Closed
Bug 1080011
Opened 11 years ago
Closed 10 years ago
Lots of ghost windows and long cc times
Categories
(Core :: DOM: Core & HTML, defect)
Tracking
()
RESOLVED
INCOMPLETE
People
(Reporter: jrmuizel, Assigned: mccr8)
Details
(Whiteboard: [MemShrink:P2])
Attachments
(1 file)
|
361.10 KB,
application/x-gzip
|
Details |
No description provided.
Comment 2•11 years ago
|
||
(I haven't seen any huge CC times recently. Medium <20ms CC times with 195 tabs + Chatzilla open)
| Reporter | ||
Comment 3•11 years ago
|
||
| Reporter | ||
Comment 4•11 years ago
|
||
| Reporter | ||
Comment 5•11 years ago
|
||
I still have the browser running now, so if there's something else that you want gathered I can try.
| Assignee | ||
Comment 6•11 years ago
|
||
A non-verbose CC log would be nice. (I don't need the GC log for that one.)
| Assignee | ||
Comment 7•11 years ago
|
||
One global window 0x12faaec00 [nsGlobalWindow #3449 inner https://bugzilla.mozilla.org/show_bug.cgi?id=1079398]
is beind held alive by a nsXPCWrappedJS (nsITimerCallback).
| Reporter | ||
Comment 8•11 years ago
|
||
Here's the concise log: http://people.mozilla.org/~jmuizelaar/cc-logs/cc-edges.92817.1412794514.log.xz
| Assignee | ||
Comment 9•11 years ago
|
||
Yeah, I'm seeing this, too. I have 664 ghost windows now. And my browser is super janky. :)
Jeff, are you using e10s? I've found a patch that causes a bunch of leaked windows in e10s mochitests, maybe it is related.
Flags: needinfo?(jmuizelaar)
| Assignee | ||
Updated•11 years ago
|
Assignee: nobody → continuation
Whiteboard: [MemShrink]
| Assignee | ||
Comment 10•11 years ago
|
||
It looks like you are not using e10s mode, judging by the memory report.
Flags: needinfo?(jmuizelaar)
| Assignee | ||
Comment 11•11 years ago
|
||
On telemetry, it looks like the 95% percentile of ghost windows increased from 3 to 5 on 10/4.
| Assignee | ||
Comment 12•11 years ago
|
||
I'm pretty sure the issue I'm seeing is bug 1080301, because at least one of leaking windows are being held alive by an object element. But as I said, Jeff's leak is non-e10s, and is via a nsITimerCallback.
| Assignee | ||
Comment 13•11 years ago
|
||
The only thing (aside from canvas elements), which seems to be CCed properly, that looks like it owns nsITimerCallbacks is nsTimerImpl. Could something be firing off a bunch of timers? I'm not sure how we can diagnose that.
Updated•11 years ago
|
Whiteboard: [MemShrink] → [MemShrink:P2]
| Assignee | ||
Comment 14•11 years ago
|
||
Jeff, does this happen regularly to you? Do you have any idea how to reproduce it?
Flags: needinfo?(jmuizelaar)
| Reporter | ||
Comment 15•11 years ago
|
||
I haven't noticed it happening again. I'll write something here if it does.
Flags: needinfo?(jmuizelaar)
| Assignee | ||
Updated•10 years ago
|
Status: NEW → RESOLVED
Closed: 10 years ago
Resolution: --- → INCOMPLETE
Updated•6 years ago
|
Component: DOM → DOM: Core & HTML
You need to log in
before you can comment on or make changes to this bug.
Description
•