ChatGPT fails to load
Categories
(Web Compatibility :: Site Reports, defect, P2)
Tracking
(Webcompat Priority:P2, Webcompat Score:6)
People
(Reporter: adam, Unassigned)
References
()
Details
(Keywords: webcompat:needs-diagnosis, webcompat:site-report, Whiteboard: [webcompat:sightline])
User Story
platform:windows,mac,linux,android impact:site-broken configuration:general affects:few branch:release user-impact-score:200 diagnosis-team:webcompat
User Agent: Mozilla/5.0 (Windows NT 10.0; Win64; x64; rv:139.0) Gecko/20100101 Firefox/139.0
Steps to reproduce:
Use ChatGPT for a while, eventually it stops working
Actual results:
At random intervals (hours,days) https://chatgpt.com/ will cease to fully load. Past chats do not show up and none of the buttons work.
Clearing data/cookie does not fix.
Restarting Firefox does fix
Expected results:
ChatGPT keeps working...
|
||
Comment 1•1 year ago
|
||
Adam, do you have https://chatgpt.com/ open in a tab or in sidebar? Only chatgo does not load, other sites load? Thank you.
HI Monica,
- in a tab
- as far as I've seen this bug is only for ChatGPT.com
It is discussed by many users here over the last year.
https://community.openai.com/t/chatgpt-has-stopped-working-in-firefox/664425/93
|
||
Updated•1 year ago
|
|
||
Updated•1 year ago
|
|
||
Comment 3•1 year ago
|
||
The linked community thread is about ESR 115, which is... very old. It seems fine with newer releases, although there is a chance we see some rare intermittent issues? The thread linked might contain some helpful info, though, we should read through it.
|
|
||
Updated•1 year ago
|
|
||
Comment 4•1 year ago
•
|
||
(In reply to Dennis Schubert [:denschub] from comment #3)
The linked community thread is about ESR 115, which is... very old. It seems fine with newer releases, although there is a chance we see some rare intermittent issues?
FWIW the reporter here (Adam) seems to have been using Firefox 139, so this probably isn't ESR-specific.
The last two posts in the linked thread seem to have found a relevant TypeError and potentially an intervention that works for the person who noticed it there
Quoting those posts:
If you look in the console you’ll a lot of warnings and in my case, an error at the top saying_
Uncaught (in promise) TypeError: Performance.mark: Expected startTime >= 0
My hunch is that maybe Firefox doesn’t allow Performance.mark to have a startTime < 0,
but OpenAI (or a package they use) is doing some funk with specific timings and state causing a race-condition.
Note: the thing about negative startTime isn't Firefox-specific; the following JS snippet (executed in DevTools console) gives a TypeError in both Firefox and Chrome:
performance.mark("my mark", {startTime: -5})
But it's possible that OpenAI has a race condition that happens to trigger this negative-startTime situation in a Firefox-specific way, I suppose.
The reporter has a screenshot of DevTools showing some of the relevant code:
https://us1.discourse-cdn.com/openai1/optimized/4X/9/1/6/9164b581a5c70e3617630bcec2e059bf5249b5b8_2_1035x424.png
and they confirmed that hacking around this negative-startTime does fix the issue for them.
|
|
||
Comment 5•1 year ago
|
||
Looks like we have earlier bug 1971874 filed on this.
|
||
Comment 6•1 year ago
|
||
Adam, have you, or has anyone else seen this recently?
I haven't in the last week or so. But it wasn't occurring consistently so I can't say anything definite.
|
|
||
Comment 8•1 year ago
•
|
||
Please let us know if you do see it again. Per bug 1971874 comment 5 (specifically the Math.max(0,...) quoted there to avoid negative values), it looks like (as of today at least) the site's JavaScript has been fixed to no longer trigger this problem, or at least the problem that was identified in the discourse thread linked here.
Description
•