Closed Bug 1814976 Opened 2 years ago Closed 2 years ago

Google doc shows error with constant reload

Categories

(Core :: Graphics, defect, P2)

defect

Tracking

()

RESOLVED WORKSFORME

People

(Reporter: mconca, Unassigned)

References

Details

Attachments

(1 file)

When any open Google Doc is left open and unused in a tab for multiple days, it eventually ends up in a state where the content is greyed out (or sometimes missing), the tab shows the "throbber" indicating it is trying to load, and there is an alert() modal dialog over the page stating:

"This error has been reported to Google and we'll look into it as soon as possible. Please reload this page to continue." (see image)

Reloading the page does indeed seem to restore it to normal. No exact steps to reproduce, other than load a doc and leave it in a tab for many days. Doesn't seem to impact Google Sheets or Slides, only Docs. I've seen it on pinned and unpinned tabs.

I reached out to Google about this. Their response indicated that, yes, they are seeing an increased error rate since the rollout of Firefox 109. The top error they are seeing is thrown when calling context.scale on an HTML canvas. The same error can be induced by making a very large canvas.

Did anything change with canvas in Firefox 109? (pinging some graphics folks)

Component: Networking → Graphics
Flags: needinfo?(jmuizelaar)
Flags: needinfo?(jmathies)
Flags: needinfo?(bhood)

What platform are we seeing this on, MacOS?

Flags: needinfo?(jmathies) → needinfo?(mconca)

I have experienced this too recently. I'm using MacOS.

Not just macOS, I also see this on Linux.

See Also: → 1814961

Windows 11 for me

Flags: needinfo?(mconca)

This feels like a network timeout/disconnect to me, with reloading the page causing a re-connect to Google's cloud servers. Miro does this as well with a shorter timeout, but handles it much more gracefully. It just reconnects as soon as the tab comes into focus again.

I see it started in Networking. Was it proven there that no timeout/disconnect was actually occurring?

Flags: needinfo?(bhood)
Severity: -- → S2
Priority: -- → P2

I imagine they're using a WebSocket or other such mechanism to keep an open Doc synced with the clowd, displaying who has it open, etc. If that connection dies and they fail to properly re-establish it, that might cause this sort of error. Just a guess

(In reply to Bob Hood [:bhood] from comment #6)

This feels like a network timeout/disconnect to me, with reloading the page causing a re-connect to Google's cloud servers.

If there's a networking issue, could it be bug 1814961?

I see it started in Networking. Was it proven there that no timeout/disconnect was actually occurring?

"The top error they are seeing is thrown when calling context.scale on an HTML canvas." from comment 1 sounded like it was not (only) a network issue. Or are you suggesting some of the page's JS code could be misbehaving after a timeout and cause the canvas to error because of bad input?

Or are you suggesting some of the page's JS code could be misbehaving after a timeout and cause the canvas to error because of bad input?

That hadn't occurred to me, but definitely sounds plausible.

Not sure if Google fixed something on their end, but I haven't seen this for about a week now.

(In reply to Florian Quèze [:florian] from comment #8)

(In reply to Bob Hood [:bhood] from comment #6)

This feels like a network timeout/disconnect to me, with reloading the page causing a re-connect to Google's cloud servers.

If there's a networking issue, could it be bug 1814961?

Please note that the patch on bug 1814961 has landed today and is now in the Firefox Nightly.

Reporter says the bug stopped happening, closing.

Status: NEW → RESOLVED
Closed: 2 years ago
Flags: needinfo?(jmuizelaar)
Resolution: --- → INCOMPLETE
Resolution: INCOMPLETE → WORKSFORME
You need to log in before you can comment on or make changes to this bug.

Attachment

General

Created:
Updated:
Size: