Closed Bug 952620 Opened 12 years ago Closed 12 years ago

Big regression in Eideticker checkerboarding metric on Galaxy Nexus starting December 12

Categories

(Firefox for Android Graveyard :: Toolbar, defect)

Product:
Firefox for Android Graveyard
Component:
Toolbar
Platform:
x86_64
Linux
Type:
defect
Priority:
Not set
Severity:
normal

Tracking

(Not tracked)

Status:
RESOLVED INVALID

People

(Reporter: wlach, Unassigned)

Details

(Whiteboard: [eideticker-regression])

We're seeing a big regression on the scrolling tests, even ones that have historically had no checkerboarding at all. Examples: http://eideticker.wrla.ch/#/samsung-gn/wikipedia/checkerboard http://eideticker.wrla.ch/#/samsung-gn/taskjs/checkerboard Is this a known issue?
tracking-fennec: --- → ?
Strangely not really seeing this issue on the g2x: http://eideticker.wrla.ch/#/lg-g2x/taskjs/checkerboard
Can we bisect this? Alternatively a profile would be nice but a bisection would be better.
Agreed. It's on the Galaxy Nexus, but not the LG. I see this for pageload too: http://eideticker.mozilla.org/#/samsung-gn/nytimes-load/timetostableframe http://eideticker.mozilla.org/#/samsung-gn/nytimes-load-poststartup/timetostableframe But I see hardly any change at on on the LG: http://eideticker.mozilla.org/#/lg-g2x/nytimes-load/timetostableframe http://eideticker.mozilla.org/#/lg-g2x/nytimes-load-poststartup/timetostableframe I do see what looks like a pageload regression on the LG that happens between Oct 24 and Oct 30. That would correlate to the landing of the "seer" code in bug 881804. A pageload regression has already been filed for it in bug 945779.
(In reply to Benoit Girard (:BenWa) from comment #2) > Can we bisect this? Alternatively a profile would be nice but a bisection > would be better. I have a profile from loading cnn.com on a Nexus 4 (Android 4.4.2) http://people.mozilla.org/~bgirard/cleopatra/#report=37d227989001939fbe432615ecd6ec8771651480
Can I get a profile to compare before the regression?
Sorry nevermind. We'd need to profile before/after the regression of panning (ideally directly from eideticker since the events would be the same).
(In reply to Benoit Girard (:BenWa) from comment #6) > Sorry nevermind. We'd need to profile before/after the regression of panning > (ideally directly from eideticker since the events would be the same). I will try to make this happen.
This is a profile of panning nytimes (mobile site) on a Nexus 4 (Android 4.4.2) http://people.mozilla.org/~bgirard/cleopatra/#report=c5fc04388ad67d08863e74c205c5356cf7abf7ac
Eugh, just did some more testing and I can't reproduce the problem. I think the problem was with the device. Rebooting it seems to have fixed the issues and I'm seeing normal looking results after the 11th.
Status: NEW → RESOLVED
Closed: 12 years ago
Resolution: --- → INVALID
If this is an anomaly, is it possible to remove the bad data? It is enough to throw off the scale of the rest of the data, making it harder to see new regressions.
tracking-fennec: ? → ---
(In reply to Mark Finkle (:mfinkle) from comment #10) > If this is an anomaly, is it possible to remove the bad data? It is enough > to throw off the scale of the rest of the data, making it harder to see new > regressions. Done.
Product: Firefox for Android → Firefox for Android Graveyard
You need to log in before you can comment on or make changes to this bug.