Closed Bug 1084350 Opened 10 years ago Closed 10 years ago

2.64% Win8 Paint regression on Aurora (v.34) Oct 10th from push 47cc92f8ab46

Categories

(Core :: Panning and Zooming, defect)

x86
Windows 8
defect
Not set
normal

Tracking

()

RESOLVED INCOMPLETE

People

(Reporter: jmaher, Unassigned)

References

Details

(Keywords: perf, regression, Whiteboard: [talos_regression])

We have a slight bump in tpaint:
http://graphs.mozilla.org/graph.html#tests=%5B%5B82,52,31%5D,%5B82,53,31%5D%5D&sel=1410950998213,1413542998213&displayrange=30&datatype=running

It looks to be related to this push for bug 1078373:
http://hg.mozilla.org/releases/mozilla-aurora/pushloghtml?changeset=47cc92f8ab46

Here is more information about tpaint:
https://wiki.mozilla.org/Buildbot/Talos/Tests#tpaint

this change made it to beta as well (luckily beta saw a larger improvement, so this regression ended up as an overall improvement when we uplifted)
Win8 isn't metro, right? On desktop windows this code should have exactly the same behaviour as before, because IsScrollPositionClampingScrollPortSizeSet should be returning false.

I can do a try push with some MOZ_CRASHes to verify.
Also the graph you linked doesn't really show the regression. Here's one for just aurora:

http://graphs.mozilla.org/graph.html#tests=[[82,52,31]]&sel=1413068212804.22,1413257354147.7703,137.27272709020588,147.26076536771782&displayrange=7&datatype=running

My try push (hopefully I picked the right syntax) is at https://tbpl.mozilla.org/?tree=Try&rev=6fb2d58f862f, and I also retriggered some of the tests on changesets around my cset on aurora. Will post back with findings.
(In reply to Kartikaya Gupta (email:kats@mozilla.com) from comment #2)
> My try push (hopefully I picked the right syntax) is at
> https://tbpl.mozilla.org/?tree=Try&rev=6fb2d58f862f, and I also retriggered
> some of the tests on changesets around my cset on aurora. Will post back
> with findings.

The retriggers are showing up on the graph in comment 2, and they seem to absolve my change (obviously I'm biased so somebody else should look too). My try push was useless since it didn't include the Win 8 PGO build which is the one that had the regression. Trychooser doesn't seem to have a way to build PGO, so I'm not sure what to do there. I'll file a bug about that.
There's already a bug about that.
Depends on: 691673
Kats,  thanks for jumping on this- I think I filed the bug prematurely as this appears to be related to the uplift. although we do see a slight bump in beta, but I suspect that was the delta from merging 34 into aurora in early September.

Let me hunt into this a bit more.
Based on discussion above this is likely not a regression from bug 1078373, so removing dependency.
No longer blocks: 1078373
Any progress here? I'm doing a cleanup of the Panning and Zooming component bugs :)
Flags: needinfo?(jmaher)
no progress, I had dropped this from my todo list.  This is a low priority regression (single platform, single test, 2.64%)
Flags: needinfo?(jmaher)
Ok, thanks.
Status: NEW → RESOLVED
Closed: 10 years ago
Resolution: --- → INCOMPLETE
You need to log in before you can comment on or make changes to this bug.