Re: g_idle_add(): limits?

[Chris Vine <chris cvine freeserve co uk>]

On Wed, 10 Mar 2010 15:10:50 +0100
"fkater googlemail com" <fkater googlemail com> wrote:
> Chris Vine:
>
> > > I make heavy use of g_idle_add(), and encounter this problem:
> > > It seems that the idle functions aren't called sometimes.
>
> [...]
>
> > Assuming that is right, any sources you add before calling
> > gtk_main_quit() without returning control to the main loop will not
> > be executed.  I think it probably also means that if you call
> > gtk_main_quit() in a GTK+ signal handler, which will have a higher
> > priority than a default idle event, a stacked idle callback will
> > also not be called (that depends on what the documentation that I
> > have referred to means by "dispatched").
>
> This might be the case here: I am definitly not calling
> gtk_main_quit() too early, since I am sleeping in a loop
> before calling it, waiting for my "pending idle function
> counter" to become 0 (but it doesn't sometimes as
> mentioned).  However, the locking situation here really
> prevents giving GTK+ back control while waiting.
>
> I assume this is the problem (since not occuring always I
> can not yet say definitly this was the cause, though).

You could test it by running 'while (gtk_events_pending())
gtk_main_iteration();' immediately before the call to gtk_main_quit(),
(or you could call 'while (g_main_context_pending(NULL))
g_main_context_iteration(NULL, FALSE)'). In either case, make it
absolutely the last thing that is done before gtk_main_quit().

Chris