Hi Justin,
FLAC is not just a specification. FLAC is both a specification and a
reference implementation.
I recall that the sample bits must be low-aligned, not high-aligned
as I would have normally assumed. I haven't looked back at the spec
or my source code to confirm my memory, though.
I've just been able to reproduce the problem with a file from Josh
Green. I have yet to analyze the samples, but the problem is not
specific to AIFF or WAV formats. Unfortunately, this indicates that
both libFLAC and the flac command-line tool are broken, so there
isn't any way to look at the flac command-line source to find a way
around the problem.
Sorry to come across as defensive. I guess I let the fact that I've
sent a hundred different 24-bit files through FLAC without ever
seeing this problem lead me to assume that it was a crusty old issue
which hadn't been reproduced. From my point of view, it was starting
to sound a lot like a rumor. But now we may be getting to the bottom
of this.
Brian Willoughby
Sound Consulting
On Mar 30, 2007, at 15:00, Justin Frankel wrote:
To clarify, I am one of the REAPER developers.
To say that FLAC itself is not broken is sort of meaningless--FLAC
itself is just a specification. So yes, it is just libFLAC that is
broken. I have spent QUITE a bit of time trying to sort out why it
fails on 24 bit mode.. If there are some specific alignment
requirements of samples, please point me to the documentation of
them, but we have experimented with many-- and indeed we aren't doing
things TERRIBLY wrong, as it does at least preserve the samples.
Perhaps it's just a little endian or OS issue. You don't need to be
overly defensive of FLAC--this isn't a criticism, it's just a request
for clarification/fixing of an issue that a lot of people seem to
experience.
-Justin
_______________________________________________
Flac-dev mailing list
[email protected]
http://lists.xiph.org/mailman/listinfo/flac-dev
