859996 - stat() called on fonts during shutdown

Status: RESOLVED DUPLICATE of bug 833087

(Core :: Graphics: Text, defect)

Description

[Gregory Szorc [:gps]]

Attachment: fs_usage output at shutdown

I'm not sure if this is the correct component, but I was running Nightly under fs_usage() (an OS X utility). One of the things I noticed was we stat64() a bunch of font files pretty late during shutdown. I have no clue why. Surely this isn't needed and adversely affects shutdown times.

The attachment is fs_usage output at shutdown time showing the issue.

I can reproduce this issue every time I shut down Nightly.app.

Comment 1

[Gregory Szorc [:gps]]

Taras says BenWa is fixing this. I assume it will be duped.

Comment 2

[Jonathan Kew [:jfkthame]]

I doubt that's something we're doing directly; at a guess, it's probably something Cocoa is doing as part of shutting down. But to get a better idea than just my blind guess, could you maybe try it under gdb with a break on stat64() and get a stack for where those calls actually come from?

Comment 3

[Jonathan Kew [:jfkthame]]

(In reply to Gregory Szorc [:gps] from comment #1)
> Taras says BenWa is fixing this. I assume it will be duped.

Is that a reference to bug 833087 (in which case it would only apply if the browser hasn't been running very long), or something else?

Comment 4

[Benoit Girard (:BenWa)]

(In reply to Jonathan Kew (:jfkthame) from comment #3)
> (In reply to Gregory Szorc [:gps] from comment #1)
> > Taras says BenWa is fixing this. I assume it will be duped.
> 
> Is that a reference to bug 833087 (in which case it would only apply if the
> browser hasn't been running very long), or something else?

That's correct the only thing similar I'm working on is that bug. Did you leave the browser running for a few minutes. We still have a few cases of timers not being canceled on shutdown with Fonts being the biggest offender I found.

Comment 5

[Gregory Szorc [:gps]]

My measurement was from a short-lived profile (I'm guessing under a minute).

Comment 6

[Jonathan Kew [:jfkthame]]

OK, in that case I think we can dupe this to 833087.