[gst-devel] Re: [gst-cvs] CVS: gstreamer/gst gstbuffer.c,1.36.2.7,1.36.2.8 gstbuffer.h,1.26.4.3,1.26.4.4
vishnu at pobox.com
vishnu at pobox.com
Sun Sep 23 10:17:02 CEST 2001
On Sun, Sep 23, 2001 at 03:04:01PM +0200, Wim Taymans wrote:
> On Sun, 2001-09-23 at 08:03, Joshua Pritikin wrote:
> > Modified Files:
> > Tag: BRANCH-EVENTS1
> > gstbuffer.c gstbuffer.h
> > Log Message:
> > 1 Add GST_BUFFER_WHERE support to bytestream.
>
> I don't like this patch at all.
>
> It screws up automatic doc generation, adds bloat, looks ugly and should
> be done in a more general way IMO, ie. not only on buffer creation but
> as a general trace feature (possibly with -finstrument-functions). I've
> added a print_stack_trace feature in my local copy using the
> -finstrument-functions option that does exactly what this patch does and
> more (and cleaner).
Fine, then put your code in CVS! Taaz & i spent too much time trying
to figure out a memory leak. *Something* had to me done. Gstreamer
is too hard to debug without an easy way to analyze buffer usage.
i see that -finstrument-functions offers an exact stack trace. i didn't
know about this gcc 3.0 feature until now -- i would have used it myself.
Obviously, that's a lot cleaner than passing __FILE__,__LINE__ info
around.
Wim, why are you keeping all these great debugging tools to yourself??
Taaz & i have been complaining about this problem all week and you
were silent.
> Another option (and an idea I have been toying with for some time) is to
> create a GstElementBox element that would fork() and let an arbitrary
> element (even bin) run in its own address space. Communication between
> the processes would be done using shm or pipes. This would make
> monitoring the element much more flexible and accurate without ugly
> hacks.
That sounds great. The co/threads also cause problems with various
memory checking tools. An option to not use co/threads would be
very welcome.
> so, let's not get into this and back out the patch before someone starts
> adding similar code to GstProps/Caps/Events/whatmore.
i'm happy to back it out. The patch has served its purpose eliciting
better suggestions. Up until now, i felt myself stuck in an
information vacuum.
--
Victory to the Divine Mother!!
http://sahajayoga.org
More information about the gstreamer-devel
mailing list