554772 - 2 reftest failures on sparc linux

Status: UNCONFIRMED

(Core :: Layout, defect)

Description

[Mike Hommey [:glandium]]

Attachment: 423385-1

I got 2 failures in reftests on sparc linux:
reftest> REFTEST TEST-UNEXPECTED-FAIL | file:///build/buildd-xulrunner_1.9.1.8-6-sparc-VDHGBV/xulrunner-1.9.1.8/layout/reftests/bugs/423385-1.html |
reftest> REFTEST TEST-UNEXPECTED-FAIL | file:///build/buildd-xulrunner_1.9.1.8-6-sparc-VDHGBV/xulrunner-1.9.1.8/layout/reftests/bugs/425247-2.html |

I'm attaching the corresponding png files. It appears the non -ref version has unexpected transparency set. What is strange, too, is that 425247-1.html doesn't fail.

I'll try to see if the failures are random or consistent.

If someone could point me at where in the code to look for something fishy, that would be appreciated.

Comment 1

[Mike Hommey [:glandium]]

Attachment: 423385-1-ref

Comment 2

[Mike Hommey [:glandium]]

Attachment: 425247-2

Comment 3

[Mike Hommey [:glandium]]

Attachment: 425247-2-ref

Comment 4

[Mike Hommey [:glandium]]

I can't reproduce this failure "by hand" (i.e. connecting to a sparc machine and exporting a remote X display) :-/

I will try with the same environment as when we got the failure (with xvfb)

Comment 5

[Mike Hommey [:glandium]]

It unfortunately happens on random tests :(

On the other hand, considering comment 4 on bug 554774, this could be a problem with xvfb.

Comment 6

[Mike Hommey [:glandium]]

It appears that contrary to bug 554774, the problem is not on xvfb end. I can reproduce some reftest failures with a remote X display. The strange thing is that the failing reftests vary from time to time. Sometimes, failures start as soon as layout/reftests/reftest-sanity/test-async.html, sometimes a bit later, at layout/reftests/reftest-sanity/blank.html, and a lot more tests now fail than the origin 2 I gave. Though at the second try, only 3 were failing...