374254 - need tests running on Vista machine

Status: RESOLVED INCOMPLETE

(Release Engineering :: General, defect, P3)

Description

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

I'm noticing a bunch of win32 reftest failures because of changes in win32 rendering, and potentially in configuration changes... for example, this simple test:

span { display: inline-block; }
<p>a<span>b</span>c</p>   vs. <p>abc</p>

fails on Vista, because one column of antialiasing on the 'b' in the first case gets cut off.

Comment 1

[Rob Campbell [:rc] (:robcee)]

we should probably extend that to include mochitest and make check for the sake of completeness. I'll put in the IT request to get a machine spun up.

Comment 2

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

Hmm, so my issue was actually caused by me having Cleartype enabled.  Disabling it fixes things.  This is tricky, because I don't know what to do here -- getting things exactly right with cleartype enabled is bound to be hard, since the win32 text rendering stuff seems to cheat a bit.

Comment 3

[Rob Campbell [:rc] (:robcee)]

I still think we should have a Vista machine for checking these things. Just out of curiosity, did anyone try running this test on a WinXP machine with cleartype enabled?

Comment 4

[John O'Duinn [:joduinn] (please use "needinfo?" flag)]

in build triage meeting, we:

- clarified subject to be about running tests on a  Vista machine containing the new win ref platform toolset
- reassigned to QA

Comment 5

[John O'Duinn [:joduinn] (please use "needinfo?" flag)]

Here's a pointer to some docs that schrep is working on, which might be helpful: 

http://wiki.mozilla.org/ReferencePlatforms/Test/Vista

Comment 6

[Chris Pearce [:cpearce (Not reading bugmail)]]

Did a Vista machine get setup and/or tested? I've been having Mochitests toolkit/content/tests/widgets/test_scale.xul and
dom/tests/mochitest/ajax/scriptaculous/ fail on my Vista laptop. Scriptaculous is failing spectacularly with 810 of 814 tests failing. The reftests pass on my Vista machine though.

It's also worth noting that the mochitests take longer to run on Vista than on WinXP.

Comment 7

[Rob Campbell [:rc] (:robcee)]

No, this hasn't been setup yet. We had a VM installed but it disappeared again. I'll bump this up to get it running.

Comment 8

[(not currently active) Ted Mielczarek]

Fixing bug 421611 would probably make this a lot easier.

Comment 9

[John O'Duinn [:joduinn] (please use "needinfo?" flag)]

(In reply to comment #2)
> Hmm, so my issue was actually caused by me having Cleartype enabled.  Disabling
> it fixes things.  This is tricky, because I don't know what to do here --
> getting things exactly right with cleartype enabled is bound to be hard, since
> the win32 text rendering stuff seems to cheat a bit.

If this is still needed, should we:
- update build+unittest refplatform to turn off ClearType? 
...or...
- setup pool of vista machines to run unittests as a newly supported UnitTest platform?

Comment 10

[John O'Duinn [:joduinn] (please use "needinfo?" flag)]

(In reply to comment #9)
> (In reply to comment #2)
> > Hmm, so my issue was actually caused by me having Cleartype enabled.  Disabling
> > it fixes things.  This is tricky, because I don't know what to do here --
> > getting things exactly right with cleartype enabled is bound to be hard, since
> > the win32 text rendering stuff seems to cheat a bit.
> 
> If this is still needed, should we:
> - update build+unittest refplatform to turn off ClearType? 
> ...or...
> - setup pool of vista machines to run unittests as a newly supported UnitTest
> platform?


Vlad: is this still needed?

Comment 11

[John O'Duinn [:joduinn] (please use "needinfo?" flag)]

(In reply to comment #10)
> (In reply to comment #9)
> > (In reply to comment #2)
> > > Hmm, so my issue was actually caused by me having Cleartype enabled.  Disabling
> > > it fixes things.  This is tricky, because I don't know what to do here --
> > > getting things exactly right with cleartype enabled is bound to be hard, since
> > > the win32 text rendering stuff seems to cheat a bit.
> > 
> > If this is still needed, should we:
> > - update build+unittest refplatform to turn off ClearType? 
> > ...or...
> > - setup pool of vista machines to run unittests as a newly supported UnitTest
> > platform?
> 
> 
> Vlad: is this still needed?

Please reopen if needed.

Comment 12

[Chris Cooper [:coop] (he/him)]

Moving closed Future bugs into Release Engineering in preparation for removing the Future component.