Closed Bug 2074540 Opened 1 month ago Closed 1 month ago

Text missing on all google docs

Categories

(Web Compatibility :: Site Reports, defect)

defect

Tracking

(Webcompat Priority:P2, Webcompat Score:5, firefox-esr140 unaffected, firefox-esr153 unaffected, firefox156 unaffected, firefox157 unaffected, firefox158+ disabled, firefox159+ verified)

VERIFIED FIXED
159 Branch
Webcompat Priority P2
Webcompat Score 5
Tracking Status
firefox-esr140 --- unaffected
firefox-esr153 --- unaffected
firefox156 --- unaffected
firefox157 --- unaffected
firefox158 + disabled
firefox159 + verified

People

(Reporter: sclements, Assigned: jnicol)

References

(Regressed 1 open bug, Regression, )

Details

(4 keywords, Whiteboard: [autowebcompat:processed][autowebcompat:repro-failed][webcompat:sightline][webcompat:japan][webcompat:core])

User Story

autowebcompat-repro-status:failed
autowebcompat-repro-reason:not_reproducible
autowebcompat-repro-report-os:all
platform:mac
impact:site-broken
configuration:common
affects:all
branch:release
diagnosis-team:graphics
user-impact-score:150

Attachments

(3 files, 1 external file, 1 obsolete file)

On 158.0a1 (2026-09-22) Nightly none of my google docs are rendering text. It looks like it starts to render the text on a restart but then it completely disappears. Refreshing didn't work, restarting the browser didn't work, logging out/back in to my moco email didn't work. I had to sign into my gmail account on safari to be able to read the doc.

I can't repro it on latest main so maybe it was fixed?

Whiteboard: [autowebcompat:processed]

Seems like the nightly build in question and latest firefox/main at the time were both https://github.com/mozilla-firefox/firefox/commit/f9a73919fe039381777330403e5f94918735a5e7, which suggests that that either this is something buildconfig related (e.g. PGO) or there's some other external variable (e.g. Google serving different versions of the sites)

User Story: (updated)
Whiteboard: [autowebcompat:processed] → [autowebcompat:processed][autowebcompat:repro-failed]
User Story: (updated)
Whiteboard: [autowebcompat:processed][autowebcompat:repro-failed] → [autowebcompat:processed][autowebcompat:repro-failed][webcompat:sightline][webcompat:japan][webcompat:core]

I can't repro it on latest main so maybe it was fixed?

Is it still working for you today?

Flags: needinfo?(sclements)
Flags: needinfo?(sclements)

(In reply to Sky Schubert [:skyschub] from comment #2)

I can't repro it on latest main so maybe it was fixed?

Is it still working for you today?

No, it's not working for me at all on the latest nightly build. I've tried different profiles and different gmail accounts and the issue is the same with google doc text missing. Also seeing that sheets has black artifacts where text should be - see screenshot - and downloading a doc in pdf viewer is blank. I removed the Nightly build and reinstalled it again (didn't delete any cached files because I didn't want to delete my profile) and the issue persists. However, opening the same profile on a local nightly artifact build does not have the same issue at all.

So perhaps a build issue which was what James suggested in slack. Anything I should look into to help diagnose it? (and ideally resolve it)? I'm on macOS Tahoe 26.6.2 btw...

Marcus, any ideas about this?

User Story: (updated)
Flags: needinfo?(mstange.moz)

Here's what one of my google docs looks like (from team notes, hence the icons). It looks as though sometimes on page load/browser restart I can initially see the text render but then it changes to this (same with sheets - I see the text initially, then the black artifacts).

OK< I've triaged this assuming it's mac-only and specific to some configuration we haven't identified yet.

I don't see the issue on my silicon mac, so its on an older 2020 64bit iMac that I see the issue.

Attached file about:support details —

[Tracking Requested - why for this release]: Would be sad to ship this even if it only affects intel mac.

Since it looks like a recent regression mozregression probably helps?

Ran it on the older iMac and mozregression points to bug 2072535.

Flags: needinfo?(mstange.moz) → needinfo?(jnicol)
Regressed by: 2072535

Set release status flags based on info from the regressing bug 2072535

Hi Sarah, could you try setting the pref webgl.power-preference-override to both -1 and 1 and see whether either of those make a difference?

Flags: needinfo?(jnicol) → needinfo?(sclements)

(In reply to Jamie Nicol [:jnicol] from comment #13)

Hi Sarah, could you try setting the pref webgl.power-preference-override to both -1 and 1 and see whether either of those make a difference?

No, neither of those values fixes it.

Flags: needinfo?(sclements) → needinfo?(jnicol)

Set release status flags based on info from the regressing bug 2072535

Webcompat Priority: --- → P2
Webcompat Score: --- → 5

Noting for tracking purposes that latest beta is unaffected - I verified 157.0b5 doesn't have the same problem so per the regressing bug; I can check when the first 158 beta build is ready early next week if that's helpful at all.

Using the shared PBO upload path results in broken rendering,
especially noticeable on Google docs/sheets and PDFs, presumably due
to a driver bug.

Assignee: nobody → jnicol
Status: NEW → ASSIGNED

Yeah the pref flip which regressed this is gated to nightly-only, so beta should be unaffected.

I can reproduce, though perhaps not quite as severely (presumably depends on the exact document that are being rendered). That's on my macbook which has an intel and AMD GPU. But I have to force enable the AMD GPU first. I think the difference on your iMac is that the AMD GPU is always enabled rather than the Intel one. (Though curiously about:support still indicates an Intel GPU is present, but even when requesting the low power GPU the AMD one still appears to be used.

Flags: needinfo?(jnicol)

(adding back autowebcompat userstory edits as we that for tracking its reproduction rate statistics)

User Story: (updated)
Attachment #9648537 - Attachment is obsolete: true

The previous patch was incomplete. The bug is actually with the glMapBufferRange(GL_MAP_UNSYNCHRONIZED_BIT) path in WebGLBuffer::BufferSubData. This is indeed called by the PBO upload path, which is why the above patch fixed a subset of the issues. But it is also used by vertex data uploads, meaning the fix is incomplete. Previously I had only tested a simple PDF, which was fixed by the initial patch, but now testing on google sheets I can reproduce the remaining issues. New patch up shortly.

The unsynchronized WebGLContext::BufferSubData path causes rendering
glitches with Metal ANGLE on certain Radeon GPUs. Fall back to
glBufferSubData on affected devices.

Pushed by jnicol@mozilla.com: https://github.com/mozilla-firefox/firefox/commit/00701cbab942 https://hg.mozilla.org/integration/autoland/rev/c2e558acc66c Avoid unsynchronized glMapBufferRange with Metal ANGLE on Radeon GFX10 GPUs. r=gfx-reviewers,lsalzman
Status: ASSIGNED → RESOLVED
Closed: 1 month ago
Resolution: --- → FIXED
Target Milestone: --- → 159 Branch

The patch landed in nightly and beta is affected.

:jnicol, please make an uplift decision for beta:

  • For beta, nominate the patch for beta uplift approval if the fix should be included in this release, or set status-firefox158 to wontfix if the fix can wait for the next release.

See Requesting an Uplift for documentation on how to request an uplift.

For more information, please visit BugBot documentation.

Flags: needinfo?(jnicol)

158 isn't affected, the pref is only enabled on Nightly

Flags: needinfo?(jnicol)
Regressions: 2079262

Verified, this is working for me on both RC 159 and Nightly 161.

Tested with:

  • Browser / Version: Firefox 159.0-candidate build 1 / Firefox Nightly 161
  • Operating System: Mac OS Sequoia 15.7.9
Status: RESOLVED → VERIFIED
See Also: → 2091596
You need to log in before you can comment on or make changes to this bug.

Attachment

General

Created:
Updated:
Size: