[Bug 907515] New: ghc 7.4.1 and 7.4.2 compile a correct gtk2hs program such that it crashes with a segmentation fault

[bugzilla at redhat.com]

Product: Fedora
https://bugzilla.redhat.com/show_bug.cgi?id=907515

            Bug ID: 907515
           Summary: ghc 7.4.1 and 7.4.2 compile a correct gtk2hs program
                    such that it crashes with a segmentation fault
           Product: Fedora
           Version: 18
         Component: ghc
          Severity: unspecified
          Priority: unspecified
          Reporter: garrett.mitchener at gmail.com

Description of problem:

I wrote a Haskell program using gtk2hs.  It works fine on Fedora 17 under ghc
7.0.4.  However it crashes with a segmentation fault when compiled on Fedora 18
or rawhide(19).  The crash occurs the first time an action added to the gtk
loop with timeoutAddFull runs.

I wrote a minimal test case that triggers the problem and posted it here:

http://hackage.haskell.org/trac/ghc/ticket/7629

It seems to be a compiler or run-time-system bug in ghc 7.4.1 and 7.4.2 or
possibly gtk2hs, because the binary built on F18 or rawhide crashes when run on
F17, and the binary built on F17 runs without a problem on F18.  

Furthermore, if I disable selinux on the kernel command line at boot time, the
crash vanishes in F18 and rawhide.  I don't know why-- the crash doesn't
trigger any selinux error messages, just a segmentation fault.

Jens Petersen reported to me that the crash does not occur on 64 bit test
machines on Fedora 18, but he confirmed that it does occur on 32 bit test
machines.

I've run into pointer alignment problems with ghc before, where alignment
pragmas in C modules are somehow broken when the C module is linked with a ghc
program, so that may be the case here?


Version-Release number of selected component (if applicable): ghc 7.4.1 and
7.4.2.


How reproducible:

Every time


Steps to Reproduce:

Compile and run the test program posted in the haskell.org trac bug report 

http://hackage.haskell.org/trac/ghc/ticket/7629

I'm also posting the bug here in case someone else runs into it.

-- 
You are receiving this mail because:
You are on the CC list for the bug.
Unsubscribe from this bug https://bugzilla.redhat.com/token.cgi?t=VZ6WAcNDc9&a=cc_unsubscribe