Re: [Linphone-developers] memleak in mediastreamer

[Ghislain MARY]

Hi Christian,

You're absolutely right, there's a memory leak there.

I just commited a fix in mediastreamer2:

commit 54d70169171bab26e257ff6f1f731e0d8eaa6fa3 (HEAD -> master, origin/master, 
origin/HEAD)
Author: Ghislain MARY <address@hidden>
Date:   Wed Jul 12 13:43:30 2017 +0200

    Fix memory leak in RFC3984 unpacker.

Thanks very much for your bug report and for pointing exactly what was 
wrong.

Cheers,

Ghislain


On 07/12/2017 01:08 PM, Christian Spielberger wrote:
> Hi,
>
> we have our own mediastreamer plugin which does h264 encoding/decoding. I found
> that there is a memleak in rfc3984_unpack. And yes, I use rfc3984_init during
> initialization and rfc3984_uninit at the end of the stream.
>
>
> ==7116== 1,964 (+1,748) (432 (+216) direct, 1,532 (+1,532) indirect) bytes in 2 
> (+1) blocks are definitely lost in loss record 6,378 of 6,569
> ==7116==    at 0x4C2AB80: malloc (in 
> /usr/lib/valgrind/vgpreload_memcheck-amd64-linux.so)
> ==7116==    by 0x557301D: ortp_malloc0 (port.c:82)
> ==7116==    by 0x557FAF6: dupb (str_utils.c:157)
> ==7116==    by 0x557FB8B: dupmsg (str_utils.c:169)
> ==7116==    by 0x7527847: update_parameter_set (rfc3984.c:313)
> ==7116==    by 0x7527935: store_nal (rfc3984.c:330)
> ==7116==    by 0x7527D43: rfc3984_unpack2 (rfc3984.c:414)
> ==7116==    by 0x7528028: rfc3984_unpack (rfc3984.c:264)
> ==7116==    by 0x16EAACB3: dec_process (in 
> /usr/local/lib/mediastreamer/plugins/libmsgstdsp.so.0.0.0)
> ==7116==    by 0x7758C3F: ms_filter_process (msfilter.c:207)
> ==7116==    by 0x77594A6: call_process (msticker.c:231)
> ==7116==    by 0x77594A6: run_graph (msticker.c:245)
> ==7116==    by 0x7759469: run_graph (msticker.c:250)
>
>
> I suppose that something like
>
> freemsg(ctx->last_sps)
> freemsg(ctx->last_pps)
>
> is missing in rfc3984_uninit.
>
> regards,
> Chris.