[FFmpeg-devel] [PATCH] Move av_open_input_file probe loop to its own method

[Michael Niedermayer]

On Sun, Mar 14, 2010 at 11:41:44PM +0100, Stefano Sabatini wrote:
> On date Friday 2010-03-12 17:53:49 -0500, Micah F. Galizia encoded:
> > On 10-03-10 05:59 AM, Michael Niedermayer wrote:
> > >On Mon, Mar 08, 2010 at 08:29:40PM -0500, Micah F. Galizia wrote:
> > >>On 10-03-07 07:42 PM, M?ns Rullg?rd wrote:
> > >>>Stefano Sabatini<stefano.sabatini-lala at poste.it>   writes:
> > >>>
> > >>>>On date Sunday 2010-03-07 12:48:45 -0500, MIcah Galizia encoded:
> > >>>>>On 10-03-07 05:52 AM, Stefano Sabatini wrote:
> > >>>>>>On date Saturday 2010-03-06 19:02:07 -0500, Micah F. Galizia encoded:
> > >>>>>>>On 10-03-06 06:17 PM, Michael Niedermayer wrote:
> > >>>>[...]
> > >>>>>>>>should be ok if tested
> > >>>>>>>
> > >>>>>>>I tested it in plain old ffmpeg, and with my shoutcast changes.
> > >>>>>>
> > >>>>>>Please test with make test, I tried it and it failed in
> > >>>>>>regtest-pbmpipe.
> > >>>>>>
> > >>>>>>cat pbmpipe.lavf.err
> > >>>>>>/home/stefano/src/ffmpeg.git/./tests/data/lavf/pbmpipe.pbm: Error while
> > >>>>>>opening file
> > >>>>>
> > >>>>>Thank you.  The attached patch fixes it.  The problem was that after
> > >>>>>the initial probe in av_open_input_file (with no actual probe data),
> > >>>>>I was setting the input format to NULL in ff_probe_input_buffer.
> > >>>>>Now all regression tests pass (make test).
> > >>>>>
> > >>>>>Also, I have improved the commenting in internal.h.
> > >>>>>
> > >>>>>Thanks again!
> > >>>>
> > >>>>Patch applied, thanks for the fish.
> > >>>
> > >>>This broke ea-cdata on FATE.
> > >>
> > >>Actually, it would have broken on any probe score lower than 25 for a file
> > >>smaller than 1MB.  The problem was that the old probe loop would reset the
> > >>file position each time before calling get_buffer. As a result, get_buffer
> > >>would never return AVERROR_EOF, so this wasn't ever a situation that needed
> > >>to be handled.
> > >>
> > >>The modified probe just reads on from where it left off in the previous
> > >>probe, so eventually, the end of file is reached before reaching
> > >>PROBE_BUF_MAX, so it would just return an error.  It has been updated now
> > >>to lower the required probe score when the end of file is reached in
> > >>addition to when PROBE_BUF_MAX is reached.
> > >>
> > >>Thanks again, and sorry for breaking FATE!
> > >>--
> > >>Micah F. Galizia
> > >>micahgalizia at gmail.com
> > >>
> > >>"The mark of an immature man is that he wants to die nobly for a cause,
> > >>while the mark of the mature man is that he wants to live humbly for one."
> > >>  --W. Stekel
> > >
> > >>  internal.h |   18 +++++++++++
> > >>  utils.c    |   97 +++++++++++++++++++++++++++++++++++++++++--------------------
> > >>  2 files changed, 84 insertions(+), 31 deletions(-)
> > >>73e23f49aa0fbc3c6a08b506c3888a40cdddab2f  probe_buffer5.diff
> > >
> > >looks good, if tested
> > 
> > Is someone going to check this in?
> 
> Applied, let's see if FATE will choke again this time! ;-)
> 
> BTW, can someone suggest some way to run a complete FATE test even
> without to commit stuff? This may help case for commit like this,
> which can introduce potentially dangerous changes.

Iam more curious why FATE was not implemented as part of make test
so that the server would just have to run make test123 like i suggested
before fate existed ...


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

The real ebay dictionary, page 2
"100% positive feedback" - "All either got their money back or didnt complain"
"Best seller ever, very honest" - "Seller refunded buyer after failed scam"
-------------- next part --------------
A non-text attachment was scrubbed...
Name: not available
Type: application/pgp-signature
Size: 189 bytes
Desc: Digital signature
URL: <http://lists.mplayerhq.hu/pipermail/ffmpeg-devel/attachments/20100315/5d5b88d2/attachment.pgp>