[FFmpeg-devel] [PATCH 2/2] smvjpeg: fix frame leak

Hendrik Leppkes h.leppkes at gmail.com
Fri Jun 21 21:57:23 CEST 2013


On Fri, Jun 21, 2013 at 9:46 PM, Hendrik Leppkes <h.leppkes at gmail.com> wrote:
> On Fri, Jun 21, 2013 at 9:36 PM, Michael Niedermayer <michaelni at gmx.at> wrote:
>> On Fri, Jun 21, 2013 at 07:51:40PM +0200, Hendrik Leppkes wrote:
>>> ---
>>>  libavcodec/smvjpegdec.c | 1 +
>>>  1 file changed, 1 insertion(+)
>>>
>>> diff --git a/libavcodec/smvjpegdec.c b/libavcodec/smvjpegdec.c
>>> index d815ef1..0603e49 100644
>>> --- a/libavcodec/smvjpegdec.c
>>> +++ b/libavcodec/smvjpegdec.c
>>> @@ -177,6 +177,7 @@ static av_cold int smvjpeg_decode_end(AVCodecContext *avctx)
>>>      MJpegDecodeContext *jpg = &s->jpg;
>>>
>>>      jpg->picture_ptr = NULL;
>>> +    av_frame_free(&s->picture[0]);
>>>      av_frame_free(&s->picture[1]);
>>>      ff_codec_close_recursive(s->avctx);
>>>      av_freep(&s->avctx);
>>
>>
>> --- ./tests/ref/fate/smvjpeg    2013-06-19 17:06:55.962610457 +0200
>> +++ tests/data/fate/smvjpeg 2013-06-21 21:35:23.202590216 +0200
>> @@ -1,13 +0,0 @@
>> -#tb 0: 1/1
>> -0,          0,          0,        1,    30720, 0x3a821807
>> -0,          1,          1,        1,    30720, 0x95168e5d
>> -0,          2,          2,        1,    30720, 0xd4d98e45
>> -0,          3,          3,        1,    30720, 0xe340a7ea
>> -0,          4,          4,        1,    30720, 0xb832a22d
>> -0,          5,          5,        1,    30720, 0x1f3eb488
>> -0,          6,          6,        1,    30720, 0x6429ce43
>> -0,          7,          7,        1,    30720, 0x3a3da232
>> -0,          8,          8,        1,    30720, 0x5c02aeff
>> -0,          9,          9,        1,    30720, 0x4c7b1c9a
>> -0,         10,         10,        1,    30720, 0x6f7a8313
>> -0,         11,         11,        1,    30720, 0xaa32fd72
>> Test smvjpeg failed. Look at tests/data/fate/smvjpeg.err for details.
>> make: *** [fate-smvjpeg] Error 139
>>
>
> Cannot reproduce, fate passes for me.

I sent an alternative patch which properly uses refcounted buffers to
ensure the memory is properly handled.
Does this pass for you too?


More information about the ffmpeg-devel mailing list