[Bug 34264] New: Crash with recoursive calls to __clone+0x5d (libc): divide by zero in 32-bit code

http://bugs.winehq.org/show_bug.cgi?id=34264

Bug #: 34264
           Summary: Crash with recoursive calls to __clone+0x5d (libc):
                    divide by zero in 32-bit code
           Product: Wine
           Version: unspecified
          Platform: x86-64
        OS/Version: Linux
            Status: UNCONFIRMED
          Severity: major
          Priority: P2
         Component: -unknown
        AssignedTo: [e-mail]
        ReportedBy: [e-mail]
    Classification: Unclassified

Created attachment 45615
  --> http://bugs.winehq.org/attachment.cgi?id=45615
STDOUT with backtrace

Hiho
I tried to install bioshock2;
When I start it up, it shows om splashed but then crashes with a backtrace.
(details below)

#### Setup ##############

Wine Version: 64bit 1.7.xxx but the same error occurred with 1.6.xxx.
My OS: 3.9.9-1-ARCH
Prefix:
  * Dedicated for bioshock
  * No winetricks packets
  * Bioshock installer possibly installed xlive
  * Fake screen 1920x1080, apps stayed all the time in windowed mode.
App:
  * bioshock2
  * using razor crack
  * using bioshock2 fake xlive.dll

#### Debug Output ########

Running with "WINEDEBUG=warn+all" gave me a huge log.
STDOUT contains just the backtrace.
STDERR contains a lot.

See attachments.

#### Related #############

The following bug reports appear related,
because they show a similar backtrace (search "__clone+0x5d")

* Bug 33704 - Naviextras Toolbox crashes              -
http://bugs.winehq.org/show_bug.cgi?id=33704
* Bug 16389 - Mechwarrior 2 fails to start            -
http://bugs.winehq.org/show_bug.cgi?id=16389
* Bug 30380 - Keppe Chat Server can't accept a Client -
http://bugs.winehq.org/show_bug.cgi?id=30380

All the related bugs have the call to "start_thread" that runs recoursive calls
to "clone" in common.
Googeling a bit makes me think that this bug exists among various versions of
wine.

#### Crash Analysis ######

I also had a look at the crash report:
* The actual crash reason seems to  division by zero (in 32bit code?)

Unhandled exception: divide by zero in 32-bit code (0x109223cc).

* The first few lines of the backtrace show core code. (thread starting code?)
  I am assuming that the first lines are the functions called first?

1 0x7bc79480 call_thread_func_wrapper+0xb() in ntdll (0x01bbea48)
    2 0x7bc7c4bf call_thread_func+0x9e() in ntdll (0x01bbeb18)
    3 0x7bc7945e RtlRaiseException+0x21() in ntdll (0x01bbeb48)
    4 0x7bc829e5 in ntdll (+0x729e4) (0x01bbf398)
    5 0xf756ecf0 start_thread+0xcf() in libpthread.so.0 (0x01bbf468)

* The next few hundred lines show recoursive (?) calls of __clone+0x5d() in
libc:

6 0xf74a67ae __clone+0x5d() in libc.so.6 (0x00000000)
    ...
    200 0xf74a67ae __clone+0x5d() in libc.so.6 (0x00000000)

* I think it is worth mentioning that this stack trace only contains functions
  from the following libs:

* libc.so.6
  * libpthread.so.0
  * ntdll

[Bug 34264] Crash with recoursive calls to __clone+0x5d (libc): divide by zero in 32-bit code

http://bugs.winehq.org/show_bug.cgi?id=34264

--- Comment #1 from karo  2013-08-14 16:05:56 CDT ---
Created attachment 45616
  --> http://bugs.winehq.org/attachment.cgi?id=45616
STDERR with plenty

[add STDERR with plenty attachment]

[Bug 34264] Crash with recoursive calls to __clone+0x5d (libc): divide by zero in 32-bit code

http://bugs.winehq.org/show_bug.cgi?id=34264

karo  changed:

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

[Bug 34264] Crash with recoursive calls to __clone+0x5d (libc): divide by zero in 32-bit code

http://bugs.winehq.org/show_bug.cgi?id=34264

Dan Kegel  changed:

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

--- Comment #2 from Dan Kegel  2013-08-15 00:33:25 CDT ---
Nah, the __clone just means the backtrace ran off the end of the stack,
it doesn't mean anything interesting.

I have no idea about the rest of your bug, but 
did you see the tip at http://forum.winehq.org/viewtopic.php?t=15588
about the divide by zero?

[Bug 34264] Divide by zero in bioshock2

http://bugs.winehq.org/show_bug.cgi?id=34264

Dan Kegel  changed:

What    |Removed                     |Added
----------------------------------------------------------------------------
            Summary|Crash with recoursive calls |Divide by zero in bioshock2
                   |to __clone+0x5d (libc):     |
                   |divide by zero in 32-bit    |
                   |code                        |

--- Comment #3 from Dan Kegel  2013-08-15 00:34:04 CDT ---
Oh, also, we don't support cracks.

[Bug 34264] Divide by zero in bioshock2

http://bugs.winehq.org/show_bug.cgi?id=34264

karo  changed:

What    |Removed                     |Added
----------------------------------------------------------------------------
             Status|UNCONFIRMED                 |RESOLVED
         Resolution|                            |INVALID
           Severity|major                       |minor

--- Comment #4 from karo  2013-08-15 05:58:59 CDT ---
Just fixed it;
Solution: Open binary in hex editor, replace divisions by zero with NOP.

Turns out this really had nothing to do with wine itself and was just errors in
the binary and I was mostly confused by the backtrace.

Is there any info on how to read these backtraces?

Thnks for the help,
karo

[Bug 34264] Divide by zero in bioshock2

http://bugs.winehq.org/show_bug.cgi?id=34264

Austin English  changed:

What    |Removed                     |Added
----------------------------------------------------------------------------
             Status|RESOLVED                    |CLOSED

--- Comment #5 from Austin English  2013-08-15 20:02:33 CDT ---
Closing.