Wine test analysis August 28th
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annotated every single failure: https://www.winehq.org/~jwhite/5aeb834dd4e7.html (although some of my annotations are out of date).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
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 <fgouget(a)free.fr> http://fgouget.free.fr/ question = ( to ) ? be : ! be; -- Wm. Shakespeare
On Fri, Aug 29, 2014 at 12:25 PM, Francois Gouget <fgouget(a)free.fr> 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 <fgouget(a)free.fr> 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 -- -Austin
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:wine-devel-bounces(a)winehq.org] 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 <fgouget(a)free.fr <mailto:fgouget(a)free.fr> > 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 <fgouget(a)free.fr <mailto:fgouget(a)free.fr> > 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 -- -Austin
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. -- Francois Gouget <fgouget(a)free.fr> http://fgouget.free.fr/ "Lotto: A tax on people who are bad at math." -- unknown "Windows: Microsoft's tax on computer illiterates." -- WE7U
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
participants (5)
-
Austin English -
Francois Gouget -
Indrek Altpere -
Jeremy White -
Maarten Lankhorst