Re: Valgrind is grinding my gears

[David Buchan <pdbuchan yahoo com>]

Great! Thanks!

Dave




________________________________
 From: "jcupitt gmail com" <jcupitt gmail com>
To: David Buchan <pdbuchan yahoo com> 
Cc: gtk-app-devel-list list <gtk-app-devel-list gnome org> 
Sent: Tuesday, November 5, 2013 4:56 PM
Subject: Re: Valgrind is grinding my gears
 

I have a valgrind file here I use for my large gtk2 program:

http://www.vips.ecs.soton.ac.uk/development/nip2a.supp

I get clean runs with this file. Run with something like:

        export G_DEBUG=gc-friendly
        export G_SLICE=always-malloc
        valgrind --suppressions=/home/john/nip2.supp \
          --leak-check=yes \
          nip2 ... > nip2-vg.log 2>&1

I've not used it for a while, it might need a bit of updating.





On 5 November 2013 21:01, David Buchan <pdbuchan yahoo com> wrote:
> Aaaah. I see.
>
> Thanks guys.
>
>
>
>
> ________________________________
>   From: Bernhard Schuster <schuster bernhard gmail com>
> To: David Buchan <pdbuchan yahoo com>
> Cc: David Nečas <yeti physics muni cz>; gtk-app-devel-list list <gtk-app-devel-list gnome org>
> Sent: Tuesday, November 5, 2013 3:59 PM
> Subject: Re: Valgrind is grinding my gears
>
>
>
> No, as soon as you use GObject derived types (or call g_types_init/gtk_init) the class structures for all 
> your gobject derived classes will be created _once_. This unevitable, but nothing to worry about (same for 
> GThread and friends), just be aware of their existance (and/or suppress them in the valgrind output).
>
> This has nothing to do with G_SLICE=always-malloc, it actually just reduces the false-positives of valgrind 
> as g_malloc internally allocates big chunks of memory and feeds them as chunks to g_malloc calls, which in 
> turn valgrind sometimes counts as "possibly lost" and clutters the output.
>
> Bernhard
>
>
> On Tue, Nov 5, 2013 at 9:54 PM, David Buchan <pdbuchan yahoo com> wrote:
>
> Hi Dave, "GObject type registration machinery is, under normal circumstances, only ever used tocreate 
> structures that will exist during the entire program lifetime." Does that mean that if I just use straight 
> old malloc() instead of g_slice(), most of the errors would go away? I gather that's essentially what using 
> G_SLICE=always-malloc would do. I can't try this out until late tonight, unfortunately. 
> ________________________________ From: David Nečas <yeti physics muni cz>
> To: David Buchan <pdbuchan yahoo com>
> Cc: gtk-app-devel-list list <gtk-app-devel-list gnome org>
> Sent: Tuesday, November 5, 2013 2:51 PM
> Subject: Re: Valgrind is grinding my gears On Tue, Nov 05, 2013 at 09:47:13AM -0800, David Buchan wrote:
> > But when I invoke Valgrind at runtime, I get a lot of errors which I can't make any sense of. I have 
> > grabbed a small sample of them here: http://pdbuchan.com/valgrind.txt I don't like ignoring errors and 
> > warnings, but I don't know what to do with these. Has anybody else come across these types of Valgrind 
> > notifications?
> Yes, everyone.  You must understand that all the GObject type
> registration machinery is, under normal circumstances, only ever used to
> create structures that will exist during the entire program lifetime.
> So although things such as class reference leaks can exists, eveything
> inside g_type_class_ref() should be ignorable – and you can clearly see
> from the log that these allocations happen once, not a thousand times.
> The same for g_thread_init(), gtk_init(), etc.  Create a suppression
> file or google one... Regards, Yeti
> _______________________________________________
> gtk-app-devel-list mailing list
> gtk-app-devel-list gnome org https://mail.gnome.org/mailman/listinfo/gtk-app-devel-list
> _______________________________________________
> gtk-app-devel-list mailing list
> gtk-app-devel-list gnome org
> https://mail.gnome.org/mailman/listinfo/gtk-app-devel-list