[FFmpeg-cvslog] r13880 - in trunk: libavformat/ffmenc.c tests/libav.regression.ref

[Baptiste Coudurier]

Michael Niedermayer wrote:
> On Tue, Jun 24, 2008 at 02:47:40AM -0700, Baptiste Coudurier wrote:
>> Stefano Sabatini wrote:
>>> On date Monday 2008-06-23 22:06:36 -0700, Mike Melanson wrote:
>>>> Mike Melanson wrote:
>>>>> bcoudurier wrote:
>>>>>> Author: bcoudurier
>>>>>> Date: Sun Jun 22 10:50:57 2008
>>>>>> New Revision: 13880
>>>>>>
>>>>>> Log:
>>>>>> store dts in ffm headers
>>>>> Congratulations; it was either this revision or the previous one that 
>>>>> finally fixed the regression test when building FFmpeg under Intel's C 
>>>>> compiler.
>>>>>
>>>>> Now I need to figure out when and why the PowerPC regressions broke...
>>>> It was either this revision or the one after it which broke 'make test' 
>>>> under gcc 4.1.2 and 4.0.4, both on PowerPC. The exact error looks like this:
>>>>
>>>> --- /home/melanson/ffmpeg/ffmpeg-main/tests/libav.regression.ref
>>>> 2008-06-22 11:21:56.000000000 -0700
>>>> +++ tests/data/lavf.regression  2008-06-23 20:18:43.000000000 -0700
>>>> @@ -15,9 +15,9 @@
>>>>   d6fdeb9f7083cc827f9510c6c4517dc0 *./tests/data/b-libav.swf
>>>>   335771 ./tests/data/b-libav.swf
>>>>   ./tests/data/b-libav.swf CRC=0xe14e8847
>>>> -3dbacdc3fccb551f8ab54c32f648e7a8 *./tests/data/b-libav.ffm
>>>> +3718a91970bbab48113eb50104a15241 *./tests/data/b-libav.ffm
>>>>   380928 ./tests/data/b-libav.ffm
>>>> -./tests/data/b-libav.ffm CRC=0x2b71a386
>>>> +./tests/data/b-libav.ffm CRC=0x00000001
>>>>   f8ad5bd78f4d012a8ce9570aa395ac54 *./tests/data/b-libav.flv
>>>>   335833 ./tests/data/b-libav.flv
>>>>   ./tests/data/b-libav.flv CRC=0xe14e8847
>>>> make: *** [libavtest] Error 1
>>>>
>>>> gcc-svn for PowerPC fails in the same way. I can't pinpoint the breakage 
>>>> quite as precisely because that build only started to work again recently.
>>> I have exactly the same error here:
>>> i686, gcc 4.2.3.
>>>
>> To be honest I don't really know what's going on here.
>>
>> I see that recompiling ffmenc.c without -O3 gives diff too:
>>
>> --- /home/bcouduri/src/ffmpeg/svn/tests/libav.regression.ref
>> 2008-06-22 01:48:35.000000000 -0700
>> +++ tests/data/lavf.regression  2008-06-24 02:44:45.000000000 -0700
>> @@ -15,7 +15,7 @@
>>  d6fdeb9f7083cc827f9510c6c4517dc0 *./tests/data/b-libav.swf
>>  335771 ./tests/data/b-libav.swf
>>  ./tests/data/b-libav.swf CRC=0xe14e8847
>> -3dbacdc3fccb551f8ab54c32f648e7a8 *./tests/data/b-libav.ffm
>> +02ccb72ae3ad8703a9fdd528d4b8f109 *./tests/data/b-libav.ffm
> 
> Why dont we finally drop .ffm and use nut instead ?
> 
> It cant be denied that you spend considerable time in fixing .ffm, i really
> do not belive getting nut useable for the ffm use case would take that long.
> 

Yes, definitely, you're right. I need to get this done pretty soon. I
need to understand nut and to figure how to use info packets to
transport encoding options.

-- 
Baptiste COUDURIER                              GnuPG Key Id: 0x5C1ABAAA
Smartjog USA Inc.                                http://www.smartjog.com
Key fingerprint                 8D77134D20CC9220201FC5DB0AC9325C5C1ABAAA