[FFmpeg-devel] First Patch for hlsenc.c for https://trac.ffmpeg.org/ticket/7281

Steven Liu lq at chinaffmpeg.org
Tue Aug 14 02:06:58 EEST 2018



> On Aug 14, 2018, at 05:06, Ronak <ronak2121-at-yahoo.com at ffmpeg.org> wrote:
> 
>> 
>> On Aug 13, 2018, at 10:49 AM, Steven Liu <lingjiujianke at gmail.com> wrote:
>>> 
>>>> 
>>>> Hi Steven,
>>>> 
>>>> Did you have a chance to review this patch? I haven't seen your feedback yet. Did you have a chance to try it out?
>>>> I would like to get this merged this week.
>>> 1. you MUST update the TARGET DURATION if the new fragment is long
>>> than old fragment, don't only test your audio files please, ffmpeg not
>>> only be used by you.
>> 
>> This patch already does that. I've confirmed with several video files. Please see the attached manifest that was generated by this patch as proof. This is a video file downloaded from here: https://www.sample-videos.com/index.php#sample-mp4-video<https://www.sample-videos.com/index.php#sample-mp4-video>.
> 
> Trying to send the attachment again, it got removed for some reason.
> 
> <output_test_video.m3u8>
>> 
>>> 2. if the temp file have problem, you should seperate the patch into
>>> two patch, one fix temp file problem, the other implement your
>>> feature.
>> 
>> The two problems are interrelated. The creation/destruction of temp files repeatedly adds to the generation latency here. In my use-case, ffmpeg generates and destroys the temp file 155000 times when: 1. This is offline VOD. I am just generating from my laptop to my laptop. No Web Servers are involved. 2. I did not explicitly ask for this.
>> 
>>> 3. I need more time to think about how to improve the refresh problem
>>> when in VOD mode.Because this patch is not a good way, that is not a
>>> full care about the other function.
>> 
>> The alternative way is the one I was proposing first; which was to write the header once; and then the segments only in the hls_window method.
>> However, that has problems like you pointed out, due to the architecture that ffmpeg has. We can still pursue that path, but it will be a much more complicated fix. I would need much more of your help to fix this.
>> Dashenc.c also suffers from this same problem. This needs to be fixed in both places, not just hlsenc.c.
>> 

Could you please resend a new version patch, use "git send-email”,  make patch use “git format-patch -v” to make patch version. I cannot find the newest patch :(.
Then the patch can be picked up by patchwork.

Thanks

Steven





More information about the ffmpeg-devel mailing list