http://bugs.winehq.org/show_bug.cgi?id=11188
--- Comment #97 from Claudio sick_soul@yahoo.it 2009-01-10 16:53:33 --- I finished my regression test. I got:
477582401b9a7b6baf188a788b9972c12495c51a is first bad commit commit 477582401b9a7b6baf188a788b9972c12495c51a Author: Andrey Turkin andrey.turkin@gmail.com Date: Tue Sep 18 00:00:45 2007 +0400
server: Implement server-side completion queues and operations on them.
:040000 040000 ab2c6bf45138e22c62cfafcaadcefd57bd9407d1 7d4520d246d44a63d09bcae4e29033dc05a2955d M dlls :040000 040000 482b39d1eff431320098ac137f8c99822b08d215 062636a29d0853fcce810e7530d0414d2fab96b1 M include :040000 040000 ce0b55aa8f8649c0d37504e474656b0de8a5b0be 3df6965ca1d3adb561422917cc1455fa79f056f2 M server
I could progress much faster as soon as I realized that I could load a saved game I have and be able to reproduce the problem reliably on the first subsequent save attempt.
During this test I had to change the concept of "BAD revision" a little in progress, as different wine versions react to the problem differently.
In newer wine versions, a failed attempt to save got me the war3 memory protection error dialog.
In older wine versions around 0.9.50 a failed attempt to save caused the process to just block forever.
In even older wine versions, between 0.9.45 and 0.9.46, a failed attempt to save caused a 0 length file to be created instead of the circa 2MB game file, and the game did not crash at all, but instead of turning the text 'green' to indicate success, the dialog text with the game's filename turned Red, and the game went on (but with a failed save).
I'll attach the bisection log, the saved game with which I could reliably reproduce the problem, and the output obtained with WINEDEBUG=+relay of the failed attempt to save. I generated this by attaching gdb to the process just before saving, and then using call putenv("WINEDEBUG=+relay"), otherwise the output generation was simply too huge.