[FFmpeg-devel] [PATCH 3/4] Implement dynamic memory allocation for MJPEG's picture

[Anatoly Nenashev]

On 28.02.2011 17:13, Anatoly Nenashev wrote:
> On 26.02.2011 22:05, Ronald S. Bultje wrote:
>> Hi,
>>
>> On Wed, Feb 16, 2011 at 5:02 PM, Anatoly Nenashev
>> <anatoly.nenashev at ovsoft.ru>  wrote:
>>> On 17.02.2011 00:38, M?ns Rullg?rd wrote:
>>>> Anatoly Nenashev<anatoly.nenashev at ovsoft.ru>    writes:
>>>>
>>>>
>>>>>   From 38608bf9c49c2ce2c7b3edbfb4d4641ca160a990 Mon Sep 17 
>>>>> 00:00:00 2001
>>>>> From: anatoly<anatoly.nenashev at ovsoft.ru>
>>>>> Date: Tue, 15 Feb 2011 11:33:56 +0300
>>>>> Subject: [PATCH 3/4] Implement dynamic memory allocation for MJPEG's
>>>>> picture
>>>>>
>>>>> ---
>>>>>   libavcodec/jpeglsdec.c |   18 +++++++++---------
>>>>>   libavcodec/mjpegbdec.c |    2 +-
>>>>>   libavcodec/mjpegdec.c  |   44
>>>>> +++++++++++++++++++++++++++-----------------
>>>>>   libavcodec/mjpegdec.h  |    2 +-
>>>>>   4 files changed, 38 insertions(+), 28 deletions(-)
>>>>>
>>>> [...]
>>>>
>>>>
>>>>> diff --git a/libavcodec/mjpegdec.c b/libavcodec/mjpegdec.c
>>>>> index 9792616..2db6760 100644
>>>>> --- a/libavcodec/mjpegdec.c
>>>>> +++ b/libavcodec/mjpegdec.c
>>>>> @@ -81,6 +81,13 @@ av_cold int ff_mjpeg_decode_init(AVCodecContext
>>>>> *avctx)
>>>>>   {
>>>>>       MJpegDecodeContext *s = avctx->priv_data;
>>>>>
>>>>> +    if (!s->picture_ptr) {
>>>>> +        s->picture_ptr = av_mallocz(sizeof(AVFrame));
>>>>> +        if (!s->picture_ptr)
>>>>> +              return AVERROR(ENOMEM);
>>>>> +        s->picture_ptr->reference = 0;
>>>>> +    }
>>>>> +
>>>>>       s->avctx = avctx;
>>>>>       dsputil_init(&s->dsp, avctx);
>>>>>       ff_init_scantable(s->dsp.idct_permutation,&s->scantable,
>>>>> ff_zigzag_direct);
>>>>>
>>>> [...]
>>>>
>>>>
>>>>> diff --git a/libavcodec/mjpegdec.h b/libavcodec/mjpegdec.h
>>>>> index 9c78ab2..a1dee59 100644
>>>>> --- a/libavcodec/mjpegdec.h
>>>>> +++ b/libavcodec/mjpegdec.h
>>>>> @@ -80,7 +80,7 @@ typedef struct MJpegDecodeContext {
>>>>>       int h_max, v_max; /* maximum h and v counts */
>>>>>       int quant_index[4];   /* quant table index for each 
>>>>> component */
>>>>>       int last_dc[MAX_COMPONENTS]; /* last DEQUANTIZED dc (XXX: am 
>>>>> I right
>>>>> to do that ?) */
>>>>> -    AVFrame picture; /* picture structure */
>>>>> +    AVFrame *picture_ptr; /* pointer to picture structure */
>>>>>       int got_picture;                                ///<    we 
>>>>> found a SOF
>>>>> and picture is valid, too.
>>>>>       int linesize[MAX_COMPONENTS];                   ///<    
>>>>> linesize<<
>>>>>   interlaced
>>>>>       int8_t *qscale_table;
>>>>> -- 
>>>>>
>>>> Why don't you simply make an array: AVFrame picture[2]?
>>> Because in this case this isn't enough. I should also add 
>>> "picture_index"
>>> field and implement switching between pictures. I don't think this 
>>> must be
>>> done in this file.
>> For mxpeg, and that's why that's in another patch (4/4) and file.
>>
>> But this is mjpeg.c. You remove picture and replace it with
>> picture_ptr, and then allocate picture_ptr. Why not leave picture, add
>> picture_ptr so no extra allocation is needed?
>>
>> Ronald
>>
>
> Ok. Dynamic memory allocation removed.
>
Simplify condition in decode_end.