Wine test analysis August 28th

[Jeremy White]

Hi Folks,

Good news - we've made progress!  We're down to only 25 issues.

I have now annotated every single failure:
   https://www.winehq.org/~jwhite/5aeb834dd4e7.html
(although some of my annotations are out of date).

It looks like the biggest win will come if Francois and Hans can figure 
out the msi timing stuff.

I think there are a number of ones that might be 'easy':
   http://test.winehq.org/data/tests/gdi32:font.html
    Bug 35573, 35760, fails in certain locales

http://test.winehq.org/data/tests/mmdevapi:capture.html
     I asked Jorg, but he may have a hard time getting to it.  It's nice 
and consistent...

http://test.winehq.org/data/tests/shell32:shlexec.html
     Nice and consistent; asked Misha to look into it, but he can't.

http://test.winehq.org/data/tests/user32:sysparams.html
     Nice and consistent; I'd invited Frederic to work on this one, but 
looks like he could use help...

Don't be shy, there are plenty of consistent failures to go around!

From there, I think nudging Jacek, Nikolay, and Huw will get us our 
next big set of wins.

Nudge, nudge, nudge.

Cheers,

Jeremy

p.s.  Reminder:  my analysis always shows up at
   http://winehq.org/~jwhite/latest.html

[Francois Gouget]

One of the failures is mshtml:events which has started ahppening on the 
11th of August.

What is mighty strange is that:
* There was no change to dlls/mshtml/tests/events.c around that time and 
  testing with a pre-11th build still shows the failure.
* I did not change the VMs around that time (or any time past the 3rd of 
  August). Also the failures started happening not only on some of my 
  VMware VMs, but also on the Windows 2008 VMs of the TestBot.
* It does not seem that test actually accesses the network though I 
  may be wrong there. Could it be a change on the Wine web server (I 
  found no matching commit in the tools repository).
* The test started failing on Linux (64 bit only) and FreeBSD (32 bit) 
  on the 15th which corresponds to the next commit round.

Testing this Windows 2000 and XP shows that the issue is that the 
iframe_onreadystatechange_complete test triggers a 'You are about to 
view pages over a secure connection' popup dialog which blocks the 
process until it is answered, thus leading to the timeout.

So given that nothing changed in the VM (still starting from the same 
snapshot), why did it start displaying this dialog on the 11th?

Here's a screenshot:
http://fgouget.free.fr/tmp/mshtml_events.png

[Austin English]

On Fri, Aug 29, 2014 at 12:25 PM, Francois Gouget  wrote:
> One of the failures is mshtml:events which has started ahppening on the
> 11th of August.
>
> What is mighty strange is that:
> - There was no change to dlls/mshtml/tests/events.c around that time and
> testing with a pre-11th build still shows the failure.
> - I did not change the VMs around that time (or any time past the 3rd of
> August). Also the failures started happening not only on some of my
> VMware VMs, but also on the Windows 2008 VMs of the TestBot.
> - It does not seem that test actually accesses the network though I
> may be wrong there. Could it be a change on the Wine web server (I
> found no matching commit in the tools repository).
> - The test started failing on Linux (64 bit only) and FreeBSD (32 bit)
> on the 15th which corresponds to the next commit round.
>
> Testing this Windows 2000 and XP shows that the issue is that the
> iframe_onreadystatechange_complete test triggers a 'You are about to
> view pages over a secure connection' popup dialog which blocks the
> process until it is answered, thus leading to the timeout.
>
> So given that nothing changed in the VM (still starting from the same
> snapshot), why did it start displaying this dialog on the 11th?
>
> Here's a screenshot:
> http://fgouget.free.fr/tmp/mshtml_events.png
>
> --
> Francois Gouget               http://fgouget.free.fr/
>                         question = ( to ) ? be : ! be;
>                               -- Wm. Shakespeare

FWIW, I've noticed a hang with mshtml:events under valgrind (and when I
attempted to bisect it, previously working commits failed). I filed a bug,
didn't investigate further:
https://bugs.winehq.org/show_bug.cgi?id=37157

[Indrek Altpere]

Hello

Perhaps the issue is related to the png url redirecting from http to https?

At line 2883 in events.c there is
        run_test(readystate_doc_str, test_onreadystatechange);

run_test(img_doc_str, test_imgload);
Meaning that imgload test runs right after readystatechange test which very likely has finished correctly.

On line 1847 in test_imgload function there is img url http://www.winehq.org/images/winehq_logo_text.png
When opening it, it actually seems to redirect to https://www.winehq.org/images/winehq_logo_text.png

Most likely some winehq.org server setting was changed to force connections to https and IE started complaining?
Does it help if the url is also fixed to be https from the start? Or does the IE expect it to be from HTTP when loading non-remote html?

Regards,

Indrek

From: wine-devel [mailto:[e-mail]] On Behalf Of Austin English
Sent: Friday, August 29, 2014 8:32 PM
To: Francois Gouget
Cc: Jeremy White; Wine Develop
Subject: Re: Wine test analysis August 28th

On Fri, Aug 29, 2014 at 12:25 PM, Francois Gouget <[e-mail]  > wrote:

One of the failures is mshtml:events which has started ahppening on the
11th of August.

What is mighty strange is that:
* There was no change to dlls/mshtml/tests/events.c around that time and
  testing with a pre-11th build still shows the failure.
* I did not change the VMs around that time (or any time past the 3rd of
  August). Also the failures started happening not only on some of my
  VMware VMs, but also on the Windows 2008 VMs of the TestBot.
* It does not seem that test actually accesses the network though I
  may be wrong there. Could it be a change on the Wine web server (I
  found no matching commit in the tools repository).
* The test started failing on Linux (64 bit only) and FreeBSD (32 bit)
  on the 15th which corresponds to the next commit round.

Testing this Windows 2000 and XP shows that the issue is that the
iframe_onreadystatechange_complete test triggers a 'You are about to
view pages over a secure connection' popup dialog which blocks the
process until it is answered, thus leading to the timeout.

So given that nothing changed in the VM (still starting from the same
snapshot), why did it start displaying this dialog on the 11th?

Here's a screenshot:
http://fgouget.free.fr/tmp/mshtml_events.png

--
Francois Gouget <[e-mail]  >              http://fgouget.free.fr/
                        question = ( to ) ? be : ! be;
                              -- Wm. Shakespeare

FWIW, I've noticed a hang with mshtml:events under valgrind (and when I attempted to bisect it, previously working commits failed). I filed a bug, didn't investigate further:
https://bugs.winehq.org/show_bug.cgi?id=37157

[Francois Gouget]

On Fri, 29 Aug 2014, Indrek Altpere wrote:
[...]
> On line 1847 in test_imgload function there is img url 
> http://www.winehq.org/images/winehq_logo_text.png
> When opening it, it actually seems to redirect to 
> https://www.winehq.org/images/winehq_logo_text.png
>
> Most likely some winehq.org server setting was changed to force 
> connections to https and IE started complaining?
> Does it help if the url is also fixed to be https from the start? Or 
> does the IE expect it to be from HTTP when loading non-remote html?

Yes, that does the trick. I sent a patch. The Linux 64 bit failures 
still seem to be there though. But from the start they looked like a 
slightly different issue.

[Maarten Lankhorst]

Hey,

On 28-08-14 17:57, Jeremy White wrote:
> Hi Folks,
>
> Good news - we've made progress!  We're down to only 25 issues.
>
> I have now annotated every single failure:
>   https://www.winehq.org/~jwhite/5aeb834dd4e7.html
> (although some of my annotations are out of date).
>
> It looks like the biggest win will come if Francois and Hans can figure out the msi timing stuff.
>
> I think there are a number of ones that might be 'easy':
>   http://test.winehq.org/data/tests/gdi32:font.html
>    Bug 35573, 35760, fails in certain locales
>
> http://test.winehq.org/data/tests/mmdevapi:capture.html
>     I asked Jorg, but he may have a hard time getting to it.  It's nice and consistent...

Seems to be a winealsa specific bug. It works fine if I enable winepulse. But if I set WINETEST_PLATFORM to wine I get 3 failures because of tests succeeding inside todo blocks. ;-) Otherwise 6 failures, of which 3 todo.

~Maarten