[Bug 14920] Wine crashes when running Jezzball

http://bugs.winehq.org/show_bug.cgi?id=14920

Jörg Höhle  changed:

What    |Removed                     |Added
----------------------------------------------------------------------------
                 CC|                            |[e-mail]
                   |                            |t

--- Comment #16 from Jörg Höhle   2009-06-22 09:58:00 ---
Let me consider the present issue as a sort of meta-bug for 16bit apps crashing
in wine on read access to 0xffffffff in winevdm on MacOS.
I don't need a duplicate entry just to report that among other applications,
Pharaoh's demo setup.exe crashes on MacOS 10.5.6+7 (while the application
itself works fine when copied from a Linux install) in wine built with gcc-4.2
from the Mac Xcode install DVD, without Fink nor MacPorts.

Maybe somebody could change the subject line accordingly?

Alexandre Julliard wrote several times now, e.g. in bug #13601:
> Note also that Xcode 3.0 and 3.1 both miscompile the 16-bit relay code, so
> currently the only way to get a fully functional Wine is to build with Xcode
> 2.x.

Xcode is a sort of red herring to me, as it's not the GUI, but presumably the
gcc inside gccX.Y.pkg, distributed as part of the Xcode SDK, that miscompiles
wine.

The location of Apple's modified GNU software (gcc is part of it) is
http://www.opensource.apple.com/source/gcc/
http://www.opensource.apple.com/ contains the source tarballs of Apple's open
source binaries and a mapping from Mac OS and Xcode release to packages and
their build numbers. This shows that both Xcode 2.5 and 3.x contain gcc 4.0.1,
although in different builds:
gcc    4.0.1 dev30+2.5.7 build 5465 ChangeLog:2005-09-28
gcc    4.0.1 dev25       build 5370 ChangeLog:2005-09-28
http://www.opensource.apple.com/tarballs/gcc/gcc-5370.tar.gz

So I invite all you Mac users to compile gcc and report here whether switching
gcc-4.0.1 builds solves any 16bit issue with wine.  If yes, we could
consecutively try to locate the exact cause of the regression in gcc.

Knowing which version of gcc works and which doesn't might help find the exact
cause of the bug. Maybe that's even something the gcc maintainers need to know,
if somebody can isolate it?
> I filed Radar bug 5935237 a couple of months ago,
> but I haven't heard anything back.

Alexandre, would you mind making the details of your bug report and findings
publicly readable for the rest of us via Open Radar?
http://openradar.appspot.com/faq