An update on this:
The 272 byte offset (68 sample) for CD-DA was removed from bincue.c and
cdrdao.c and the regression tests fixed up. The code was emulating the
offset on the original author's CD drive. Unfortunately, there isn't
any standard for the read offset manufacturer's use. It's just sloppy
hardware (The exception being certain Plextor models that had no
offset.) This is an interim fix until we decide how something more
sophisticated should look and behave.
There are numerous issues to address before we add audio read offset
correction to the library. Briefly:
1. Should the audio read interface remain sector-based when offset
correction will result in sector reads that straddle sector boundaries?
2. Should the library attempt to overread into the lead-in or lead-out
to get missing samples on either the first or the last track (depending
on whether the read offset correction is negative or positive)? Note
that overread is not supported by all drives and locks up some systems!
3. How do we detect a drive's read offset without resorting to
accessing a proprietary database such as the Accurate Rip database?
