[vlc-devel] vlc: svn commit r25381 (funman)

Rafaël Carré funman at videolan.org
Tue Mar 4 20:33:05 CET 2008

Hash: SHA1

Laurent Aimar a écrit :
> On Fri, Feb 29, 2008, Rafaël Carré wrote:
>> Rémi Denis-Courmont a écrit :
>>> Le Wednesday 27 February 2008 21:46:32 Subversion daemon, vous avez écrit :
>>>> r25381 | funman | 2008-02-27 20:46:31 +0100 (Wed, 27 Feb 2008) | 2 lines
>>>> Changed paths:
>>>>    M /trunk/src/input/stream.c
>>>> fix #1404
>>> This seems to be somewhat of a revert of [24790]. I don't know which commit is 
>>> right or wrong, but can someone triple check?
>> I think [24790] is wrong, because it tries to fill the buffers when the
>> exact end of the file has been reached already.
>  In fact it was the wanted behaviour to detect end of file when seeking
> exactly at the end (it is done the same way with stream mode iof block).
>  Reverting [24790] just broke access that use pf_block() mode (like mapped
> file).
>  What problem do you have with it ?

I reverted this commit because AStreamSeekBlock() doesn't have to return
EOF. This function returns VLC_SUCCESS if the seeking was successfull,
or VLC_EGENERIC if this was not possible.

I consider that seeking to the exact end of file is possible, in ASF
demux we use seeking to skip the last padding bytes in each packet (a
different padding is used in each packet, so they all have the same size).
Obviously the old behavior would cause a problem when skipping the
padding of the last packet (going to the exact end of file).

Did you think the same way than me ?

We may consider setting (access_t*)->info.b_eof to VLC_TRUE , in order
to detect EOF, but I'm in no way an access expert.

- --
Rafaël Carré

Version: GnuPG v1.4.8 (Darwin)
Comment: Using GnuPG with Mozilla - http://enigmail.mozdev.org


More information about the vlc-devel mailing list