Log Message:
openal-soft: updated to 1.21.1

openal-soft-1.21.1:

Improved alext.h's detection of standard types.

Improved slightly the local source position when the listener and source
are near each other.

Improved click/pop prevention for sounds that stop prematurely.

Fixed compilation for Windows ARM targets with MSVC.

Fixed ARM NEON detection on Windows.

Fixed CoreAudio capture when the requested sample rate doesn't match the
system configuration.

Fixed OpenSL capture desyncing from the internal capture buffer.

Fixed sources missing a batch update when applied after quickly restarting
the source.

Fixed missing source stop events when stopping a paused source.

Added capture support to the experimental Oboe backend.

openal-soft-1.21.0:

Updated library codebase to C++14.

Implemented the AL_SOFT_effect_target extension.

Implemented the AL_SOFT_events extension.

Implemented the ALC_SOFT_loopback_bformat extension.

Improved memory use for mixing voices.

Improved detection of NEON capabilities.

Improved handling of PulseAudio devices that lack manual start control.

Improved mixing performance with PulseAudio.

Improved high-frequency scaling quality for the HRTF B-Format decoder.

Improved makemhr's HRIR delay calculation.

Improved WASAPI capture of mono formats with multichannel input.

Reimplemented the modulation stage for reverb.

Enabled real-time mixing priority by default, for backends that use the
setting. It can still be disabled in the config file.

Enabled dual-band processing for the built-in quad and 7.1 output decoders.

Fixed a potential crash when deleting an effect slot immediately after the
last source using it stops.

Fixed building with the static runtime on MSVC.

Fixed using source stereo angles outside of -pi...+pi.

Fixed the buffer processed event count for sources that start with empty
buffers.

Fixed trying to open an unopenable WASAPI device causing all devices to
stop working.

Fixed stale devices when re-enumerating WASAPI devices.

Fixed using unicode paths with the log file on Windows.

Fixed DirectSound capture reporting bad sample counts or erroring when
reading samples.

Added an in-progress extension for a callback-driven buffer type.

Added an in-progress extension for higher-order B-Format buffers.

Added an in-progress extension for convolution reverb.

Added an experimental Oboe backend for Android playback. This requires the
Oboe sources at build time, so that it's built as a static library included
in libopenal.

Added an option for auto-connecting JACK ports.

Added greater-than-stereo support to the SoundIO backend.

Modified the mixer to be fully asynchronous with the external API, and
should now be real-time safe. Although alcRenderSamplesSOFT is not due to
locking to check the device handle validity.

Modified the UHJ encoder to use an all-pass FIR filter that's less harmful
to non-filtered signal phase.

Converted examples from SDL_sound to libsndfile. To avoid issues when
combining SDL2 and SDL_sound.

Worked around a 32-bit GCC/MinGW bug with TLS destructors. See:
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=83562

Reduced the maximum number of source sends from 16 to 6.

Removed the QSA backend. It's been broken for who knows how long.

Got rid of the compile-time native-tools targets, using cmake and global
initialization instead. This should make cross-compiling less troublesome.