Threading problems in (Win)DBus

[Havoc Pennington]

Hi,

Olivier Hochreutiner wrote:
> the problem disappears. I think this clearly means that
> have_connection_lock (or sometimes io_path_acquired; see asserts I
> reported) is overwritten by an overflow. Now the 1-billion-dollar
> question: where does the overflow (or underflow, or corrupted pointer)
> occur ?
> 

valgrind should be helpful for figuring this out (or on Windows I think 
you can get a free eval copy of Purify).

Re: your other post, I'm paranoid enough that I wouldn't quite assume 
DBusMutex isn't broken; it has significant logic beyond just using the 
standard system primitives. So there could be something wrong there.

Havoc