Vorbis dev - Dec 2001 - bug report for oggenc in CVS: vorbis-tools/oggenc/encode.c

I was wondering why my CD ripping was producing short files from some tracks.

A superficial examination of the code shows this piece of code in
vorbis-tools/oggenc/encode.c:

Line 120:

                /* Main encode loop - continue until end of file */
        while(!eos)
        {
        [...]
                /* While we can get enough data from the library to analyse, one
                   block at a time... */
                while(vorbis_analysis_blockout(&vd,&vb)==1)
                {
                [...]
                        /* If we've gone over a page boundary, we can do
actual output,
                           so do so (for however many pages are available) */

                        while(!eos)
                        {
                        [...]
                                if(ogg_page_eos(&og))
                                        eos = 1;
                        }
                }
        }

Now, backtracing through vorbis/lib, it seems to me that the inner eos is
being used to stop the output when we've temporarily run out of input,
which is fine. But the SAME eos flag is used to stop the outer loop. I
feel that either eos should be reset after the inner loop, or that there
should be a separate eos-like flag used for the inner loop.

Am I right of havs I misconstrued things. Certainly when I' getting short
track rips:

        - the input raw data is _not_ all consumed

        - oggenc exits "normally", with a "Done encoding file
message" just like it"s
          seen end of input

I'm off to rebuild oggenc to test this theory, but meanwhile would someone
who
knows the code properly care to see if I'm right?

Thanks,

--
Cameron Simpson, DoD#743        cs@zip.com.au    http://www.zip.com.au/~cs/

"I'm a lawyer."   "Honest?"   "No, the usual
kind."

--- >8 ----
List archives:  http://www.xiph.org/archives/
Ogg project homepage: http://www.xiph.org/ogg/
To unsubscribe from this list, send a message to
'vorbis-dev-request@xiph.org'
containing only the word 'unsubscribe' in the body.  No subject is
needed.
Unsubscribe messages sent to the list will be ignored/filtered.

On Mon, Dec 03, 2001 at 05:13:04AM +0000, Cameron Simpson
wrote:
> I was wondering why my CD ripping was producing short files from some
tracks.
> 
> A superficial examination of the code shows this piece of code in
vorbis-tools/oggenc/encode.c:
[...]

You misunderstand the code.  The eos is set by libvorbis, and
libvorbis only sets this when it gets a buffer submission of zero
samples.  A zero sample buffer submission happens only if the fread()
returns zero, and fread returns zero iff it hits end of file.  A pipe
waiting for more input is not an EOF condition.

This might be a bug due to oggenc deciding to stop reading early for
some other reason, but the fault would not be in the code you highlight.

(BTW, the suggestion of 'resetting' this flag disobeys libvorbis usage
documentation and would eventually result in a segfault).

Monty

--- >8 ----
List archives:  http://www.xiph.org/archives/
Ogg project homepage: http://www.xiph.org/ogg/
To unsubscribe from this list, send a message to
'vorbis-dev-request@xiph.org'
containing only the word 'unsubscribe' in the body.  No subject is
needed.
Unsubscribe messages sent to the list will be ignored/filtered.

>Now, backtracing through vorbis/lib, it seems to me that the inner eos is
>being used to stop the output when we've temporarily run out of input,
>which is fine. But the SAME eos flag is used to stop the outer loop. I
>feel that either eos should be reset after the inner loop, or that there
>should be a separate eos-like flag used for the inner loop.
The eos flag is set once an output page is produced that has the eos flag
set - this flag only gets set once all input has been consumed. The inner
loop stops on eos once the encode is complete - usually, it breaks out. 
Then the right thing to do is to also finish the outer loop - which is what
the code does.

Not sure why you're getting short encodes though - are you encoding from
stdin? On windows? I had a report a while ago about short encodes under
those circumstances, but couldn't reproduce it here. 

Michael

--- >8 ----
List archives:  http://www.xiph.org/archives/
Ogg project homepage: http://www.xiph.org/ogg/
To unsubscribe from this list, send a message to
'vorbis-dev-request@xiph.org'
containing only the word 'unsubscribe' in the body.  No subject is
needed.
Unsubscribe messages sent to the list will be ignored/filtered.