Open Bug 1941709 Opened 1 month ago Updated 29 days ago

Wikipedia header-bar has graphical artifacts in its background-color, on Samsung Galaxy S2

Categories

(Core :: Graphics: WebRender, defect)

defect

Tracking

()

People

(Reporter: dholbert, Unassigned, NeedInfo)

References

(Regression)

Details

(Keywords: regression)

Attachments

(3 files)

I just turned on my Samsung Galaxy S2 for the first time in a while and noticed it's gotten a ~recent regression in Nightly graphical rendering.

STR:

  1. Load https://en.m.wikipedia.org/wiki/Main_Page
  2. Look at the logo header-bar

ACTUAL RESULTS:
There's a light-gray bar within the header-bar, roughly as tall as the "Wikipedia" logo itself and slightly shorter than the search box.
That light gray bar has a small "tail" on the left edge.

EXPECTED RESULTS:
No such light gray bar; the header should have a consistent solid-gray background color.

I think this is a hardware WebRender bug (or a driver bug that's tripped by WebRender), because the issue goes away if I enable about:config pref gfx.webrender.software and restart Firefox. (See also bug 1884791 from earlier in the year with the same device.)

Regression range:
Newest known good nightly: 2024-10-14
Oldest known bad nightly: 2024-10-15

I'll come up with a pushlog in a bit (have to do so manually on Fenix).

Attachment #9459530 - Attachment description: screenshot of bug → screenshot of bug (zoomed in as much as fenix will allow, to make it more apparent)
Attachment #9459530 - Attachment description: screenshot of bug (zoomed in as much as fenix will allow, to make it more apparent) → screenshot of bug (zoomed in somewhat, to make it more apparent)

(In reply to Daniel Holbert [:dholbert] from comment #0)

Regression range:
Newest known good nightly: 2024-10-14
Oldest known bad nightly: 2024-10-15

I'll come up with a pushlog in a bit (have to do so manually on Fenix).

Regression pushlog: https://hg.mozilla.org/integration/autoland/pushloghtml?fromchange=4fabfc1559c9&tochange=495128bbc378

Skimming for WebRender stuff in there, I wonder if this is bug 1922323? (or maybe bug 1923790 or bug 1788866 which are graphics but not webrender).

Tentatively marking as regression from the webrender change - nical, maybe you could take a look when you've got cycles? Let me know if there's any additional diagnostic data I can capture to help. (As noted above, things work fine with software WR.)

Flags: needinfo?(nical.bugzilla)
Keywords: regression
Regressed by: 1922323

Looking at the site in devtools, it looks like this solid-gray background area is a background-color (not an image or anything fancy), on an element that extends across most of the page (though it's covered up by a white .mw-body element in the main content area). This is the element that provides this background-color styling:

#mw-mf-page-center {
	width: 100%;
	min-height: 100%;
	position: relative;
	background-color: var(--background-color-interactive,#eaecf0);
}

--background-color-interactive here is #eaecf0 (the same as the fallback value). That gray color ends up being the background for the header area, which renders with this glitch for some reason.

Summary: Wikipedia header-image has graphical artifacts, on Samsung Galaxy S2 t → Wikipedia header-bar has graphical artifacts in its background-color, on Samsung Galaxy S2

Also, if I tap the "Search" textfield, there's a transition where the "Wikipedia" logo and hamburger menu fade out of view -- and when that happens, this glitchy gray area starts dynamically flickering and changing its glitches during that animation. (And it often ends up with a slightly different set of artifacts as a result -- i.e. more/less light-gray bars than it started with.)

I'm not sure whether to rate this S3 or S2, if this occurs in any place that breaks something I'd reconsider it as possibly S2, not everyone knows how to disable hardware webrender so I think we need to either fix this or blocklist.

The driver was built in 2017 (but probably from an older codebase) and the number of telemetry pings that report this driver version is < 0.1% so I'm not especially concerned by the idea of blocklisting this version.

dholbert@ - thoughts?

Severity: -- → S3
Flags: needinfo?(dholbert)

S3 seems probably fine given it's gone a few months without notice/reports. Thanks!

Flags: needinfo?(dholbert)

(blocklisting seems maybe reasonable too. This device is running LineageOS to limp forward a bit to newer Android, but still hasn't had an update available in many years.)

You need to log in before you can comment on or make changes to this bug.

Attachment

General

Created:
Updated:
Size: