http://bugs.winehq.org/show_bug.cgi?id=27158
3vi1 winehq.org@eternaldusk.com changed:
What |Removed |Added ---------------------------------------------------------------------------- CC| |julliard@winehq.org, | |winehq.org@eternaldusk.com
--- Comment #10 from 3vi1 winehq.org@eternaldusk.com 2011-05-15 12:28:08 CDT --- Wish I had thought to search on some different terms this morning. I would have got here sooner and saved myself a bit of regression testing. :)
At any rate, I too tracked it to http://source.winehq.org/git/wine.git/commit/8e9b4e0a5c6bee5f72ca2314db26233... and verified that reverting that patch seems to correct the main problem.
I wish I could help fix it, but knowing less than nothing about XInput1/2 I don't entirely follow the intention of the original patch yet. The only thing really clear is that each time the one-second minimum is passed, it's merging input that should have been ignored.