[FFmpeg-devel] [PATCH] Print AV_NOPTS_VALUE in seek test

Ramiro Polla ramiro
Tue May 6 01:05:52 CEST 2008


Michael Niedermayer wrote:
> On Mon, May 05, 2008 at 05:35:37PM +0100, Ramiro Polla wrote:
>> Hello,
>>
>> Ramiro Polla wrote:
>>> Michael Niedermayer wrote:
>>>> On Thu, May 01, 2008 at 09:28:37PM +0100, Ramiro Polla wrote:
>>>>> Hello,
>>>>>
>>>>> Attached patch makes seek_test print AV_NOPTS_VALUE instead of some big 
>>>>> negative and apparently meaningless number.
>>>> Complicating the code for no apparent reason, the regression tests are
>>>> just for developers knowing what they do. And someone not knowing what
>>>> the big meaningless number is wont be able to make any sense of
>>>> "AV_NOPTS_VALUE" either.
>>> Seek test fail on MinGW because MSVCRT doesn't properly printf() big 
>>> numbers. Instead of printing
>>> pts:-102481911520608.625000
>>> it prints
>>> pts:-102481911520608.600000
>> Could you please reconsider this?
>> Seek tests with latest MinGW CVS only fail because of this printf. The 
>> tests don't really fail, but the way the results are printed do. This can 
>> easily be avoided in FFmpeg's side by writing cleaner output.
> 
> Well if you really insist then try
> #define ts2float(ts) ((ts) == AV_NOPTS_VALUE ? 0.0/0.0 : ts)

MSVCRT prints "-1.INF00" or something like that, which is much different 
than the POSIX "nan". Anyways POSIX allows other stuff to print out 
instead of "nan", so it wouldn't even be compatible in all 
POSIX-compliant systems. I'll try to think of some other way.

>> Attached patch prints out -1 instead of AV_NOPTS_VALUE.
> 
> as reimar said, this is just wrong ...

I searched a bit and read some stuff about negative timestamps... I 
wasn't aware of them. Of course, otherwise it wouldn't make sense to 
define AV_NOPTS_VALUE in the first place.

Ramiro Polla




More information about the ffmpeg-devel mailing list