1737127 - Comparitech broadband speed test has a glitch in the meter

Status: RESOLVED WORKSFORME

(Core :: Graphics, defect)

Description

[B.J. Herbison]

Attachment: The off kilter view.

Steps to reproduce:

• Visit https://www.comparitech.com/internet-providers/speed-test/ and start the test.

Expected:

• The graph looks reasonable during the test.

Actual:

• See the attached screen capture.

The meter goes away at the end of the test.

Reported in https://chat.mozilla.org/#/room/#nightly:mozilla.org but the problem also exists in Release 93.

Comment 1

[Timea Cernea [:tbabos][inactive]]

Hi B.J. Herbison,

I was not able to reproduce this on the latest Firefox Release 93.0 or Beta 94.0b10 on MacOS 10.15.
Could you please mention on which OS is this reproducible for you?
Also, please also go through the troubleshooting steps as well and let us know if this is still reproducible. (Safe Mode and New Profile).

Comment 2

[B.J. Herbison]

I can still reproduce it on Ubuntu 20.04 using XFCE using clean profiles on Release and Nightly. The glitch happened most of the way through the Download test.

Comment 3

[B.J. Herbison]

The person who discovered the glitch was using Windows 10.

Comment 4

[Timea Cernea [:tbabos][inactive]]

I can reproduce now, I expected to see it from the very beginning of the test but it can happen during the Upload test as well, which was my case on MacOS 10.15.
Reproduced on all latest Firefox versions on MacOS 10.15 and Windows 10 as well and please note that this is not reproducible on Chrome.

This also seems to be an old regression, the last good build which did not have this issue was in 2020-06-18.
Unfortunately, mozregression couldn't bisect further so here is the last pushlog: https://hg.mozilla.org/mozilla-central/pushloghtml?fromchange=cbec1e2c0426c8ab7440fe0931d24c25ceef5ed0&tochange=44e820d9be261e3f7a83156122514ed693442d98

Further manual bisection found the push that introduced the issue: https://hg.mozilla.org/mozilla-central/pushloghtml?changeset=341563fe54635619fd28b8928d5876dbcc5a21e6

From here I am not sure which issue to point out as a regressor. Moving this over to the Graphics component for further investigations, thanks for the report!

Comment 5

[Darkspirit]

Gnome Xwayland, Debian Testing, Intel
Glitch occurs during Upload test.
After the glitch shows up, Context Menu > View image opens blob:null/afb65b7e-b839-4d8b-85dd-b3a700cc17c5 which also shows the bug.

Also reproducible with WebRender on 2019-01-01, but not with Basic.

Also reproducible with 2018-01-10:
MOZ_DISABLE_CONTENT_SANDBOX=1 mozregression --launch 2018-01-10 --pref gfx.webrender.all:true layers.acceleration.force-enabled:true gfx.webrender.enabled:true gfx.webrendest.enabled:true gfx.webrender.layers-free:true gfx.webrender.blob-images:true image.mem.shared:true layout.display-list.retain:false -a https://www.comparitech.com/internet-providers/speed-test/

MOZ_DISABLE_CONTENT_SANDBOX=1 mozregression --good 2017-08-10 --bad 2018-01-10 --pref gfx.webrender.all:true layers.acceleration.force-enabled:true gfx.webrender.enabled:true gfx.webrendest.enabled:true gfx.webrender.layers-free:true gfx.webrender.blob-images:true image.mem.shared:true layout.display-list.retain:false -a https://www.comparitech.com/internet-providers/speed-test/

14:37.35 INFO: Last good revision: e070277ec199fa96fa490ed52d33646a376d0d80 (2017-11-15)
14:37.35 INFO: First bad revision: a3f183201f7f183c263d554bfb15fbf0b0ed2ea4 (2017-11-16)
14:37.35 INFO: Pushlog:
https://hg.mozilla.org/mozilla-central/pushloghtml?fromchange=e070277ec199fa96fa490ed52d33646a376d0d80&tochange=a3f183201f7f183c263d554bfb15fbf0b0ed2ea4

There are some Canvas, WebRender, SVG changes:

4a74a02c20ba4649bfdb5bdd49f5907c6f22b2d5 sotaro — Bug 1411472 - Add support of PersistentBufferProviderShared r=nical
60ed7a647f38bc9dc34d2f54838d01e03b29d9d2 Ethan Lin — Bug 1415407 - Add webrender support for filters (hue-rotate, opacity, saturate). r=kats
b98f2fff0e174ca3e7a677255bd9049431860f3d Botond Ballo — Bug 1415963 - Take into account stroke (not just fill) when calculating the contribution of an SVG element to a clip rect for a clip path. r=mstange
0ad4e810bab1313edd3d53b71f69afa692f5dc46 Andrew Osmond — Bug 1401672 - Make display items for the same WebRenderBridgeParent/Child share the ImageKey for shared surfaces. r=jrmuizel

Disabling browser.tabs.drawInTitlebar in first bad build doesn't help.
MOZ_DISABLE_CONTENT_SANDBOX=1 mozregression --launch 2017-11-16 --pref gfx.webrender.all:true layers.acceleration.force-enabled:true gfx.webrender.enabled:true gfx.webrendest.enabled:true gfx.webrender.layers-free:true gfx.webrender.blob-images:true image.mem.shared:true layout.display-list.retain:false browser.tabs.drawInTitlebar:false -a https://www.comparitech.com/internet-providers/speed-test/

The glitch also occurs with "OpenGL" compositing, but not with Basic
MOZ_DISABLE_CONTENT_SANDBOX=1 mozregression --launch 2017-11-16 --pref layers.acceleration.force-enabled:true gfx.webrender.enabled:false gfx.webrendest.enabled:false gfx.webrender.layers-free:true gfx.webrender.blob-images:true image.mem.shared:true layout.display-list.retain:false -a https://www.comparitech.com/internet-providers/speed-test/

Might the canvas use some invisible OpenGL Gtk window with Gtk double buffering enabled? (bug 1502519 comment 27: OpenGL double buffering is expected, but additional Gtk double buffering causes glitches)

Comment 6

[BugBot [:suhaib / :marco/ :calixte]]

:tbabos, since this bug is a regression, could you fill (if possible) the regressed_by field?
For more information, please visit auto_nag documentation.

Comment 7

[B.J. Herbison]

I no longer see the glitch.