2018-08-24 17:41 GMT+02:00, James Almer <jamr...@gmail.com>:
> On 8/24/2018 12:33 PM, Carl Eugen Hoyos wrote:
>> 2018-08-24 17:31 GMT+02:00, James Almer <jamr...@gmail.com>:
>>> On 8/24/2018 7:19 AM, Carl Eugen Hoyos wrote:
>>>> 2018-08-24 0:17 GMT+02:00, James Almer <jamr...@gmail.com>:
>>>>> Fixes assertion failures when trying to mux such streams.
>>>> Shouldn't this be 1/2?
>>>> And does this assert now for libavformat users that use
>>>> new libopus (but not libavcodec) or do I misunderstand?
>>> This asserts for any stream with >= 80ms packets. It doesn't need to be
>>> a direct encode from the libopus wrapper, since it can also happen
>>> during be a remux.
>> Doesn't this indicate that the assert is wrong?
>> (That invalid input can trigger the assert)
> Invalid input (say, a packet reporting a frame size of the equivalent of
> 1ms) would assert before and after this patch. Do you consider an assert
> that triggers on invalid input to be wrong?

I wanted to write "definitely", I am a little puzzled now that you seem to

Yes, I think so: It probably depends on the definition of "invalid" but
I mean invalid data with valid api usage.

Carl Eugen
ffmpeg-devel mailing list

Reply via email to