After an evaluation, GNOME has moved from Bugzilla to GitLab. Learn more about GitLab.
No new issues can be reported in GNOME Bugzilla anymore.
To report an issue in a GNOME project, go to GNOME GitLab.
Do not go to GNOME Gitlab for: Bluefish, Doxygen, GnuCash, GStreamer, java-gnome, LDTP, NetworkManager, Tomboy.
Bug 747612 - Valgrind 3.7.0 chokes on match-leak-kinds in gst.supp
Valgrind 3.7.0 chokes on match-leak-kinds in gst.supp
Status: RESOLVED FIXED
Product: GStreamer
Classification: Platform
Component: gstreamer (core)
git master
Other Linux
: Normal normal
: 1.6.1
Assigned To: GStreamer Maintainers
GStreamer Maintainers
Depends on:
Blocks:
 
 
Reported: 2015-04-10 10:57 UTC by Vincent Penquerc'h
Modified: 2015-09-29 12:00 UTC
See Also:
GNOME target: ---
GNOME version: ---


Attachments
make suppressions compatible with valgrindd 3.7.0 (1.02 KB, patch)
2015-04-20 14:00 UTC, Vincent Penquerc'h
committed Details | Review

Description Vincent Penquerc'h 2015-04-10 10:57:49 UTC
+{
+   closures aren't valgrind friendly (bgo#739850)
+   Memcheck:Leak
+   match-leak-kinds: possible
+   fun:calloc
+   ...
+   fun:g_cclosure_new
+}

The "match-leak-kinds: possible" causes valgrind 3.7.0 to error out (amusingly for debugging, on a line number in the set two positions before the one with the match-leak-kinds.

This feature was added in 3.9.0:
http://valgrind.10908.n7.nabble.com/Valgrind-3-9-0-is-available-td47723.html

I half expect this will be deemed wontfix, and that is fine if it is.
Comment 1 Vincent Penquerc'h 2015-04-20 10:41:19 UTC
I'm thinking one could make this a configure generated file, with this match-leak-kinds being parameterized and replaced with "" if unsupported.
Comment 2 Tim-Philipp Müller 2015-04-20 10:52:04 UTC
No, please, let's not make this configure-generated :)

Is this "match-leak-kinds: possible" line really needed in the first place?
Comment 3 Vincent Penquerc'h 2015-04-20 11:23:13 UTC
Well, I guess the supressions are pretty encompassing:

{
   closures aren't valgrind friendly (bgo#739850)
   Memcheck:Leak
   match-leak-kinds: possible
   fun:calloc
   ...
   fun:g_closure_new_simple
}

So removing it might hide other actual leaks.

I'm running with 3.10.1 and these removed now, I've hit an unrelated leak in core, which I will fix, then report whether I saw anything from closures when done.
Comment 4 Tim-Philipp Müller 2015-04-20 11:34:31 UTC
Well, seeing that we are extremely unlikely to use this API directly, this is ultimately GLib stuff then. If we catch memory leaks in GLib, great, but we are more interested in stuff in GStreamer so I don't think it's a problem to simply ignore some known leaks in GLib here.
Comment 5 Vincent Penquerc'h 2015-04-20 14:00:11 UTC
Created attachment 301997 [details] [review]
make suppressions compatible with valgrindd 3.7.0

If we don't, then here's a patch to remove the lines.
And it turns out the leak in the tests was already fixed by you yesterday (push_buffer_async) :)
Comment 6 Tim-Philipp Müller 2015-09-29 12:00:21 UTC
ommit 6babecd339cce45223ddb8312400811ba956a16d
Author: Vincent Penquerc'h <vincent.penquerch@collabora.co.uk>
Date:   Mon Apr 20 14:54:59 2015 +0100

    gst.supp: make suppressions compatible with valgrind 3.7.0
    
    "match-leak-kinds: possible" is a recent valgrind addition,
    and valgrind 3.7.0 errors out on it.
    
    https://bugzilla.gnome.org/show_bug.cgi?id=747612