Re: [Valgrind-users] Valgrind trunk on Mac

[Julian S. <js...@ac...>]

On Wednesday, July 18, 2012, Nikolas Zimmermann wrote:

> > Thanks for the bisecting.  If you use ToT with r12466 backed out, does it
> > work?  Can you try with both --tool=none and --tool=memcheck ?
> 
> I'll try it as soon as trunk builds again on Mac OSX 10.7:
> m_debuginfo/readmacho.c: In function 'vgModuleLocal_read_macho_debug_info':
> m_debuginfo/readmacho.c:1090: error: too few arguments to function
> 'vgModuleLocal_read_debuginfo_dwarf3' m_debuginfo/readmacho.c:1106: error:
> too few arguments to function 'vgModuleLocal_new_dwarf3_reader'

Fixed, r12754.  Urr.  We should have noticed that earlier.

> > Urr, seems like we have a bunch of problems on MacOS to do with OSX 10.8
> > and XCode 4.3, not just the above bug.  Unfortunately I only have a 10.7
> > based machine to test with.
> 
> Out of curiosity: is it XCode faults? Some non-standard mmap
> implementation?

From memory, the list of problems I know about are

* assertion failures w.r.t. alignment for 32 bit processes 
  when V is built by XCode 4.3

* link failures in some build configurations (not sure which ones)

* #include path problems for system includes, again unclear which
  configurations

* compilation of V never finishes, something to do with building mig,
  unknown what configuration

* the bug you mention, which was itself caused by a fix for a different
  problem

* almost certainly, more problems that I don't know about

J