Open Bug 1650436 Opened 6 years ago Updated 1 year ago

Janky scrolling on bose.com with WebRender enabled

Categories

(Core :: Graphics: WebRender, defect, P3)

defect

Tracking

()

Tracking Status
firefox-esr68 --- unaffected
firefox-esr78 --- unaffected
firefox78 --- unaffected
firefox79 --- wontfix
firefox80 --- wontfix
firefox81 --- wontfix
firefox82 --- wontfix
firefox83 --- fix-optional

People

(Reporter: emilio, Unassigned)

References

(Blocks 2 open bugs)

Details

(Keywords: perf, regression)

(I started filing this as a perf bug, thus the template)

Basic information

Steps to Reproduce: Load https://www.bose.com/en_us/support/article/no-sound-from-connected-device0.html. This is more noticeable in Android / GeckoView, but also happens in desktop, both with and without WR.

Expected Results: Smooth scrolling.

Actual Results: Scrolling janks

More information

Happens with and without WR enabled.

Profile URL: https://share.firefox.dev/3dVVahj (there doesn't seem to be much going on there unfortunately...).

Given scrolling is not janky when scrolling with keyboard, I suspect this may be a site issue... 68 definitely seems much snappier than nightly. I'm not on a good enough connection to do a proper bisection right now.

75 looks good too.

Hmm... Though now I don't seem to consistently repro on desktop... I can repro on my profile always though... On android nightly the page is pretty unusable :(

Ah, my android device now uses WR by default, which explains why I thought that this reproed with and without WR (I thought I initially had it disabled), and that I could repro on desktop on my profile but not while running mozregression.

On Android this page kills Firefox's performance.

Component: Panning and Zooming → Graphics: WebRender
Summary: Janky scrolling on bose.com → Janky scrolling on bose.com with WebRender enabled

https://share.firefox.dev/2YUz3Dy is a hopefully more useful profile from desktop.

Severity: -- → S3
Keywords: perf
Priority: -- → P3

Managed to reproduce the issue on a device with macOS 10.15.5 as well.

Regression range, picked up manually:

  • Last good: 2020-06-18-21-28-28
  • First bad: 2020-06-19-09-21-44
  • Pushlog: URL

Marking 79 as affected per the regression range.

Christian, can you get a tighter regression range?

Flags: needinfo?(cristian.fogel)

This was the best I could get; limitations being that it didn't seem to reproduce via mozregression which I found odd.
Checked with people from our Mobile team and they're limited to manually getting the regression ranges as well.

Let me know if there is something I can help you with, to push this further.

Flags: needinfo?(cristian.fogel)

Loading this page in GVE with WR enabled on a Pixel2 totally destroys the browser. It starts to load, then either goes white right away or on scrolling, and nothing else loads after that. It was happening on the 2020-06-17 and 2020-06-01 builds as well so I'm not sure how much I trust the regression range from macOS. bug 1654852 is making it hard for me to actually use mozregression for this.

Even seeing it on the 2020-01-01 GVE build.

Kats, should we be tracking this for release builds? The priority is low but your last comment makes this sounds pretty bad.

Flags: needinfo?(kats)

It is bad, but it's not a recent regression. It's probably worth tracking and prioritizing though.

Flags: needinfo?(kats)

From Kats - might have to do with blob images. May be a dupe of Bug 1662297 - Zillow using massive amounts of memory.

See Also: → 1662297
See Also: → 1680999
Webcompat Priority: --- → ?
Webcompat Priority: ? → P2

I tried to reproduce this, but couldn't. Emilio: can you still repro? Otherwise I think we close it as WFM.

Webcompat Priority: P2 → ---
You need to log in before you can comment on or make changes to this bug.