[a52-dev] Re: [a52-cvs] CVS: a52dec/liba52 parse.c,1.7,1.8

[David I. L. <dl...@vt...>]

* Michel Lespinasse <wa...@us...> [20011214 06:14]:
> Modified Files:
> 	parse.c 
> Log Message:
> Fixed one long-standing bug, where we had to memset the state to 0
> before calling a52_frame(). This should not be necessary anymore.
> 

Cool.  Please release this soon! ;)

IMHO the API should be changed a bit:
* move struct defs from a52.h to liba52/a52_private.h (or whatever)
* leave an annonymous typedef for a52_state_t
* change: a52_init(..) to return an initialized a52_state_t*
* add: sample_t field to a52_state_t
* add: sample_t *a52_samples(a52_state_t *) to get the field
* add: a52_free(a52_state_t *) to free state* and samples_t*

User code shouldn't need to see the internals of your structs right?  If
that's not the case I'd still argue that pieces that need to be accessed
should be done through functions.  I'm not sure the requirements of more
complex user code though.  This will help protect binary breakage if you
change struct sizes or field orders.  Yeah yeah, shared lib thing.  But
it does make the a52.h public header nice and small.

Is this something you'd accept a patch for?

-dave
-- 
David I. Lehn <dl...@vt...>  | http://www.lehn.org/~dlehn/
Computer Engineering Graduate @ Virginia Tech in sunny Blacksburg, VA