851785 - FrameRect triggers decoding

Status: RESOLVED FIXED

(Core :: Graphics: ImageLib, defect)

Description

[Timothy Nikkel (:tnikkel)]

We trigger a decode for basically every image we download because imgStatusTracker::SyncNotify calls FrameRect on the image, which asks for decoded frames, which triggers a decode. This is the only caller of FrameRect, so it shouldn't be a problem to make FrameRect not ask for decoded frames.

Comment 1

[Timothy Nikkel (:tnikkel)]

This causes some of our async_load_tests.js tests to fail for a very simple reason: without this inadvertent decode we don't trigger a decode at all in these tests. But checkAllCallbacks wants us to be decoded fully.

Should we insert a requestDecode in this test or just not check if the image is decoded?

Comment 2

[Timothy Nikkel (:tnikkel)]

(In reply to Timothy Nikkel (:tn) from comment #1)
> But checkAllCallbacks wants us to be decoded fully.

We only check this on a static request. So maybe we only have these expectations of being fully decoded on static requests? Before bug 843895 that was always true because static requests used ExtractFrame with sync decode, but after bug 843895 we just wrap them in a wrapper that always returns the first frame.

Comment 3

[Timothy Nikkel (:tnikkel)]

Talked to Joe on irc about this.

Comment 4

[Timothy Nikkel (:tnikkel)]

Attachment: fix aync_load_tests.js to not expect a decode

Comment 5

[Timothy Nikkel (:tnikkel)]

Attachment: stop from asking for a decode in FrameRect

Comment 6

[Timothy Nikkel (:tnikkel)]

Attachment: stop from asking for a decode in FrameRect

Comment 7

[Timothy Nikkel (:tnikkel)]

https://hg.mozilla.org/integration/mozilla-inbound/rev/9502b3cccfe3
https://hg.mozilla.org/integration/mozilla-inbound/rev/25c76053d44a

Comment 8

[Ryan VanderMeulen [:RyanVM]]

https://hg.mozilla.org/mozilla-central/rev/9502b3cccfe3
https://hg.mozilla.org/mozilla-central/rev/25c76053d44a