[gst-devel] RE : Re: Memory leaks

[Stefan Kost]

hi,


Steve Fink wrote:
> On 9/22/07, James Doc Livingston <doclivingston at gmail.com> wrote:
>> On Sat, 2007-09-22 at 09:31 -0700, Steve Fink wrote:
>>> I think it was 0.10.13 originally? But I get the same thing with current CVS.
>>>
>>> ==00:00:03:44.899 25856== 719,760 bytes in 8,997 blocks are definitely
>>> lost in loss record 2,505 of 2,505
>>> ==00:00:03:44.899 25856==    at 0x4021620: malloc (vg_replace_malloc.c:149)
>>> ==00:00:03:44.899 25856==    by 0x415A2C5: g_malloc (gmem.c:131)
>>> ==00:00:03:44.899 25856==    by 0x4169E97: g_slice_alloc (gslice.c:777)
>>> ==00:00:03:44.899 25856==    by 0x416A214: g_slice_alloc0 (gslice.c:784)
>>> ==00:00:03:44.899 25856==    by 0x40FF336: g_type_create_instance (gtype.c:1549)
>> <snip>
>>
>> You'll want to run this with the environment variables
>> G_SLICE=always-malloc and G_DEBUG=gc-friendly set, otherwise you'll get
>> the above false positive from glib's slice allocator.
> 
> I am. I run valgrind with the shell script below. How can I tell if
> they are false positives? (I *am* observing memory use going up, after
> all.)
> 
> export GLIBCPP_FORCE_NEW=1
> export GLIBCXX_FORCE_NEW=1
> export G_SLICE=always-malloc
> export G_DEBUG=gc-friendly
> 
> DP="/home/sfink/"
> SUPP="--suppressions=/usr/lib/valgrind/glibc-2.3.supp
> --suppressions=${DP}gstreamer/common/gst.supp
> --suppressions=${DP}gst-plugins-base/common/gst.supp
> --suppressions=${DP}gst-plugins-good/common/gst.supp"
> 
> OPTS="-v --trace-children=yes --track-fds=yes --time-stamp=yes
> --tool=memcheck --leak-check=yes --leak-resolution=high
> --freelist-vol=10000000 --num-callers=40 ${SUPP}"
> 
> exec valgrind $OPTS "$@"

With those suppresions in use the shown leaks might be real. Start with the
"definitely lost" and look at the code. Its hard, but thats the game.

Stefan