[FFmpeg-devel] [PATCH 3/5] avcodec: Factor updating palette out

[Andreas Rheinhardt]

James Almer:
> On 3/17/2021 8:59 PM, Andreas Rheinhardt wrote:
>> Because the properties of frames returned from ff_get/reget_buffer
>> are not reset at all, lots of returned frames had palette_has_changed
>> wrongly set to 1. This has been changed, too.
>>
>> Signed-off-by: Andreas Rheinhardt <andreas.rheinhardt at gmail.com>
>> ---
>>   libavcodec/8bps.c           | 11 +----------
>>   libavcodec/cinepak.c        |  9 +--------
>>   libavcodec/decode.c         | 14 ++++++++++++++
>>   libavcodec/gdv.c            |  5 +----
>>   libavcodec/idcinvideo.c     |  9 +--------
>>   libavcodec/imx.c            |  5 +----
>>   libavcodec/internal.h       |  9 +++++++++
>>   libavcodec/interplayvideo.c |  9 +--------
>>   libavcodec/kmvc.c           |  8 +-------
>>   libavcodec/msrle.c          | 11 ++---------
>>   libavcodec/msvideo1.c       | 10 +---------
>>   libavcodec/qpeg.c           |  9 +--------
>>   libavcodec/qtrle.c          | 10 +---------
>>   libavcodec/rawdec.c         | 13 ++-----------
>>   libavcodec/rscc.c           | 13 ++-----------
>>   libavcodec/smc.c            |  9 +--------
>>   libavcodec/tscc.c           | 10 +---------
>>   17 files changed, 41 insertions(+), 123 deletions(-)
>>
> 
> [...]
> 
>> diff --git a/libavcodec/decode.c b/libavcodec/decode.c
>> index 5a00aeedae..efa8a9ac8d 100644
>> --- a/libavcodec/decode.c
>> +++ b/libavcodec/decode.c
>> @@ -2051,3 +2051,17 @@ FF_ENABLE_DEPRECATION_WARNINGS
>>         return 0;
>>   }
>> +
>> +int ff_copy_palette(void *dst, const AVPacket *src, void *logctx)
> 
> All the arguments for dst are uint8_t*.

Actually, most of them are uint32_t*. The only (?) exception is
rawdec.c. (An earlier version of this patch used "uint32_t
dst[AVPALETTE_SIZE]" with a cast for rawdec, but then I noticed that
this is of by a factor of sizeof(uint32_t).)

> 
>> +{
>> +    buffer_size_t size;
>> +    const void *pal = av_packet_get_side_data(src,
>> AV_PKT_DATA_PALETTE, &size);
> 
> Same, av_packet_get_side_data() returns an uint8_t*.

Yes, but it actually is an array of uint32_t, hence void*. (void* is a
better return value for av_packet_get_side_data() anyway, as most of the
side data types are structures. Maybe we should change that.)

> 
>> +
>> +    if (pal && size == AVPALETTE_SIZE) {
>> +        memcpy(dst, pal, AVPALETTE_SIZE);
>> +        return 1;
>> +    } else if (pal) {
>> +        av_log(logctx, AV_LOG_ERROR, "Palette size %d is wrong\n",
>> size);
>> +    }
>> +    return 0;
>> +}
> 
> [...]
> 
>> diff --git a/libavcodec/internal.h b/libavcodec/internal.h
>> index b57b996816..0fb3107979 100644
>> --- a/libavcodec/internal.h
>> +++ b/libavcodec/internal.h
>> @@ -393,6 +393,15 @@ int ff_int_from_list_or_default(void *ctx, const
>> char * val_name, int val,
>>     void ff_dvdsub_parse_palette(uint32_t *palette, const char *p);
>>   +/**
>> + * Check whether the side-data of src contains a palette of
>> + * size AVPALETTE_SIZE; if so, copy it to dst and return 1;
>> + * else return 0.
>> + * Also emit an error message upon encountering a palette
>> + * with invalid size.
>> + */
>> +int ff_copy_palette(void *dst, const AVPacket *src, void *logctx);
> 
> Should be in libavcodec/decode.h instead.

It was in decode.h until I noticed that most decoders don't include that
header. internal.h includes several other functions that are only used
by decoders (like ff_reget_buffer). But, yes, will move it.

> 
> Or maybe avpacket.c and packet_internal.h, for that matter.

Given that it is only used by decoders, decode.c seems the appropriate
place (is it noticeable that I dream of a day when all the decoding code
is really disabled when there are no decoders enabled?).

> 
>> +
>>   #if defined(_WIN32) && CONFIG_SHARED && !defined(BUILDING_avcodec)
>>   #    define av_export_avcodec __declspec(dllimport)
>>   #else