Bump all dependent ports that need a bump.
Assign MAINTAINER of audio/libsamplerate and audio/libsndfile to Jan Stary.
ok landry@ sthen@ aja@ jasper@ jeremy@ Jan Stary (MAINTAINER)
- make it work with 24-bit formats (as with azalia and envy)
- use the correct member of struct sio_par for the buffer size
- check both poll(2) revents as well as calculated buffer positions
to know when we can read/write data
sun_driver.c:
- make it work with 24-bit formats (as with azalia and envy)
jackd/engine.c:
- don't spam error messages every time a client quits
add README.OpenBSD
* deal with sio_{read,write} possibly moving less data than requested
* check for out-of-sync conditions and restart if they occur
makes running jackd over aucat much more reliable
- add sndio backend
playback under aucat server is pretty solid. for full-duplex,
starting aucat with 'aucat -l -b 2048' and jackd with just
'jackd -d sndio' seems to work best for me. we want low latency
anyway, right ;)
- fix build on alpha by not comparing a va_list to NULL. problem
noticed and fix tested by merdely@.
- fix possible 100% cpu usage in full-duplex mode
- fix net client/server
- add @bin markers to PLIST
- don't bother trying to fix compiler optimization flag tests. they're
badly broken/linux spcific, and we don't use them anyway.
- the net source client became a "real" client (executable), instead of
an "in process" client (plugin)
- previous sun_driver patches accepted upstream, plus:
* fix cycle timing in duplex mode by moving calculations and function
calls out of the poll() synchronization loop
* shorten xrun and silence writing/reading messages
* remove unused variables
* attempt to stop continuous playback underruns in duplex mode that are
caused by the capture buffer overrunning
* try harder to make sure we have pristine buffers after
sun_driver_start()
* use native endian audio encodings
* report a delay when poll() times out, instead of effectively running
a null cycle
now jack works on my r10k sgi :)
nanosleep(2) instead of usleep(3), which is limited to 1 second.
- stop clients from getting "zombified" after connection graph
changes by ignoring errors from read(2)ing data that isn't used.
- in each call to sun_driver_write(), bzero the buffer that will be
written to the audio device. stops jackd from "buzzing" or
"humming" when the last samples from a client weren't true silence.
- consistently use bzero instead of memset to zero out buffers.
- add missing returns in error conditions.
