Google Spreadsheets have black and blank cells; GPU Helper "Connection Invalid error for service com.apple.hiservices-xpcservice"
Categories
(Core :: Graphics: Canvas2D, defect)
Tracking
()
| Tracking | Status | |
|---|---|---|
| firefox-esr115 | --- | unaffected |
| firefox-esr153 | --- | unaffected |
| firefox160 | + | fixed |
| firefox161 | + | fixed |
| firefox162 | + | verified |
People
(Reporter: cpeterson, Unassigned)
References
(Regression)
Details
(Keywords: regression)
Attachments
(1 file)
|
233.80 KB,
image/png
|
Details |
On my old Intel iMac running macOS 15.7.9 (Sequoia), Google spreadsheets show blacks and blank cells instead of data. Enabling gfx.software.webrender = true and restarting fixes the problem.
I see the following GPU Helper errors logged to mozgression's console:
2026-11-11T16:09:39.655000: INFO : b'2026-11-12 00:09:39.654 Firefox Nightly GPU Helper[5169:187703] Failure on line 687 in function id scheduleApplicationNotification(LSNotificationCode, NSWorkspaceNotificationCenter *): noErr == _LSModifyNotification(notificationID, 1, &code, 0, NULL, NULL, NULL)'
2026-11-11T16:09:39.759000: INFO : b'2026-11-12 00:09:39.758 Firefox Nightly GPU Helper[5169:187712] Connection Invalid error for service com.apple.hiservices-xpcservice.'
2026-11-11T16:09:39.763000: INFO : b'2026-11-12 00:09:39.758 Firefox Nightly GPU Helper[5169:187703] Error received in message reply handler: Connection invalid'
| Reporter | ||
Comment 1•4 days ago
|
||
@lsalzman: this bug is a regression in Fx 158 from this pushlog. mozregression couldn't bisect any further, so I don't know if this regressor is CFF font bug 2072852, ImageBridgeParent.cpp bug 2073200, or PatternT::IsWeak() bug 2072448.
| Reporter | ||
Comment 2•4 days ago
|
||
[Tracking Requested - why for this release]:
This bug is a regression since Fx 158 that makes Google Sheets unusable on my Intel iMac running macOS 15.7.9 (Sequoia).
Comment 3•4 days ago
|
||
(In reply to Chris Peterson [:cpeterson] from comment #1)
@lsalzman: this bug is a regression in Fx 158 from this pushlog. mozregression couldn't bisect any further, so I don't know if this regressor is CFF font bug 2072852, ImageBridgeParent.cpp bug 2073200, or PatternT::IsWeak() bug 2072448.
Can you try reverting each of these changes individually in a local build to narrow down which one?
Comment 4•4 days ago
|
||
I am tracking for releases in flight, including on the release channel in case we have a safe fix this week.
Comment 5•4 days ago
|
||
The bug is marked as tracked for firefox160 (release), tracked for firefox161 (beta) and tracked for firefox162 (nightly). However, the bug still isn't assigned.
:bhood, could you please find an assignee for this tracked bug? If you disagree with the tracking decision, please talk with the release managers.
For more information, please visit BugBot documentation.
Comment 6•3 days ago
|
||
I tried to reproduce on a 2019 16" MacBook Pro (Intel UHD 630 + AMD Radeon Pro 5300M/5500M, macOS 15.8.1), with hardware WebRender and accelerated Canvas2D on, using both GPUs (webgl.power-preference-override=1 to force AMD). The Google Sheet I was using rendered fine on both.
Chris, could you attach your about:support (Copy raw data to clipboard)? I'd mainly like the GPU and the WebGL renderer string. Can you also share a spreadsheet that reproduces, or a copy with the data removed, or describe what's in the cells that go black (fills, images, conditional formatting, charts)? And does setting gfx.canvas.accelerated=false, with gfx.software.webrender left false, also fix it? That would narrow it to accelerated Canvas2D rather than WebRender.
Comment 7•3 days ago
|
||
ANGLE WebGL was enabled on Nightly in 158 (and rode the trains in 160). So almost certainly a regression from that. I believe software webrender disables accelerated canvas, which would explain why that solves it.
This is almost certainly a dup of bug 2091596. But if Chris can provide the WebGL renderer string we can confirm. I don't think bug 2092355 is related - WebGL fails to initialise altogether in that bug, rather than initialising but rendering being broken.
Updated•3 days ago
|
| Reporter | ||
Comment 8•3 days ago
|
||
(In reply to Timothy Nikkel (:tnikkel) from comment #6)
does setting gfx.canvas.accelerated=false, with gfx.software.webrender left false, also fix it? That would narrow it to accelerated Canvas2D rather than WebRender.
Yes, gfx.canvas.accelerated=false (with gfx.software.webrender = true) also fixes the problem.
(In reply to Jamie Nicol [:jnicol] from comment #7)
This is almost certainly a dup of bug 2091596. But if Chris can provide the WebGL renderer string we can confirm. I don't think bug 2092355 is related - WebGL fails to initialise altogether in that bug, rather than initialising but rendering being broken.
My renderer string from about:support is:
Google Inc. (AMD) -- ANGLE (AMD, ANGLE Metal Renderer: AMD Radeon Pro 575X, Version 15.7.9 (Build 24G830))
| Reporter | ||
Comment 9•3 days ago
|
||
Do you still need my about:support? Or need me to test other builds?
Comment 10•2 days ago
|
||
Chris, the patch in bug 2091596 landed in 162.0a1 20261113085116 — could you try latest Nightly to check if that issue is fixed for you to help confirm the dupe? TY!
Updated•2 days ago
|
Updated•2 days ago
|
| Reporter | ||
Comment 11•2 days ago
|
||
(In reply to Jan Brasna [:janbrasna] from comment #10)
the patch in bug 2091596 landed in 162.0a1 20261113085116 — could you try latest Nightly to check if that issue is fixed for you to help confirm the dupe?
Good news! The bug is fixed in the latest Nightly.
I'll close this bug as fixed in 160–162, copying the status flags from PDF bug 2091596.
| Reporter | ||
Updated•2 days ago
|
Updated•15 hours ago
|
Description
•