87819 - Transparent GIFs with transparency on all four sides render non-transparent portions against the left and top sides

Status: RESOLVED FIXED

(Core :: Graphics: ImageLib, defect, P2)

Description

[mikol]

With some (but not all) transparent GIFs, mozilla renders the non-transparent
portions of the GIF against the top and left edges instead of twoards the center
of the GIF as it should. Please see the URL for an example.

Comment 1

[mikol]

Upon further investigation it appears that the sample GIFs provided at the URL
were created with the GIMP. The content layer rect does not extend beyond the
boundaries of the visible portion, which means that there is a a wide border of
pixels that are both transparent and not part of the content layer (as GIMP sees
it). If I set the content layer rect to be the same size as the image, the new
GIF will render the transparency as expected.

Comment 2

[James Paige]

Could be related to Bug 86508

That bug tracks a similar symptom in the first frame of animating gifs.

Comment 3

[Diego Biurrun]

Confirming on Linux, build 2001062606 and 0.9.1 Netscape 4.77 renders it correctly.

Comment 4

[Dwayne]

Confirming on build 2001-06-22-04, Win98

OS Should be changed to all.

Comment 5

[Diego Biurrun]

Setting OS to ALL as per confirmation.

Comment 6

[anthonyd]

over to pavlov, image lib boy (er man).

Comment 7

[James Paige]

I was hoping that the patch for bug 86508 was going to fix this too, but it
appears it has not :(

Comment 8

[ArronM (:paper)]

Attachment: Fix for 1 frame images w/offset > 0

The patch for Bug 86508 fixed the problem for animated gifs.  This one will fix
it for non-animated ones. :)

Comment 9

[Diego Biurrun]

Your patch fixes the problem for me on Linux.  My build is very crash-prone,
though.  I'll keep testing, no idea if this is related, probably not.

Pavlov, can you r= this?

Comment 10

[ArronM (:paper)]

Attachment: Revision 2: Added Division By 0 Checking

There were some seemingly random times that DrawScaledImage's sr.width would be
0, which caused a division by 0.  This updated patch fixes this.

Comment 11

[Diego Biurrun]

Second patch also works for me, crashes are gone, though I still believe this is
unrelated.

Pavlov, can you give r= ?

Comment 12

[ArronM (:paper)]

Attachment: Rev 3; Corrected for partial image draws

Rev 2 didn't work well when the image was partially on the screen, and one
scrolled to see more of the image.

Comment 13

[Diego Biurrun]

Third patch works fine here.  Somebody needs to have a look and review....

Comment 14

[ArronM (:paper)]

Attachment: Another example

Happy face.  The face should start at 400x200, and extend to 414x214.

The gif is a one frame image, size 14x14, starting at 400x200.

Comment 15

[Diego Biurrun]

Someone, anyone, please review this. Arron M, did you try the necessary steps to
get a review for this?  These URLs may help you:

http://www.mozilla.org/hacking/
http://www.mozilla.org/hacking/code-review-faq.html
http://www.mozilla.org/hacking/reviewers.html

Feel free to contact me for assistance.

Comment 16

[ArronM (:paper)]

If you count buging pavlov (the only active imglib reviewer) every other day to
r=, yes.  But now he's gone for 2 (?) weeks, so I'm going to try it backwards
and see if I can get an SR first

Comment 17

[Diego Biurrun]

Yes, bugging Pavlov counts.  Getting r= and sr= is the worst part of this whole
Mozilla thing.  I'm currently in the loop again with bug 46490 and boy, do you
need patience.  In mails always mention that you have already been ignored n
times, seems like few are willing to review the first time around.  Ask around
on IRC (#mozillazine, #mozilla) who else might be suitable to review your work
and try to get that sr= in parallel.  I wish you luck and patience, you will
need it.

Comment 18

[Stuart Parmenter]

Comment on attachment 82433 [details] [diff] [review]
Rev 3; Corrected for partial image draws

r=pavlov

Comment 19

[tor]

Comment on attachment 82433 [details] [diff] [review]
Rev 3; Corrected for partial image draws

sr=tor

Comment 20

[ArronM (:paper)]

This will not make 1.1 (too close to branching time)

Comment 21

[Christian :Biesinger (don't email me, ping me on IRC)]

checked in.