150881 - scaled transparent png badly rendered

Status: RESOLVED FIXED

(Core :: Graphics: ImageLib, defect, P5)

Description

[Radek Doulik]

Page with such tables is incorrectly rendered, and when you scroll down, it's
visible even more. try to browse
http://primates.ximian.com/~rodo/obrazky/aquarium/page-1.html

Comment 1

[Radek Doulik]

happens with

Mozilla 1.0
Mozilla/5.0 (X11; U; Linux ppc; en-US; rv:1.0.0) Gecko/20020606

Comment 2

[Radek Doulik]

Attachment: example of incorrect rendering

Comment 3

[riscky]

just wondering what is the problem... what or where badly rendered area.

Comment 4

[Radek Doulik]

Riscky: see the attached screenshot, image shadows are wrong (they are not part
of the image, image is without them and they are added as background images of
table cells, where some of them have col and row span > 1)

Comment 5

[Radek Doulik]

Any news on this? Why do you keep it unconfirmed? Don't you experience the same
problem on http://primates.ximian.com/~rodo/obrazky/aquarium/page-1.html page?

Comment 6

[Simón Medrano]

Attachment: Incorrect rendering of PNG images with partial transparency on Windoze.

Has something to do with the transparency of the .PNG images, not the tables.
But it looks uglier on my Windoze macchine.

Comment 7

[Bernd]

> Any news on this? Why do you keep it unconfirmed?
One of the reasons is that it does not have a reduced testcase. 

Comment 8

[Andrew Schultz]

Attachment: testcase

this is just the shadow_e.png from the URL scaled up.  make the browser window
small enough to scroll and bad things will happen (as in the previous
screenshots)

Comment 9

[Andrew Schultz]

this is probably a dupe of bug 146202, except that the URL there changed, and so
I don't really know what the problem was.  See also bug 75558, which describes
somewhat similar stuff, but it worksforME.

marking NEW
==> Imglib

Comment 10

[Alfred Kayser]

Retest, as it worksforme, 20030408080, W2K.
Can probably be closed.

Comment 11

[Andrew Schultz]

testcase still exhibits the bug with linux trunk 20030412
might be linux-only

Comment 12

[tor]

Taking bug.

Comment 13

[tor]

Attachment: patch for linux composited image partial updates

Fix for the problem on linux.  This patch also affects the beos, xlib, and
xprint gfx platforms, none of which I'm able to test at the moment.  The
respective module owners should review/test the patch for their platforms.

Comment 14

[Sergei Dolgov]

Honestly, i don't see this bug in my development builds.
Maybe timeless can confirm if it exists or not in BeOS with standard gfx source.

Comment 15

[tor]

Sergei: BeOS probably wouldn't have exhibited the problem, but the fix
required changing one of the backend functions (RectStretch) that
gfx/src/beos uses.  I'm CCing the other gfx owners so they can check
to see if my untesting changes to their code break things.

Comment 16

[Sergei Dolgov]

tor
i compiled Mozilla for BeOS with this patch. Seems OK (as it was before).
But maybe there are some testcase for image scaling where something may be
broken with this patch? To be absolutely sure.

Comment 17

[tor]

Attachment: same as previous plus win32 fix

Win32 was doing scaled composited images like this (also the OS/2 approach):
   1) read screen region and scale to size of image
   2) composite image
   3) scale back to screen region and paint

While this works for shrinking images, it gives horrible artifacts in the
background when scaling up.  This patch fixes that, though still shows some
artifacts when scrolling the test url for this bug.

Comment 18

[tor]

Attachment: plug memory leak in win32 path

Comment 19

[tor]

Attachment: win32: fix problem with unscaled alpha images

Comment 20

[tor]

*** Bug 75558 has been marked as a duplicate of this bug. ***

Comment 21

[tor]

*** Bug 176330 has been marked as a duplicate of this bug. ***

Comment 22

[tor]

Attachment: fix xprint build bustage

Comment 23

[Asa Dotzler [:asa]]

Comment on attachment 135336 [details] [diff] [review]
fix xprint build bustage

a=asa (on behalf of drivers) for checkin to Mozilla 1.6 beta

Comment 24

[tor]

Checked in.  Leaving bug open for os/2 work.

Comment 25

[Sergei Dolgov]

tor,
BeOS build is broken.
RectStretch() 
at line 498 in nsImageBeOS lacks arguments.
(while RectStretch at 478 has full set of them.)

Comment 26

[tor]

BeOS bustage fix checked in.

Comment 27

[Glenn Randers-Pehrson]

The "testcase" attachment above, the test case that I posted recently to bug
#75558, and the "exteeeemely stretched" example attached to bug #75558 all WFM
after the checkin.

Mozilla/5.0 (Windows; U; Win95; en-US; rv:1.6b) Gecko/20031123 Firebird/0.7+
(Mike Tigas (Nova))

Thanks!

Comment 28

[tor]

Attachment: fix os/2 gfx

Comment 29

[Hermann Schwab]

requesting blocking 1.4.2, 
Seems to have fixed bug 220576 on Win98.

Comment 30

[Mike Kaply [:mkaply]]

Comment on attachment 136612 [details] [diff] [review]
fix os/2 gfx

r=mkaply on patch
sr=blizzard
a=mkaply for 1.6
a=mkaply for 1.4.2 (I'm doing an OS/2 specific safer patch for that)

Comment 31

[Mike Kaply [:mkaply]]

all is well.

Everything checked in.

Comment 32

[Hermann Schwab]

re comment 30:
> a=mkaply for 1.4.2 (I'm doing an OS/2 specific safer patch for that)
Does this mean it will be fixed in 1.4.2 for os/2 only?
If there will come a 1.4.2, will it also be for other OS than OS/2, at least
Linux and Windows?

Comment 33

[Mike Kaply [:mkaply]]

Does anyone feel a need to fix this on 1.4.2 for other platforms?

The reason I want to fix it for OS/2 is because we are still doing important
1.4.2 releases.

Comment 34

[Ben Bucksch (:BenB)]

I need to fix this for Beonex (or rather a customer) on the 1.4 branch, but we
can do that in our own 1.4-based branch.

Comment 35

[tor]

*** Bug 231974 has been marked as a duplicate of this bug. ***

Comment 36

[tor]

Checked in on 1.4 branch - verbal approval from mkaply.

Comment 37

[Roland Mainz]

This may have caused bug 250786 ("scaled gifs render as garbage") ...