368157 - [X11] DrawTile and canvas drawWindow don't get along

Status: RESOLVED WORKSFORME

(Core :: Graphics, defect, P4)

Description

[David Baron :dbaron: (⌚️UTC-5, no longer working on Mozilla)]

I have a testcase that works fine in the browser, works fine in svg:foreignObject, but when displayed in reftest (which uses canvas and drawWindow), the tiled background image ends up showing garbage.

The difference between it and the testcase that works fine is that in the testcase that works fine (for which it happens to be the reference), the code in nsCSSRendering::PaintBackgroundWithSC is able to use DrawImage instead of DrawTile.

The testcase in question is layout/reftests/bugs/368020-2-ref.html , which is part of the patch in bug 368020.

Steps to reproduce:

1. If the patch to bug 368020 hasn't landed yet, get the files 368020-2.html and 368020-2-ref.html out of it and put them in layout/reftests/bugs/

2. Create a file layout/reftests/debug-painting.list containing the line:
== bugs/368020-2.html bugs/368020-2-ref.html

3. run the command "./firefox [-profile <dir>] -reftest /path/to/srcdir/layout/reftests/debug-painting.list" and redirect the output to a file.

Expected results:  "LAYOUTATEST PASS" contained in file

Actual results: "LAYOUTATEST UNEXPECTED FAIL" contained in file.

[note: I plan to change "LAYOUTATEST" to "REFTEST", just haven't gotten around to it.  I missed it when renaming before checkin.]

Below the "UNEXPECTED FAIL" message will be two data URLs for PNG images showing the test and reference.  In these images, I see that the background for the reference rendering shows pieces of garbage from my GNOME desktop or toolbars.

I see this in a trunk build from yesterday or today on Fedora Core 6 with ati-fglrx video drivers.

Comment 2

[Boris Zbarsky [:bzbarsky]]

Um.. wrong textbox.  :(

Comment 3

[David Baron :dbaron: (⌚️UTC-5, no longer working on Mozilla)]

Yeah, 368020-2 passes on Windows.

Comment 4

[David Baron :dbaron: (⌚️UTC-5, no longer working on Mozilla)]

Hrm, seems like this might be fixed by my patch on bug 371434.  Not sure why, though, and need to look at the images to see if it's really fixed.  It could be that I broke both now.

Comment 5

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

The tinderbox is saying that test bugs/368020-2 is passing after my checkin of bug 368247, though sayrer says it continues to fail on his local build.  I've changed that reftest to be random on gtk2 instead of fails on gtk2 for now.

Comment 6

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

368020-2 is still passing on tinderboxes; this is going to end up being an X/driver issue.  I don't think it's worth trying to do a workaround, because drawWindow isn't frequently used.