[Paranoia] Getting paranoia to quit on CD eject

[Peter Jones]

On Sun, 2004-10-17 at 08:03 -0400, Bill Davidsen wrote:
> On that topic, it would be nice if there were more control over error 
> handling, such as a limit on errors per track, after which the track 
> would be abandoned, etc.

Right now, pretty much everything about errors and warnings is on my
"can stand some serious work" list.  If nothing else, the way messages
are reported from the libraries to the front end (and consequently from
the libraries to gstreamer/grip/abcde/etc) is really inflexible.  I
think having the frontend register a callback is probably the best plan,
but I'd love to hear other peoples' opinions.

It might be nice to have the callbacks be able to alter the plan for
ripping the disk, as well, but that's not nearly as important.

And, of course, this would be for libcdda_interface level errors, not
sg/SG_IO/ioctl level errors.  Still, the stuff e.g. grip does to display
our percentage bars is pure pain -- I think it actually causes things to
be slower, because libcdda_* get stuck into a select() loop talking to
X, but that's just my quick reading of an strace, so it may not be 100%
accurate.  Anyway, it looks quite silly.
-- 
        Peter