Re: nested gdk_threads_enter() hangs: linux/gcc vs. win32/mingw?

[Juergen Dankoweit t-online de]

Hello Felix,

Am Samstag, den 14.05.2005, 18:01 +0200 schrieb Felix Kater:
> Hi,
>
> first let me thank you!
>
>
> Juergen Dankoweit <Juergen Dankoweit T-Online de> wrote:
>
> > Funny. On FreeBSD I have no problems with nested calls of
> > gdk_threads_enter()/gdk_threads_leave().
>
> [...]
>
> > This mutex-semaphore is gdk-wide.
>
> [...]
>
> > As you can see, you MUST not call them nested!
>
>
> If you don't have problems with nested calls of gdk_threads_enter()
> either (like me on win32) then there must be some more magic behind the
> gdk_threads_enter thing than just a global semaphore... 

Well let me speculate: 
There could be two things:
(1) the operating system could detect such dead-locks
(2) the operating system sets a timeout for waiting on semaphores

Let me repeat I realy don't know.

> How can it
> determine if a call is a nested call from the same thread which is
> allowed -- and a call from a foreign thread which is not allowed?


Well let me point out that I wrote one thing not really exact:
This semaphore works only in the process-context in which your
application runs. It is not system-wide - sorry for that mistake.

On the layer under this the process management takes care of this or
even the Xlib.

Regards

JÃrgen