[Webkit-unassigned] [Bug 26379] Reconsider image decoding architecture/APIs

[bugzilla-daemon at webkit.org]

https://bugs.webkit.org/show_bug.cgi?id=26379





------- Comment #33 from pkasting at google.com  2009-06-14 23:13 PDT -------
(In reply to comment #27)
>           (as the image data was not copied and the cairo_image_surface was
> created exactly once, at least to what I think is right)

BTW Brent, I think this is correct, and may actually be a bug.  From my reading
of some of the Cairo docs today, cairo_image_surface_create_for_data() doesn't
copy the data, which means your returned surface is at the merge of the
lifetime of the RGBA32Buffer -- which is managed by the ImageDecoder, which may
delete it (see e.g. clear()).  I don't think it's safe to have the frames
returned from the ImageSource depend on the vagaries of the ImageDecoder.  I
assumed this code was copying since wx copies, Skia refs, and when we used to
have a different backing implementation here, we had all kinds of problems
until we ensured that the lifetimes were separate.

So you may want to see if you can write this in a way that allocates the buffer
in some kind of a refcounted object.


-- 
Configure bugmail: https://bugs.webkit.org/userprefs.cgi?tab=email
------- You are receiving this mail because: -------
You are the assignee for the bug, or are watching the assignee.