[FFmpeg-devel] [PATCH 14/18] lavf: move AVStream.{need_parsing, parser} to AVStreamInternal

[James Almer]

On 10/11/2020 9:12 AM, Anton Khirnov wrote:
> Quoting Andreas Rheinhardt (2020-10-09 16:36:27)
>> Anton Khirnov:
>>> Those are private fields, no reason to have them exposed in a public
>>> header.
>>> ---
>>>  libavdevice/v4l2.c           |   2 +-
>>
>> This is a problem: There is no requirement to update libavdevice at the
>> same time as libavformat; one might use a newer version of libavformat
>> together with an older version of libavdevice. This effectively makes
>> the offsets of all the fields accessed by libavdevice avpriv, regardless
>> of whether they are in AVStream or AVStreamInternal.
> 
> Right. I wish avdevice was either purged from earth or merged into lavf,

Merging it into lavf is for that matter the only proper way to remove
avdevice_register_all() while also maintaining the ability to select an
outdev in functions like av_guess_format(). So i wouldn't be against it.

> in that order of preference.
> Shall I postpone this patch until the bump then?
> 
>>> diff --git a/libavformat/avformat.h b/libavformat/avformat.h
>>> index 9c0b25d37b..b1aab0c8ea 100644
>>> --- a/libavformat/avformat.h
>>> +++ b/libavformat/avformat.h
>>> @@ -1053,12 +1053,10 @@ typedef struct AVStream {
>>>       */
>>>      int codec_info_nb_frames;
>>>  
>>> -    /* av_read_frame() support */
>>> -    enum AVStreamParseType need_parsing;
>>
>> sizeof enum types is compiler-specific IIRC. Compilers may choose a type
>> smaller than int if all the constants fit into a smaller type. Could
>> this cause a problem here?
> 
> Theoretically, I guess. In practice, probably no.
> But changed the type of unused8 to enum AVStreamParseType, just in case.
> 
>>
>> There is no need to keep the superfluous whitespace here and in the
>> other places.
> 
> Ok, fixed locally.
>