[FFmpeg-devel] [PATCH] avformat/avienc: Remove unneeded seekable tests

[Mats Peterson]

"Ronald S. Bultje" <rsbultje at gmail.com> skrev: (6 mars 2016 14:37:25 CET)
>Hi,
>
>On Sun, Mar 6, 2016 at 7:37 AM, Mats Peterson <
>matsp888-at-yahoo.com at ffmpeg.org> wrote:
>
>> On 03/06/2016 01:29 PM, Ronald S. Bultje wrote:
>>
>>> Hi Mats,
>>>
>>> On Sun, Mar 6, 2016 at 6:51 AM, Mats Peterson <
>>> matsp888-at-yahoo.com at ffmpeg.org> wrote:
>>>
>>> On 03/06/2016 12:09 PM, Michael Niedermayer wrote:
>>>>
>>>> Once again, why are these tests unneeded?
>>>>
>>>
>>>
>>> You shouldn't send the same message 5x within 2 seconds. Please be
>>> patient.
>>>
>>> Why are they unneeded? Imagine this situation:
>>>
>>> ----
>>> int a;
>>>
>>> if (condition)
>>>      a = create_a();
>>>
>>> [..]
>>>
>>> if (condition)
>>>      do_something_with_a(a);
>>> ----
>>>
>>> versus:
>>>
>>> ----
>>> int a = create_a();
>>>
>>> [..]
>>>
>>> if (condition)
>>>      do_something_with_a(a);
>>> ----
>>>
>>> Oddly, both codes do exactly the same thing, assuming they don't
>change
>>> global state. In this case, avio_tell() shouldn't change global
>state,
>>> that
>>> is, we're assuming that it will not cause bitstream corruption
>simply
>>> because it's unseekable, which is a fair assumption.
>>>
>>> On the positive side, with that assumption, code fragment #2 is much
>>> simpler than #1, which is why the patch is net positive.
>>>
>>> Ronald
>>> _______________________________________________
>>> ffmpeg-devel mailing list
>>>
>>
>>
>> Well, that part I understand. It's for simplification purposes.
>>
>> By the way, I just noticed that avio_tell() (and avio_seek() will
>work on
>> stdout as well. That's the part I *don't* understand: How can they
>work on
>> an "officially" unseekable stream like stdout, when ftell() and
>lseek()
>> won't? Is it due to the buffering in FFmpeg? And where's the limit
>where a
>> normally "unseekable stream" like stdout is no longer seekable with
>> avio_seek()? Hope you understand what I mean.
>
>
>See the implementation [1]. If not seekable, and the seek is forward,
>it
>fills buffer until it reaches that position. If not seekable and seek
>is
>backward but fits in buffer, it works as expected by changing buffer
>index.
>Etc. See also avio_tell [2]. Line 230 of avio_seek tells you the
>shortcut
>that makes this all work without calls to fseek/ftell.
>
>Ronald
>
>[1] https://ffmpeg.org/doxygen/trunk/aviobuf_8c_source.html#l00209
>[2] https://ffmpeg.org/doxygen/trunk/avio_8h_source.html#l00447
>_______________________________________________
>ffmpeg-devel mailing list
>ffmpeg-devel at ffmpeg.org
>http://ffmpeg.org/mailman/listinfo/ffmpeg-devel

OK, thanks.
-- 
Mats Peterson
http://matsp888.no-ip.org/~mats/