Open Bug 964448 Opened 6 years ago Updated 6 years ago

many JS strings build up during browser-chrome tests

Categories

(Testing :: General, defect)

defect
Not set

Tracking

(Not tracked)

People

(Reporter: froydnj, Unassigned)

References

(Blocks 1 open bug)

Details

Looking at about:memory dumps for browser-chrome tests, I see several megabytes of JS strings building up.  A significant majority of them begin with "Console message", but there are various other messages, some from devtools, some random messages about thinks, some that look like sessionstore entries (!), and so forth.

Depending on which snapshots of about:memory you look at, these "notable" strings take from 1-3% of about:memory's explicit tree, which is certainly enough to care about.

What I'm not clear on is who's holding on to these strings for so long.  nsConsoleService doesn't hold on to many messages (there are several strings with  more copies than the number of messages that nsConsoleService buffers, browser-test.js doesn't hold on to messages from nsIConsoleService for very long, and so forth.  Are the devtools tests generating all of these and holding on to them?
I also saw some console message strings in bug 946546. In my case it wasn't a leak; I was more worried about fragmentation from an unnecessary allocation. But there is some mention of caching in the bug... maybe related?
You need to log in before you can comment on or make changes to this bug.