[FFmpeg-devel] [Bugreport] and [Patch]: duration of TS streams is wrong when used near 2^33 wrap

[Måns Rullgård]

Thorsten Jordan wrote:
> Michael Niedermayer schrieb:
>> Hi
>>
>> On Mon, Oct 29, 2007 at 04:52:52PM +0100, Thorsten Jordan wrote:
>>> Rich Felker schrieb:
>>>> On Mon, Oct 29, 2007 at 01:01:38PM +0100, Michael Niedermayer wrote:
>>>>>> I'm inclined to say that rather than trying to guess the unknown, we
>>>>>> should
>>>>>> simply report the duration of these files as unknown.
>>>>> yes, fully agree
>>>> Me too. I'm inclined to always agree with this principle. Returning
>>>> guesses in fields documented as providing reliable information is very
>>>> bad.
>>> following that argumentation, the current implementation to compute
>>> duration would have to be removed too, because it is a guess as well, as
>>> Mans said. But who would want that?
>>> The proposed change wouldn't make the current situation much worse imho,
>>> but ok, it is your decision.
>>
>> the change makes the code worse
>> as it turns negative durations (which would be ignored) into valid positive
>> ones, in practice such negative durations are more likely caused by
>> timestamp resets instead of 33bit wraping
> Sorry, but i doubt that this is the case. PTS wrap occurs every 26.5
> hours, but how often do PTS resets occur? In fact i never saw one yet
> and i am working much with TS of TV broadcasts. They may happen, but
> surely less often than simple wraps.

Roughly half of all DVDs have at least one PTS reset.  Whether you see
discontinuities in broadcast streams depends on the broadcaster.  I
don't remember the details, but I know I've seen it happen on live TV
streams.

-- 
M?ns Rullg?rd
mans at mansr.com