On Thu, 2003-02-13 at 08:23, Jason R Thorpe wrote:
> On Thu, Feb 13, 2003 at 08:19:52AM +0200, Arto Huusko wrote:
> 
>  > However, what I was talking about was a *GTK* signal, and I believe
>  > the whole idea of GTK is that you can, and have to, do very much
>  > inside *GTK* signal handlers.
> 
> Is this "GTK signal" like a GUI event dispatch, or something?  If it is
> an actual signal handler, then you're pretty screwed if you're trying to
> do something that is not async-signal-safe.

Yes, with "GTK signal" I mean stuff like event dispatch and so on. GTK+ uses
"signals" to pass information between its objects. These signals have
nothing to do with real signals.

But anyway the problem with my app is that it is buggy; it exhibited the
same behaviour on Linux (well, no assertion there, it just hangs), and by
this I guess that the GDK lock is not recursive (which is a real bummer...
and it's odd that this worked with GNU pth).

The slight problem still remains with NetBSD's pthread library: after I get
the assertion, only kill -9 whacks the process. From the stack trace I sent
it can be seen that abort() has been called, but the process doesn't actually
end.