On 12/9/12 6:26 PM, Måns Rullgård wrote:
Luca Barbato <[email protected]> writes:
On 12/9/12 5:37 PM, Måns Rullgård wrote:
Luca Barbato <[email protected]> writes:
On 12/9/12 1:32 PM, Måns Rullgård wrote:
Luca Barbato <[email protected]> writes:
On 12/09/2012 05:34 AM, Luca Barbato wrote:
Hi, I'm wondering if we could completely hide the avio checksum fields
from the public structure by moving it to an internal field or extend it
to be able to use any hash function from avutil and fully expose it to
our users.
The current situation is less than optimal.
Sort of related, I'd provide some wrapper api over our crypto/hash stuff
along those lines.
typedef struct AVHash {
const AVClass *av_class;
int context_size;
void *context;
int (*init) (struct AVHash);
void (*update)(struct AVHash, const uint8_t *src, const int len);
void (*close) (struct AVHash, uint8_t *dst);
} AVHash;
That's something I've been thinking of doing for a long time.
Anything to add to the the drafted api?
You've already made more progress than I ever did on this.
I'd craft it, implement it for md5, sha and crc and then make AVIO use
an AVHash pointer.
The open question is if we want to add a muxer capability to signal it
is internally used and let people use it in their external code.
What is it currently used for?
ogg and nut (and probably transogg or whichever will be the name of
the new container) have some optional checksumming of fields.
avio has a way to register a checksum function so that from that point
till you request the final result, all the io operations will update
the checksum automatically.
It is quite nice since the resulting code is really terse:
ffio_init_checksum
as many avio operations as you want
ffio_get_checksum
What happens if you seek?
The checksum update happens on fill_buffer/flush_buffer and get_checksum
finalizes it with the possible remaining trailer.
So it works only on read and write operation.
the current situation is annoying since ff_crc04C11DB7_update is the
only function matching the callback signature and you can't use
something else since the state isn't really forwarded properly.
With a generic checksum/hash interface, this would all be solved, right?
Exactly.
lu
_______________________________________________
libav-devel mailing list
[email protected]
https://lists.libav.org/mailman/listinfo/libav-devel
