Multiple issues after client restart

[dbkr]

Steps to reproduce

1. Reload app from one of the develop builds on 02/02 to develop as of now (10:00 03/02)

Multiple reports of this have come in now, some reloading from yesterday's nightly build and some reloading develop. I saw it when restarting nightly this morning.

Clear cache & reload fixes it (which technically is a workaround, which would make this 'minor', but I'm going to apply common sense here and make it critical)

Outcome

• Many rooms have lost their names (shown as 'empty room')
• Often directed to old versions of rooms first, and then have to click through manually to their newest version
• Room list is missing message previews until I switch to the room
• Receiving notifications for strange rooms that don't usually notify me

Operating system

No response

Browser information

No response

URL for webapp

No response

Application version

No response

Homeserver

No response

Will you send logs?

No

[jryans]

I believe this list of commits has all the new things I applied this morning when the issues started, though I am not 100% certain.

[dbkr]

Thanks. I got the bug with the nightly build which finished at 8:47: before any of the 3 commits today, so that rules out those three.

[jryans]

...though after scanning those changes, I can't see how they could trigger anything like these issues... 😰

[jryans]

Given the areas affected, I would have expected JS SDK changes, but I do not recall seeing any in the changelog before updating...

[jryans]

Using /devtools to try viewing room state makes it clear that state events seem to just be... missing from the room model.

[jryans]

I do notice that the IndexedDB store worker appears to startup twice, which doesn’t seem to happen on app.

[jryans]

The sync data in IndexedDB for the affected session is extremely minimal: it only seems to have some read receipts... perhaps 2 copies of the store logic are racing?

[jryans]

I'm only noticing duplicate runs of the store worker in my main Firefox session, though some are affected on the desktop build as well.

[jryans]

...and now the worker only starts up once. 😵

[jryans]

Well, I've run out of theories to investigate, so I suppose I'll clear cache now.

[turt2live]

Fwiw, I haven't yet checked nightly or web, but mobile (android) is suddenly using default naming on a handful of rooms and also lost their avatar. This might be a server issue.