: ----
NextAction: ----
OS: ----
Pri: 3
Type: Bug



Sign in to add a comment

rebaseline-cl may not de-dupe virtual test baselines after rebaselining.

Project Member Reported by qyears...@chromium.org, Dec 14 2016

Issue description

See bug 663411#3 -- if a test with a corresponding virtual test is modified, then in general, both the test and virtual test should be rebaselined -- new baselines should be downloaded for both.

Then, in the optimize/de-dupe step, if the baselines for the virtual and non-virtual tests are equal, then 

Next step is to confirm this (by making a CL that modifies a test that has virtual tests and baselines, and then running webkit-patch rebaseline-cl).

The consequence of this bug is that some duplicate baselines are uploaded and stored.
 
That makes sense. What do you think of this solution? [1]

[1] https://codereview.chromium.org/2575383004
Labels: TE-NeedsTriageHelp
Labels: -TE-NeedsTriageHelp
Status: Started (was: Unconfirmed)
Owner: qyears...@chromium.org
Bug update: I can't set jack@nottheoilrig.com as the owner, but it seems like there should be *some* owner for "started" issues. jack@'s change above has now been reviewed.
Thanks for updating this, I plan to get back to it in the next week.
I've just been interrupted by holidays and other work.
No problem :-)
Ping jack@: any updates on https://codereview.chromium.org/2575383004? Could that CL be synced and committed?
Status: Assigned (was: Started)
Owner: ----
Status: Available (was: Assigned)
The next step here is to take Jack's started change, and make a new change and adjust it.
Cc: robertma@chromium.org skobes@chromium.org
Perhaps Robert or someone else can pick this up? I just hit this bug yesterday in https://chromium-review.googlesource.com/c/chromium/src/+/1113310 (see patchset 3 vs patchset 4).
I don't understand the original description: 
> Then, in the optimize/de-dupe step, if the baselines for the virtual and non-virtual tests are equal, then 
(then what? Quinten, could you elaborate?)

I spent a lot of effort on rebaseline/optimize last year. The current behaviour is that baseline optimizer WILL remove virtual baselines if they are redundant (identical to the non-virtual ones). Is this not desired and why?
In the case of the CL I mentioned, the virtual baselines were _not_ removed even though they were identical.

Sign in to add a comment