Google sheet loading hangs with half of the screen black
Categories
(Core :: JavaScript Engine: JIT, defect, P2)
Tracking
()
| Performance Impact | high |
People
(Reporter: jstutte, Unassigned)
References
Details
(Keywords: perf:pageload)
Loading this sheet:
Nightly's web content tab seems to hang in an endless loop inside Javascript execution:
https://share.firefox.dev/3cIoA3P
Until yesterday it worked fine.
|
||
Comment 1•5 years ago
|
||
Looking at the profile quickly, we are hitting the Warp JIT, have approx acceptable GC behaviour, etc.
There is an extreme amount of time in setTimeout handlers which seems like something is broken. Probably needs to be compared against chrome and use user-agent overrides to ensure we are getting the same code as Chrome.
|
|
||
Comment 2•5 years ago
|
||
Actually, the setTimeouts in the marker-chart show that it isn't thousands of timeouts, but individual ones that take 600ms. The bulk of the time is in the JIT code itself (and not even the C++ fallback paths it sometimes uses). It is unclear why the page does not settle and continues to fire the timeouts over and over again.
|
||
Updated•5 years ago
|
|
|
||
Comment 3•5 years ago
|
||
Running the spreadsheet myself, I see the same black screen hang with Chrome. The document is quite complex and gdocs seems to load it all before finishing its internal render.
That said, I see Fx Nightly taking several seconds longer to fully load (20s vs 13s doing a stopwatch by hand).
|
Reporter | |
Comment 4•5 years ago
|
||
Hmmm, since the doc worked without problems (in Firefox) until March 17 I would assume, Google changed something in GDocs then? BTW, it is not important for me to recover that doc, I will not touch it to not interfere with investigations.
|
|
||
Comment 5•5 years ago
|
||
Profiles of specific region of comparable code that is largely pure JS JIT:
Fx: https://share.firefox.dev/3vVwUG5
Cr: https://share.firefox.dev/3rjHkMm
(Sample counts are not directly comparable, but "Traced running time" in sidebar is closer).
Chrome does much more aggressive inlining here. Iain and I are discussing a few experiments here.
|
||
Updated•5 years ago
|
|
||
Updated•4 years ago
|
|
||
Comment 6•4 years ago
|
||
Can this issue be retested and have a new profile completed to determine if it is still an issue?
|
|
Reporter | |
Comment 7•4 years ago
•
|
||
(In reply to Frank Doty [:fdoty] from comment #6)
Can this issue be retested and have a new profile completed to determine if it is still an issue?
Sure (though I switched machine in the meantime, meaning that it is generally faster). It seems to load after a few secs normally, and the profile shows that the thread settles after a while, too.
Tested on FF 102.0.1 (release).
I think this means a works-for-me, then.
|
|
||
Comment 8•4 years ago
|
||
Closing this bug as works-for-me based on this comment
(In reply to Jens Stutte [:jstutte] from comment #7)
(In reply to Frank Doty [:fdoty] from comment #6)
Can this issue be retested and have a new profile completed to determine if it is still an issue?
Sure (though I switched machine in the meantime, meaning that it is generally faster). It seems to load after a few secs normally, and the profile shows that the thread settles after a while, too.
Tested on FF 102.0.1 (release).
I think this means a works-for-me, then.
Description
•