Revert "Insure that the alsa plugin links against libao, since the plugin calls ao_is_big_endian from libao."

This reverts commit d11c3f1a.

This change was the wrong way to 'fix' the problem reported in trac #1779.
In normal use the plugin should already have access to the symbols from
the library that loaded it, so creating a circular dependency between
them isn't helpful.

If some other user is dynamically loading the core libao itself as a
plugin to their application, then they need to use RTLD_GLOBAL to ensure
its symbols are made available to its plugins (in the same way that libao
needs to do this when dynamically loading alsa for alsa's own plugins).

It's been out of date since 2004 and v0.8.5, so it's not really helping
anyone to still have it in this repo.

See https://packages.qa.debian.org/liba/libao.html for a link to where
the repository containing the debian packaging is found.

The autoreconf tool does the same tool detection the older
shell script does, and is now sufficiently widely deployed
to depend on. The new script is what is currently used in
other Xiph.Org projects.

Note this is a change to the new pattern of needing to
invoke ./configure separately after running ./autogen.sh.

This condenses the flag-heavy compiler invocations, making
warnings easier to see. The old behaviour can be recovered
with `make V=1`.

Replace references with AM_CPPFLAGS. This worked as far back
as automake 1.6, our declared minimum version.

Add newer generated files from the autotools build.

Behavior for default_driver was different from driver-specific options
(no surrounding whitespace allowed). Comments are properly ignored now.

For consistency with plugin API docs and other plugins. Fixes trac #1971.

Fixes trac #1999.

Fixes trac #2022.

Fixes trace #2023.

Fixes trac #2129.

Fixes trac #2204.

git-svn-id: http://svn.xiph.org/trunk/ao@19479 0101bb08-14d6-0310-b084-bc0e0c8e3800

git-svn-id: http://svn.xiph.org/trunk/ao@19478 0101bb08-14d6-0310-b084-bc0e0c8e3800

git-svn-id: http://svn.xiph.org/trunk/ao@19470 0101bb08-14d6-0310-b084-bc0e0c8e3800

Patch by Tristan Matthews. https://review.xiph.org/184/


git-svn-id: http://svn.xiph.org/trunk/ao@19083 0101bb08-14d6-0310-b084-bc0e0c8e3800

git-svn-id: http://svn.xiph.org/trunk/ao@19079 0101bb08-14d6-0310-b084-bc0e0c8e3800

git-svn-id: http://svn.xiph.org/trunk/ao@19076 0101bb08-14d6-0310-b084-bc0e0c8e3800

I didn't list all the trac tickets; that's what the repository
commit log is for. I prefer CHANGES just list the major highlights.


git-svn-id: http://svn.xiph.org/trunk/ao@19075 0101bb08-14d6-0310-b084-bc0e0c8e3800

Fixes 'make distcheck'.


git-svn-id: http://svn.xiph.org/trunk/ao@19074 0101bb08-14d6-0310-b084-bc0e0c8e3800

Also move the link arguments to the end for platforms which require that.

Thanks to Craig Turner for the bug report.


git-svn-id: http://svn.xiph.org/trunk/ao@19005 0101bb08-14d6-0310-b084-bc0e0c8e3800

Insure that the alsa plugin links against libao, since the plugin calls ao_is_big_endian from libao.
Closes Trac #1779


git-svn-id: http://svn.xiph.org/trunk/ao@18984 0101bb08-14d6-0310-b084-bc0e0c8e3800

Add logic to ALSA driver such that it also tries greater bitdepths if device open fails with the requested depth. 



git-svn-id: http://svn.xiph.org/trunk/ao@18883 0101bb08-14d6-0310-b084-bc0e0c8e3800

git-svn-id: http://svn.xiph.org/trunk/ao@18882 0101bb08-14d6-0310-b084-bc0e0c8e3800

use device names (or partial names) or, if you have it, the audio
device UID (UIDs look like
"AppleHDAEngineOutputDP:8,5,1,0:0:{2D4C-05ED-00000000}") with the
-d/--device option.  With the partial matching you can use "-d
speaker" or "-d headphones" or "-d hdmi" or "-d spdif" or "-d iMic"
etc. to select the output device.  (It's a case insensitive match.)

pull out the attempt to set the AudioHardware's CFRunLoop to NULL
because that's just wrong (and fortunately was being ignored by
CoreAudio)



git-svn-id: http://svn.xiph.org/trunk/ao@18881 0101bb08-14d6-0310-b084-bc0e0c8e3800

patch from Kyle McKay <mackyle@gmail.com>


git-svn-id: http://svn.xiph.org/trunk/ao@18868 0101bb08-14d6-0310-b084-bc0e0c8e3800

git-svn-id: http://svn.xiph.org/trunk/ao@18783 0101bb08-14d6-0310-b084-bc0e0c8e3800

enable swap buffer if driver backend changes output channel count


git-svn-id: http://svn.xiph.org/trunk/ao@18781 0101bb08-14d6-0310-b084-bc0e0c8e3800

the period size first, and potentially base other params off the
minimum period.  Some of the plug devices will have rather high
minimum periods, and ending up with too small a buffer in relation (2x
or smaller) guarantees skipping.



git-svn-id: http://svn.xiph.org/trunk/ao@18780 0101bb08-14d6-0310-b084-bc0e0c8e3800

Add latency management the Pulse plugin, along with "buffer_time" option
Add drain() bug workaround to pulse ao_plugin_close(); should eliminate the long wait on teardown



git-svn-id: http://svn.xiph.org/trunk/ao@18779 0101bb08-14d6-0310-b084-bc0e0c8e3800

the output device is disabled; it will queue as fast as it can, but
simply hold the data, meaning the close wait can be huge.

In this case, look or a close wait > 1s, and if so, drain instead of
wait.



git-svn-id: http://svn.xiph.org/trunk/ao@18778 0101bb08-14d6-0310-b084-bc0e0c8e3800

disable underrun detection for only Pulse plugins to avoid the
'immediate underrrun on stream start' bug:
https://github.com/kinetiknz/cubeb/commit/1aa0058d0729eb85505df104cd1ac072432c6d24

Also tweak the ALSA latency configuration in general; the old code was
'trying too hard' AFAICT.



git-svn-id: http://svn.xiph.org/trunk/ao@18777 0101bb08-14d6-0310-b084-bc0e0c8e3800

git-svn-id: http://svn.xiph.org/trunk/ao@18776 0101bb08-14d6-0310-b084-bc0e0c8e3800

of 2s whether there's audio to drain or not.  Add a workaround to the
ALSA plugin to eliminate this behavior when used with Pulse's ALSA
emulation.

(Rather than draining, we query the stream depth, wait, then spike the
stream explicitly)



git-svn-id: http://svn.xiph.org/trunk/ao@18775 0101bb08-14d6-0310-b084-bc0e0c8e3800