[FFmpeg-devel] [PATCH] RTMP seek support

Howard Chu hyc
Mon Apr 12 18:06:52 CEST 2010

Michael Niedermayer wrote:
> On Mon, Apr 12, 2010 at 08:27:32AM -0700, Howard Chu wrote:
>> Michael Niedermayer wrote:
>>> On Sun, Apr 11, 2010 at 04:40:06PM -0700, Howard Chu wrote:
>>>> Instead of these magical conventions that aren't at all obvious from
>>>> reading the code or its documentation, why not use:
>>>>     current_ts, target_ts, flag.
>>>> For exact seek you can still use current_ts = INT64_MIN as a special case
>>>> but otherwise you drop one useless parameter.
>>> an application might prefer to do seeking with all 3 timestamps different,
>>> above where just examples.
>> OK.
>> We've gotten sidetracked anyway. So one more time with this patch. To
>> actually implement flv_read_seek2 correctly, there would need to be some
>> corresponding av_url_read_fseek2() function to propagate these timestamps
>> down to the ByteIOContext and URLContext. Since there is no such API at the
>> moment, there's not much more I can do with this.
> you could add such api ;)
> anyway that patch looks ok given it works and passes tests

Ah, there's a problem. url_seek() is returning AVERROR(EPIPE) not 
AVERROR(ENOSYS), so av_url_read_fseek() returns failure. (But the seek request 
was issued.)

I was assuming this patch was going in already


and this patch depended on that assumption. oops.


   -- Howard Chu
   CTO, Symas Corp.           http://www.symas.com
   Director, Highland Sun     http://highlandsun.com/hyc/
   Chief Architect, OpenLDAP  http://www.openldap.org/project/

More information about the ffmpeg-devel mailing list