http://bugs.winehq.org/show_bug.cgi?id=24521
Matt Helsley mhelsley@linux.ucla.edu changed:
What |Removed |Added ---------------------------------------------------------------------------- CC| |mhelsley@linux.ucla.edu
--- Comment #6 from Matt Helsley mhelsley@linux.ucla.edu 2010-10-03 01:42:59 CDT --- So finally hit this bug with Civ. Knowing it existed I took a look at the Civ V process using lsof to see which files it had open. It turns out it has the same directory open via about 600 file descriptors. So it has all the appearances of a file descriptor "leak" in one of the API implementations. The offending directory looks like this in lsof:
Civilizat 32577 me 160r DIR 8,33 4096 789724 /home/me/.wine/drive_c/windows/system32/mui
And the command to figure this out is:
lsof -p <pid of CivilizationV.exe> | grep system32/mui | wc -l
which reports 598 of the 1025 open files (using the standard 1024-fd limit) referencing that directory. I skimmed the rest of the open files and they are:
Libraries (.so, .dll) Fonts Sound files Game data packs (e.g. .fpk) FIFOs (There were quite a few! I can't tell if this is how it's supposed to work or also a symptom..) a .db file or two
But mostly that directory.
Other details: The whole Civ screen freezes *except* the cursor. Even kill -SIGSTOP doesn't "stop" the cursor. Switching VTs just made the whole screen except the cursor black. Given what little I do know about X+wine I doubt this latter information about the screen contents is pertinent -- I'm including it just in case I'm wrong.