1513604 - 3GB memory use on youtube livechat (1.2GB in SVG emojis. Rest in heap-unclassified,detached, and GC)

Status: RESOLVED DUPLICATE of bug 1509911

(Core :: Graphics: ImageLib, defect, P3)

Description

[Mayank Bansal]

Attachment: memory-report.json.gz

User Agent: Mozilla/5.0 (Windows NT 10.0; Win64; x64; rv:66.0) Gecko/20100101 Firefox/66.0

Steps to reproduce:

enable WR
go to youtube livechat like : https://www.youtube.com/watch?v=RLxCfyQSBCs
start playing. Do the following in succession for 10-12 times.

1. Show chat
2. Click on the emoji button to load the emoji keyboard for chatting . (This takes 100% CPU, but thats another bug)
3. Wait till emojis load
4. Hide the chat


Actual results:

after repeating this for 4-5 times, you will see a rise in memory. This is partly due to bug 1509911.
I have done this in WR only. For non-WR, there is another bug that I filed


Expected results:

not so, hopefully.

Comment 1

[Mayank Bansal]

Attachment: aboutsupport.txt

Comment 2

[Andrew Osmond [:aosmond] (he/him)]

It looks like we save a copy of the source data each time it is loaded. That adds up over time.

Comment 3

[Andrew Osmond [:aosmond] (he/him)]

Bug 1202085 made us reload the SVGs because it appears the emoji panel is a controlled document / for a service worker. We remove the emojis from the cache when the worker is closed (e.g. the chat is hidden) but something is keeping the requests alive longer than I would think it should. We can't reuse the cache even if it wasn't removed because when the chat is reopened, it is for a new controlled document.

Only concern I have now is what exactly is keeping the "uncached" requests alive and holding onto the memory. A memory minimize had no impact, I just had to wait.

Comment 4

[Mayank Bansal]

(In reply to Andrew Osmond [:aosmond] from comment #3)
> Only concern I have now is what exactly is keeping the "uncached" requests
> alive and holding onto the memory. A memory minimize had no impact, I just
> had to wait.

could be this : https://bugzilla.mozilla.org/show_bug.cgi?id=1509911#c2

Comment 5

[Andrew Osmond [:aosmond] (he/him)]

(In reply to Mayank Bansal from comment #4)
> (In reply to Andrew Osmond [:aosmond] from comment #3)
> > Only concern I have now is what exactly is keeping the "uncached" requests
> > alive and holding onto the memory. A memory minimize had no impact, I just
> > had to wait.
> 
> could be this : https://bugzilla.mozilla.org/show_bug.cgi?id=1509911#c2

Yes, if the iframe remains alive, that would do it.

Comment 6

[Andrew Osmond [:aosmond] (he/him)]

Duping because if we fix the iframe issue, and the image memory issues should go away.