[FFmpeg-devel] [PATCH 1/2] avformat/framehash: don't share write_header code with framecrc muxers

James Almer jamrial at gmail.com
Wed Apr 13 16:50:06 CEST 2016


On 4/13/2016 6:56 AM, Michael Niedermayer wrote:
> On Wed, Apr 13, 2016 at 12:15:42AM -0300, James Almer wrote:
>> On 4/12/2016 11:31 PM, Michael Niedermayer wrote:
>>> On Tue, Apr 12, 2016 at 07:32:44PM -0300, James Almer wrote:
> [...]
>>> we never would want to use a old version for checking
>>> anything as it simply checks less at no advantage
>>> also its IMO important that the extra fields are enabled for the
>>> actual fate tests, this patchset would (please corect me if iam wrong)
>>> remove this support from the actually used (in fate) framecrc
>>
>> True. I wrote this with the intention of keeping framecrc intact, which
>> is pretty much the opposite of your intention to add the extra output to it.
>>
>> In that case guess I'll just turn framecrc into an alias of framehash with
>> adler32 as default. The commit updating ref files will be massive, though,
>> because AVHash and current framecrc initialize adler32 with different
>> values.
> 
> couldnt the initial value be made user setable / setable based on
> which muxer is used to acess framehash?
> so that framecrc could use framehash but with its current value ?
> 
> huge changes like this would increase the git repo size proportionally
> and would make diffing accross hard

AVHash would need to be adapted to do that. adler32, crc32 and murmur3 hardcode
their initial values, and there's no way to change it with the current API.

I'll not convert framecrc into a framehash alias for now anyway, so not an issue.


More information about the ffmpeg-devel mailing list