Re: [gnet-dev] Possible Memory Leak

[Hubert Sokolowski <h sokolowski wsisiz edu pl>]

On Wed, 26 Jan 2005 13:13:14 +0100
Carsten Burstedde <c burstedde de> wrote:

> I experience the following:
> 
> 1. valgrind does not like the glib mem profile functions, so I removed
> them.
> 2. the memory seems to be allocated in g_ptr_array_sized_new.
> 3. when I do instead
> 
>    context = g_main_context_default ();
>    loop = g_main_loop_new (context, FALSE);
> 
>    ...
> 
>    g_main_loop_unref (loop);
>    g_main_context_unref (context);
> 
> the leak disappears completely.  So it is probably in the source 
> handling of the main context and finally cleaned up.  But then the 
> static GMainContext *default_main_context is dangling nonzero inside 
> glib, which is dangerous.  Maybe propose a patch to
> g_main_context_unref 
> to fix this?
this looks like a bug in glib. you could report this to
gtk-app-devel-list gnome org mailing list.
I am not sure if context returned by g_main_context_default ();
should be freed by the user, the docs does not says about it.

> 
> 4. In my _real_ gnet programs, valgrind says
> ==29743== warning: Valgrind's pthread_getschedparam is incomplete
> ==29743==          your program may misbehave as a result
> And it does no longer respond to the Ctrl-D on stdin and network 
> connects for example.  Is there a possibility to fix this (redhat 9, 
> glib 2.6.1)?
yeah, valgrind is useless for threaded applications.

when you remove gnet code valgrind reports memleaks, run it with
valgrind --tool=memcheck --leak-check=yes  progrname

hs

ps. remember about gnet_init :)