advpack.c must be compiled with optimization?

[Dan Kegel]

Oddly, if I compile the wine tree without -O2, the following
test fails:

../../../wine advpack_test.exe.so advpack.c
advpack.c:441: Test failed: Expected C:\Program Files, got C:\
advpack.c:443: Test failed: Expected size 17, got 4
make: *** [advpack.ok] Error 2

Compiling advpack/advpack.c with at least -O makes
the problem go away.

[Maarten Lankhorst]

Hello Dan,

2008/5/27 Dan Kegel :
> Oddly, if I compile the wine tree without -O2, the following
> test fails:
>
> ../../../wine advpack_test.exe.so advpack.c
> advpack.c:441: Test failed: Expected C:\Program Files, got C:\
> advpack.c:443: Test failed: Expected size 17, got 4
> make: *** [advpack.ok] Error 2
>
> Compiling advpack/advpack.c with at least -O makes
> the problem go away.

Out of curiosity I tried to do the same, it doesn't fail for me,
however I do get an 'unexpected pass into todo block' in
fusion/asmcache.c Valgrinding it gives even different results. (And a
lot of warnings of heapfree being used on a malloc in strdup)

Cheers,
Maarten

[Dan Kegel]

On Mon, May 26, 2008 at 11:14 PM, Maarten Lankhorst
 wrote:
> > Oddly, if I compile the wine tree without -O2, the following
> > test fails:
> >
> > ../../../wine advpack_test.exe.so advpack.c
> > advpack.c:441: Test failed: Expected C:\Program Files, got C:\
>
> Out of curiosity I tried to do the same, it doesn't fail for me,
> however I do get an 'unexpected pass into todo block' in
> fusion/asmcache.c Valgrinding it gives even different results. (And a
> lot of warnings of heapfree being used on a malloc in strdup)

I think http://www.winehq.org/pipermail/wine-patches/2008-May/055234.html
fixes the latter problem...

[Michael Karcher]

Am Montag, den 26.05.2008, 22:16 -0700 schrieb Dan Kegel:
> Oddly, if I compile the wine tree without -O2, the following
> test fails:
>
> ../../../wine advpack_test.exe.so advpack.c
> advpack.c:441: Test failed: Expected C:\Program Files, got C:\
> advpack.c:443: Test failed: Expected size 17, got 4
> make: *** [advpack.ok] Error 2
>
> Compiling advpack/advpack.c with at least -O makes
> the problem go away.

Evil bug of using uninitialized stack space caused by not checking
return values. It usually slips through, as the local variable was
"initialized" in a previous run of get_dest_dir. Working patches are
available and will be submitted in the next hour.

Regards,
  Michael Karcher