it to signed big endian data. pull in a modified copy of the kernels'
auconv.[ch] to help with this (and remove some home grown code that did
some of the same things.)
- swap big endian samples for WAV.
- update audioplay(1) and audiorecord(1) audio driver xrefs.
XXX fix support for unsigned linear samples -- it seems they need
to be converted to signed for WAV???
big-endian systems properly yet -- need to add RIFX support or swap
samples while writing. more on this coming soon.
- new "-F <format>" option for audiorecord. supported values of <format>
are "sun", "wav" and "none", plus a few aliases.
- split sun & wav support out into their own files.
- don't assume that WAV header elements are properly aligned for the
type of data they hold.
XXX: still need to clean up libaudio interfaces a lot.
Also, fix up for and enable WARNS=2. This fixes problems playing from
pipes or standard input, and playing from funny filesystems or other
cases that can't be mmap(2)ed.
- rename audio_get_sun_encoding() to audio_sun_to_encoding()
- add audio_encoding_to_sun()
play.c:
- adapt
record.c:
- convert NetBSD encoding to sun encoding if possible, otherwise don't
output a header at all.
fixes PR#10380 from Michael Eriksson <eramore@era-t.ericsson.se>.
fixes PR bin/9329. Thanks to Wolfgang Rupprecht (wolfgang@wsrcc.com)
for filing the PR and identifying the bug. This change is slightly different
than the one provided in the PR.
Also, document the default encoding in the manpage.
This default is consistent with the default for audiorecord under solaris.
standard error, arrange for a non-zero exit status and continue with the
next file. Addresses PR bin/8748.
* GCC vs. ANSI: a cast doesn't yield an lvalue.
* Use MAP_FAILED.
only return a header length of 0 if we are using -f (force), not
in the case when we know how long the header is. this stops from
playing the audio header.
