flac dev - Oct 2004 - FLAC Process_single... read:write ratio ?

I'm just trying to wrap the flac decoder into a push model class... as
i'm
having a few threading problems in corner cases.

One thing i need to fnid out is... if i call process_single() and let the 
read callback give the decoder exactly one packet/frame.

Is the write_callback garaunteed to only be fired once ?

Also one other thing... this where is always crashes...

Line 374 in bitbuffer.c, at bit_buffer_clear(...) the state is ABORTED, and 
blurbs and bits are both ~2^32 and it tries to do a memset.

I know the issue is primarily a threading one... is it normal for this to be 
called in the ABORTED state ? How does the blurbs and bits get so high ? (is 
it subtracting from 0 ?)

Cheers,

Zen.

--- illiminable <ogg@illiminable.com> wrote:
> I'm just trying to wrap the flac decoder into a push model class...
> as i'm 
> having a few threading problems in corner cases.
> 
> One thing i need to fnid out is... if i call process_single() and let
> the 
> read callback give the decoder exactly one packet/frame.
> 
> Is the write_callback garaunteed to only be fired once ?
yes
> Also one other thing... this where is always crashes...
> 
> Line 374 in bitbuffer.c, at bit_buffer_clear(...) the state is
> ABORTED, and 
> blurbs and bits are both ~2^32 and it tries to do a memset.
> 
> I know the issue is primarily a threading one... is it normal for
> this to be 
> called in the ABORTED state ? How does the blurbs and bits get so
> high ? (is 
> it subtracting from 0 ?)
yes, it's probably underflow.

FLAC__bitbuffer_clear() is only called in the decoder from
FLAC__stream_decoder_flush().  is it you calling flush or
reset from somewhere that triggers it, or is it indirectly
through a seek/init/...?

the problem is probably before that anyway though, the blubs/
bits should not underflow.

Josh


__________________________________________________
Do You Yahoo!?
Tired of spam?  Yahoo! Mail has the best spam protection around 
http://mail.yahoo.com