[FFmpeg-devel] [PATCH] lavc: remove libfaac wrapper

[Michael Niedermayer]

On Thu, Sep 29, 2016 at 08:12:14PM -0300, James Almer wrote:
> On 9/29/2016 7:21 PM, Michael Niedermayer wrote:
> > On Thu, Sep 29, 2016 at 09:54:42PM +0100, Josh de Kock wrote:

[...]

> 
> > if libfaac support is droped, muxers wont be tested against it anymore
> > most likely. It could be done with command line faac and remuxing but
> > that special case makes it so unwieldy that i doubt anyone will do it
> > 
> > also quite some testcases in trac tickets used libfaac as it was
> > default at the time,
> > regression testing future changes against these tickets may be
> > unreliable without libfaac
> 
> Bugs in libfaac or the lavc wrapper wouldn't happen anymore because the
> wrapper just wont exist anymore. And if some issue can't be reproduced
> with aacenc or libfdk-aac, then it wouldn't be an issue to begin with.

I never talked about bugs in libfaac or bugs in the libfaac wraper,
in fact i dont remember any bugs there. What i meant are
bugs that use testcases which as part of the testcase used libfaac
(libfaac being one of the few options that were available in the past
 for aac encoding results in it having beenfrequently used)

Many bugs are sensitve to changes, for example a bug in a muxer
is not unlikely to depend on some of the packet sizes, durations,
timestamps, exact initial padding value, ...
the closer one can match all that the more likely testing a testcase
in a old ticket with a modern ffmpeg actually does a useful test.

droping the encoder that was used in such old tickets makes testing
the old testcases less meaningfull.

Its as if your muxer shows a issue with mpeg1 video and mp3 and you
drop these encoders. the muxer might still show the issue with h264
and flac but then it quite possibly wont or it possibly wont with the
sample and command line referenced in the ticket

or if you want to you could also say that fewer encoders being
available limits the broadness/coverage of tests of the muxers and core
and each encoder (not specific to aac encoders) does likely have its
peculiarities.

or you could see it as in I and carl test old tickets for regression
these old tickets used libfaac. libfaacs removial will make such
regression tests less meaningfull as the bugs are less likely to
reproduce with a different encoder.

[...]
-- 
Michael     GnuPG fingerprint: 9FF2128B147EF6730BADF133611EC787040B0FAB

Many things microsoft did are stupid, but not doing something just because
microsoft did it is even more stupid. If everything ms did were stupid they
would be bankrupt already.
-------------- next part --------------
A non-text attachment was scrubbed...
Name: signature.asc
Type: application/pgp-signature
Size: 181 bytes
Desc: Digital signature
URL: <http://ffmpeg.org/pipermail/ffmpeg-devel/attachments/20160930/a23448f9/attachment.sig>