123401 - image loader should cache ioservice.

Status: RESOLVED FIXED

(Core :: Layout, defect, P3)

Description

[Doug Turner (:dougt)]

Comment 1

[Doug Turner (:dougt)]

Attachment: caches io service and img loader

Comment 2

[Stuart Parmenter]

Comment on attachment 67776 [details] [diff] [review]
caches io service and img loader

r=pavlov

Comment 3

[Boris Zbarsky [:bzbarsky]]

doug, is this patch still relevant?  We no longer need the ioservice at all in
this file....

Comment 4

[Doug Turner (:dougt)]

if you say so, mark it as invalid.  I remember that the image loader, at one
point, smacked the service manager pretty hard for these two objects.

Comment 5

[Boris Zbarsky [:bzbarsky]]

Yeah, at this point the prescontext is what would be smacking the service
manager for the ioservice.  But it already has a cached on it could use... patch
coming up for that.

And I'd be interested in how much both nsImageLoader and nsImageLoadingContent
hit the service manager looking for the image loader...  We could conceivably
cache the image loader in those two places...

Comment 6

[Boris Zbarsky [:bzbarsky]]

Attachment: use the cached ioservice in the prescontext

Comment 7

[Doug Turner (:dougt)]

Comment on attachment 121621 [details] [diff] [review]
use the cached ioservice in the prescontext

r=dougt

assuming that this doesn't cause any nasty cycles.  Please check this by using
the reference count log.

Comment 8

[Boris Zbarsky [:bzbarsky]]

Doug, I did check... note that we're _already_ caching the IOService here; all
I'm doing is passing it to NS_NewURI.

Comment 9

[Doug Turner (:dougt)]

i didn't see any calls to GetIOService in nsPresContext -- is is probably called
via the nsIPresContext interface.

Comment 10

[Boris Zbarsky [:bzbarsky]]

Yep.

Comment 11

[Boris Zbarsky [:bzbarsky]]

Patch checked in.  Marking fixed.