409162 - Messed background with --enable-system-cairo

Status: RESOLVED DUPLICATE of bug 413583

(Core :: Graphics, defect)

Description

[ojab]

User-Agent:       Mozilla/5.0 (X11; U; Linux i686; en-US; rv:1.9b3pre) Gecko/2007121912 Minefield/3.0b3pre
Build Identifier: Mozilla/5.0 (X11; U; Linux i686; en-US; rv:1.9b3pre) Gecko/2007121912 Minefield/3.0b3pre

In builds with --enable-system-cairo it is absolutely messed background on some (as far as I understand, rather long) pages and even Page source window.
I've tried with cairo-1.5.4 & gairo-git.


Reproducible: Always

Steps to Reproduce:
1. Go to http://labs.mozilla.com/2007/10/prism/ page for example
Actual Results:  
Messed background

Expected Results:  
normal background

about:buildconfig

--disable-installer --disable-accessibility --enable-application=browser --disable-dbus --prefix=/usr --enable-system-cairo --with-system-jpeg --with-system-zlib --enable-native-ucon --enable-image-decoders=all --enable-places --enable-storage --enable-safe-browsing --enable-libxul --enable-system-lcms --enable-system-hunspell --enable-system-sqlite3 --enable-image-encoder=all --enable-canvas --enable-oji --enable-mathml --enable-places_bookmarks --disable-tests --disable-debug --disable-pango

Comment 1

[ojab]

Attachment: --with-system-cairo

screenshot from --with-system-cairo build

Comment 2

[ojab]

Attachment: with bundled cairo

screenshot from build without --with-system-cairo option

Comment 3

[ojab]

Maybe this bug is really invalid, but what should I done with my broken firefox build? :)

I can provide you with any additional info for investigation, just tell me what exactly do you need.

Comment 4

[Frank Wein [:mcsmurf]]

I'm not sure, but I think enable-system-cairo should only be used if you really know what you are doing. The cairo version we use is a cairo version with some special patches in it. Also see http://lxr.mozilla.org/seamonkey/source/gfx/cairo/README for what version of cairo we currently use.

Comment 5

[ojab]

So in general it shouldn't works correctly with system libs?

Comment 6

[Stuart Parmenter]

I'd generally mark these bugs invalid but we should work with current git versions of cairo and probably 1.5.4.  You might take a look and see if we have any patches that need to be pushed upstream in to cairo.  It would be good to work out if something broke between our current in-tree version and git.  I expect vlad will update our in-tree cairo soon.

Comment 7

[ojab]

after rm -rf mozilla  building from scratch all works fine.

Comment 8

[ojab]

oops. No. Now upper part of the page looks good, but background in the bottom (from "The User Experience" title and below) is still messed.

Comment 9

[Brandon Wright]

I can confirm this. It seems the cause is related to bug 413583.

Comment 10

[Brandon Wright]

(In reply to comment #9)
> I can confirm this. It seems the cause is related to bug 413583.

It also appears the patch to change the ratio of the fixed type (fixed-24-8.patch at http://lxr.mozilla.org/seamonkey/source/gfx/cairo/README) has a severe effect. Pages greater than 65535 pixels in width/length will cause an overflow or error and cairo won't handle the background. The mozilla patch increases the whole integer component of fixed to 24 bits.

The default cairo libraries on most systems won't be changed, as I imagine there are many programs that need 16 bits of decimal precision. The only real solution here is fixing the rendering backend to prevent needing the ratio change.

Comment 11

[Brandon Wright]

In my testing of the issue, the fixed type is actually signed, so the overflow happens at 32767px. I wasn't able to reproduce this with images, but using a <div> tag with a CSS height of 32800 pixels will cause this error, but 32700 pixels won't.

Comment 12

[Vladimir Vukicevic [:vlad] [:vladv] (needinfo me, slow to respond)]

(In reply to comment #10)
> The default cairo libraries on most systems won't be changed, as I imagine
> there are many programs that need 16 bits of decimal precision. The only real
> solution here is fixing the rendering backend to prevent needing the ratio
> change.

That is incorrect; cairo 1.6 will ship with 24.8 as the default fixed point format.  Firefox will require cairo 1.6 to build with system cairo under linux.  Whether that version of cairo has the workaround for the buggy_repeat issue is up to the vendors -- if they are able to fix the X server bug, then their version of cairo should not have the workaround.  So, this bug as it stands is invalid, but it's technically a dup of bug 413583, so marking as such.