On Wed, Jul 4, 2012 at 12:15 AM, Andrea Canciani ranma42@gmail.com wrote:
Do you think you could write a test that (semi-)reliably causes the deadlock you're fixing?
Yes, I guess it should be possible. Could you point me to a test which triggers a deadlock (in particular, how should the deadlock be handled?)
http://bugs.winehq.org/show_bug.cgi?id=28362 says that the mshtml tests hang occasionally. The testsuite handles them by, well, hanging :-) Scripts that care about this add their own timeout.
Yes, I sent the e-mail just before finding out about this issue. I will probably need to set up a proper wine development environment to find out more about it (I have a ubuntu 12.04 x86 vm, but even wine/master fails the testsuite on it).
The testsuite does not, in general, fully pass on any machine but Alexandre's, so just generate a baseline by running the tests several times with "make -k test" without your patch, and saving the logs. Then repeat after applying your patch.
(https://code.google.com/p/winezeug/source/browse/trunk/buildbot/dotests_blac... has a list of all the tests that the buildbot failed on while it was running, and you can run just the known good or known bad tests with https://code.google.com/p/winezeug/source/browse/trunk/buildbot/dotests.sh but that's overkill (or underkill) for you, and slightly out of date anyway.) - Dan