[Bug 33307] heap corruption with gcc 4.8.0

[wine-bugs at winehq.org]

http://bugs.winehq.org/show_bug.cgi?id=33307

Dan Kegel <dank at kegel.com> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
             Status|UNCONFIRMED                 |NEW
     Ever Confirmed|0                           |1

--- Comment #22 from Dan Kegel <dank at kegel.com> 2013-04-22 00:23:59 CDT ---
I did some bisecting of the wine source.
Configuring with --disable-tests CC="ccache gcc-4.8" helps speed things up.

You have to test both with
  rm -rf ~/.wine; ./wine64 notepad
and
  rm -rf ~/.wine; WINEDEBUG=warn+heap ./wine64 notepad
because sometimes one will crash, sometimes the other will.
You also want to watch the log for strange warnings with munged paths.

-O2 fails; -O1 works; -O2 -fno-builtin-memcpy fails with warn+heap.
So it's not quite just -fno-builtin-memcpy.  But memcpy seems a big part of it.

Compiling everything with -O2, but with -O1 for kernel32, ntdll, and setupapi
seems to work, but has one heap warning.
Compiling everything with -O2 -fno-builtin-memcpy, but with -O1 for kernel32,
ntdll, and setupapi seems to work, with no heap warnings.

kernel32/process.c, and maybe build_argv() in particular, seem to be one
sensitive spot; when I was testing without warn+heap, that seemed to be the
worst bit.

-- 
Configure bugmail: http://bugs.winehq.org/userprefs.cgi?tab=email
Do not reply to this email, post in Bugzilla using the
above URL to reply.
------- You are receiving this mail because: -------
You are watching all bug changes.