On 06/14/2017 01:06 PM, Martin Storsjö wrote:
> On Wed, 14 Jun 2017, John Stebbins wrote:
>
>>
>> On 06/14/2017 12:13 PM, Martin Storsjö wrote:
>>> On Wed, 14 Jun 2017, John Stebbins wrote:
>>>
>>>> On 06/13/2017 11:16 PM, Martin Storsjö wrote:
>>>>> On Tue, 13 Jun 2017, John Stebbins wrote:
>>>>>
>>>>>> On 06/13/2017 12:13 PM, Martin Storsjö wrote:
>>>>>>> On Mon, 12 Jun 2017, John Stebbins wrote:
>>>>>>>
>>>>>>>> If AVCodecParameters.codec_tag is 'hvc1' use it instead of 'hev1' for
>>>>>>>> h.265 streams. QuickTime (and other Apple software) requires 'hvc1'.
>>>>>>>> ---
>>>>>>>> libavformat/movenc.c | 11 +++++++++--
>>>>>>>> 1 file changed, 9 insertions(+), 2 deletions(-)
>>>>>>>>
>>>>>>>> diff --git a/libavformat/movenc.c b/libavformat/movenc.c
>>>>>>>> index e389029..8d89a7a 100644
>>>>>>>> --- a/libavformat/movenc.c
>>>>>>>> +++ b/libavformat/movenc.c
>>>>>>>> @@ -726,7 +726,10 @@ static int mov_write_hvcc_tag(AVIOContext *pb,
>>>>>>>> MOVTrack *track)
>>>>>>>>
>>>>>>>> avio_wb32(pb, 0);
>>>>>>>> ffio_wfourcc(pb, "hvcC");
>>>>>>>> - ff_isom_write_hvcc(pb, track->vos_data, track->vos_len, 0);
>>>>>>>> + if (track->tag == MKTAG('h','v','c','1'))
>>>>>>>> + ff_isom_write_hvcc(pb, track->vos_data, track->vos_len, 1);
>>>>>>>> + else
>>>>>>>> + ff_isom_write_hvcc(pb, track->vos_data, track->vos_len, 0);
>>>>>>>> return update_size(pb, pos);
>>>>>>>> }
>>>>>>>>
>>>>>>>> @@ -786,6 +789,8 @@ static int mp4_get_codec_tag(AVFormatContext *s,
>>>>>>>> MOVTrack *track)
>>>>>>>> return 0;
>>>>>>>>
>>>>>>>> if (track->par->codec_id == AV_CODEC_ID_H264) tag =
>>>>>>>> MKTAG('a','v','c','1');
>>>>>>>> + else if (track->par->codec_id == AV_CODEC_ID_HEVC &&
>>>>>>>> + tag == MKTAG('h','v','c','1')) ;
>>>>>>>> else if (track->par->codec_id == AV_CODEC_ID_HEVC) tag =
>>>>>>>> MKTAG('h','e','v','1');
>>>>>>>> else if (track->par->codec_id == AV_CODEC_ID_AC3) tag =
>>>>>>>> MKTAG('a','c','-','3');
>>>>>>>> else if (track->par->codec_id == AV_CODEC_ID_DIRAC) tag =
>>>>>>>> MKTAG('d','r','a','c');
>>>>>>>> @@ -4496,7 +4501,9 @@ AVOutputFormat ff_mp4_muxer = {
>>>>>>>> .write_packet = mov_write_packet,
>>>>>>>> .write_trailer = mov_write_trailer,
>>>>>>>> .flags = AVFMT_GLOBALHEADER | AVFMT_ALLOW_FLUSH |
>>>>>>>> AVFMT_TS_NEGATIVE,
>>>>>>>> - .codec_tag = (const AVCodecTag* const []){
>>>>>>>> ff_mp4_obj_type, 0 },
>>>>>>>> + .codec_tag = (const AVCodecTag* const []){
>>>>>>>> + ff_codec_movvideo_tags, ff_codec_movaudio_tags,
>>>>>>>> + ff_codec_movsubtitle_tags, 0 },
>>>>>>>> .priv_class = &mp4_muxer_class,
>>>>>>> I guess this change is because libavformat refuses to allow you to set a
>>>>>>> custom codec_tag unless it is found in the codec_tag list, right? Does
>>>>>>> this have any other practical implications? I feel a little uneasy about
>>>>>>> this part of it...
>>>>>>>
>>>>>>>
>>>>>> Yes, this change was to fix result of validate_codec_tag in mux.c.
>>>>>>
>>>>>> I was curious about this myself. I could only find 2 places that this
>>>>>> codec_tag array is used. One in
>>>>>> validate_codec_tag which is the instance that needed fixing, and another
>>>>>> in avconv where it is checking if the source
>>>>>> codec_tag is valid for the output format when doing stream copy. In
>>>>>> both of these cases ff_mp4_obj_type is wrong for
>>>>>> codec_tag so I'm not sure why this was used in the first place.
>>>>>> According to git history, this was used from the
>>>>>> beginning, but makes no sense. I also looked at who calls
>>>>>> validate_codec_tag just to be sure it's being used
>>>>>> consistently. It's only used by avformat_write_header.
>>>>>>
>>>>>> I also ran fate to check for regressions.
>>>>> Ok, that makes sense.
>>>>>
>>>>> In that case though, I would very much prefer to do this in two separate
>>>>> steps; one that removes the .codec_tag part from the AVOutputFormat, with
>>>>> a thorough explanation of this (pretty much what you just wrote) and
>>>>> saying that this will allow the caller to indicate that actual codec tag
>>>>> he wants to be used, and then this one on top that allows you to use hvc1.
>>>>>
>>>>>
>>>> Ok, I'll do that. There are 2 additional places ff_mp4_obj_type is used
>>>> for codec_tag. psp and ismv also use it.
>>>> Should I just do the same thing to them, or would you rather something
>>>> more tailored for those use cases (as it is done
>>>> for ipod)? I don't happen to know what fourcc are permitted in these
>>>> variants. There's nothing in the code that
>>>> currently limits what can be put in these.
>>> Hmm, after removing codec_tag in this case, what happens if you try to mux
>>> an unsupported codec? Would it be even better to use another list like the
>>> current one, but which contains the actual real tags that can end up
>>> written in the mp4 file, so you can have both HEVC alternatives in there?
>>>
>>>
>> I think you misread something. I didn't "remove" codec_tag. I set it to {
>> ff_codec_movvideo_tags,
>> ff_codec_movaudio_tags, ff_codec_movsubtitle_tags, 0 }
> Yep, and now on a second thought I think that's wrong. Wouldn't that allow
> you to (attempt to) mux codecs which really aren't supposed to go into an
> mp4, say SVQ3?
>
> So instead it'd be better to mirror the actual list of codecs that are
> allowed in mp4, that currently are in ff_mp4_obj_type, but with sensible
> tag names (including both hev1 and hvc1)?
>
>Understood. I assumed mov and mp4 allowed pretty much the same stream types. But I see this is incorrect. I'll roll that into a new patch. Do you want to have a more limited list for psp and/or ismv? If so, do you happen to know a good source for what should be in that list? -- John GnuPG fingerprint: D0EC B3DB C372 D1F1 0B01 83F0 49F1 D7B2 60D4 D0F7
signature.asc
Description: OpenPGP digital signature
_______________________________________________ libav-devel mailing list [email protected] https://lists.libav.org/mailman/listinfo/libav-devel
