http://bugs.winehq.org/show_bug.cgi?id=17421
Summary: Regression: Diablo2 fails to run when Wine is compiled with -march=amdfam10 Product: Wine Version: 1.1.15 Platform: PC-x86-64 OS/Version: Linux Status: UNCONFIRMED Severity: normal Priority: P2 Component: -unknown AssignedTo: wine-bugs@winehq.org ReportedBy: liquitsnake@gmx.net
Since Wine 1.1.13, when it is compiled -march=amdfam10 in CFLAGS, Diablo 2 fails to run. When attempting to log into BattleNet, I am greeted by the error message box: "Diablo II Exception. UNHANDLED EXCEPTION: ACCESS_VIOLATION (c0000005)" No suspicious console log is present, but I didn't enable any special debug channels.
Wine 1.1.12 worked fine. I will do a regression test later today
I am aware that some will not consider using some "obscure" CFLAGS a bug. However, this march was recently introduced and should not be considered obscure. I also have encountered other problems in other games starting with 1.1.13, but I didn't investigate further into those. I suspect they may be related. Bug 14348 seems to have the same issue, but uses normal precompiled packages. For that reason, I'm opening a new bug, since I doubt that these bugs are directly related. Feel free to close as a dup if this turns out to be wrong.
Gentoo Linux on AMD64 GCC 4.3.2, glibc 2.8, Wine optimization level -O3 same results with Gentoo's ebuild aswell as custom compile from git-1.1.15
http://bugs.winehq.org/show_bug.cgi?id=17421
haarp liquitsnake@gmx.net changed:
What |Removed |Added ---------------------------------------------------------------------------- Keywords| |regression
http://bugs.winehq.org/show_bug.cgi?id=17421
--- Comment #1 from haarp liquitsnake@gmx.net 2009-02-16 08:01:23 --- Oh, and for the record: 1.1.13+ works fine when compiled with another march, like k8-sse3 for example.
http://bugs.winehq.org/show_bug.cgi?id=17421
--- Comment #2 from haarp liquitsnake@gmx.net 2009-02-16 08:42:06 --- Disregard the last comment. I had to completely remove the march argument for it to work.
http://bugs.winehq.org/show_bug.cgi?id=17421
Dmitry Timoshkov dmitry@codeweavers.com changed:
What |Removed |Added ---------------------------------------------------------------------------- Summary|Regression: Diablo2 fails to|Diablo2 fails to run when |run when Wine is compiled |Wine is compiled with - |with -march=amdfam10 |march=amdfam10
http://bugs.winehq.org/show_bug.cgi?id=17421
haarp liquitsnake@gmx.net changed:
What |Removed |Added ---------------------------------------------------------------------------- CC| |robertshearman@gmail.com
--- Comment #3 from haarp liquitsnake@gmx.net 2009-02-18 00:11:39 --- Ok, I tracked the problem down to actually being between 1.1.6 and 1.1.7. Regression testing points me to this:
0a2efbb1ccb73da5943baa204aad670ed9b9c9a0 is first bad commit commit 0a2efbb1ccb73da5943baa204aad670ed9b9c9a0 Author: Rob Shearman robertshearman@gmail.com Date: Wed Oct 15 14:13:26 2008 +0100
rsaenh: Make some functions used in only one file static.
:040000 040000 12c9e401b287eedf849134619d6b7cfe19598aa4 5a483c687b16625d9f653325e7056eb941716303 M dlls
CC'ing author
http://bugs.winehq.org/show_bug.cgi?id=17421
--- Comment #4 from Rob Shearman robertshearman@gmail.com 2009-02-18 07:15:34 --- Have you tried doing: git reset --hard 0a2efbb1ccb73da5943baa204aad670ed9b9c9a0 git revert 0a2efbb1ccb73da5943baa204aad670ed9b9c9a0
And then re-compiling. I really doubt that 0a2efbb1ccb73da5943baa204aad670ed9b9c9a0 is the cause of the problems you are seeing, and if it is then there is something much more subtle going on.
http://bugs.winehq.org/show_bug.cgi?id=17421
--- Comment #5 from haarp liquitsnake@gmx.net 2009-02-18 10:00:57 ---
git reset --hard 0a2efbb1ccb73da5943baa204aad670ed9b9c9a0
Diablo fails to work
git revert 0a2efbb1ccb73da5943baa204aad670ed9b9c9a0
works. This was with CFLAGS="-pipe -fno-ident -fomit-frame-pointer -march=k8-sse3 -O3 -fprefetch-loop-arrays -ftracer" I know that the last 2 are experimental. Removing them gets rid of this bug for 1.1.7. My own fault then. Sorry for bothering you Rob...
BUT In 1.1.15, it's back! So along the way something else must've broken. Interestingly, Changing the optimization level down to -O2 also allows Diablo to work there. Now here's the dilemma. I doubt that many people use this level, probably only those with lots of CPU cache. And even then, most will use pre-compiled packages. On the other hand, -O3 is a fairly common CFLAG that should, in theory, pose no problems (aside from making an app slower/faster). Should Wine be required to work properly on this level? I think it should. But others may have other opinions on this. In the meantime, I'll try to find out what broke it again between 1.1.7 and 1.1.15.
http://bugs.winehq.org/show_bug.cgi?id=17421
haarp liquitsnake@gmx.net changed:
What |Removed |Added ---------------------------------------------------------------------------- CC| |Andrew.Talbot@talbotville.co | |m
--- Comment #6 from haarp liquitsnake@gmx.net 2009-02-18 12:26:25 --- Now that's interesting. The new ( -O3 ) bug is in the same component as the first regression:
4be41680a3e18ea2b295415398fad25ac396a984 is first bad commit commit 4be41680a3e18ea2b295415398fad25ac396a984 Author: Andrew Talbot andrew.talbot@talbotville.com Date: Wed Jan 21 22:08:39 2009 +0000
rsaenh: Declare some functions static.
:040000 040000 70110a79b212c87f9981b94b1193bcb0d953240e 32935260eb81d986494dad4034d8ede652192ce3 M dlls
CC'ing author
http://bugs.winehq.org/show_bug.cgi?id=17421
--- Comment #7 from Andrew Talbot Andrew.Talbot@talbotville.com 2009-02-18 17:27:57 --- What is the minimum combination of compiler flags needed for the bug to show?
http://bugs.winehq.org/show_bug.cgi?id=17421
--- Comment #8 from haarp liquitsnake@gmx.net 2009-02-18 18:59:37 --- export CFLAGS="-march=k8-sse3 -O3" Alternatively, -march=amdfam10. Haven't tested any other marches.
http://bugs.winehq.org/show_bug.cgi?id=17421
--- Comment #9 from Dmitry Timoshkov dmitry@codeweavers.com 2009-02-18 21:43:10 --- Sounds like a compiler bug.
http://bugs.winehq.org/show_bug.cgi?id=17421
haarp liquitsnake@gmx.net changed:
What |Removed |Added ---------------------------------------------------------------------------- Status|UNCONFIRMED |RESOLVED Resolution| |DUPLICATE
--- Comment #10 from haarp liquitsnake@gmx.net 2009-03-07 11:33:29 --- I just discovered bug 17406. Sounds awfully similar. Marking this as dup.
*** This bug has been marked as a duplicate of bug 17406 ***
http://bugs.winehq.org/show_bug.cgi?id=17421
Austin English austinenglish@gmail.com changed:
What |Removed |Added ---------------------------------------------------------------------------- Status|RESOLVED |CLOSED
--- Comment #11 from Austin English austinenglish@gmail.com 2009-03-10 15:11:41 --- Closing.