Re: Question about threads

[Havoc Pennington <hp redhat com>]

Mike Dreibelbis <maddoc en com> writes:
> Thanks Havoc for your response, however I may not have been
> totally clear in my original post. 
>
> At the current time I'm not using any mutex's and the threads are
> detached when they are created. Based upon this I don't understand how
> they could be affecting a call to gtk_widget_destroy(), which seems to
> be the case because it hangs upon exit on *every* one. (But only on
> the sparcstation, it works fine under Linux on x86) The dialog boxes are
> all part of the main thread and in fact the threads which are spawned
> are actually of the "work crew" variety and have no interaction with the
> GUI or shared data at all. (At this time anyways).  In fact I can
> execute the program so that no threads are spawned and I get the same
> behaviour. So yes, it doesn't make sense to me that it should be any
> deadlock in this case. I must be missing something here...

Right. Yeah, if you only use GTK from one thread, then g_thread_init()
and the gdk_threads_enter/leave stuff should not be required. You
should be able to simply use a totally unlocked GTK. So I guess that
solves your immediate problem, though it would be nice to get to the
bottom of this deadlock.

In a call such as gtk_widget_destroy(), with locks turned on GTK will
be grabbing and releasing locks quite a few times, so presumably a bug
in either GLib, GTK, or the Sparc thread implementation causes it to
hang. It's hard to speculate on where the problem might be...

Havoc