Home | About | Community | Download | Documentation | Planet |
PulseAudio 7.0 Release Notes
Changes at a Glance
- LFE channel synthesis with low-pass filtering
- New libsoxr based resamplers
- Socket activation support for TCP
- The "srbchannel" IPC mechanism enabled by default
- More flexible jack detection support when using UCM
- Exiting due to SIGTERM isn't considered a failure
Notes for Application Developers
(nothing important happened)
Notes for Packagers
There's a new optional dependency on libsoxr. The minimum supported version is 0.1.1. libsoxr is used for the soxr-mq, soxr-hq and soxr-vhq resamplers.
Changes in More Detail
LFE channel synthesis with low-pass filtering
When the speaker system has a subwoofer, it's often desirable to synthesize the LFE channel audio when the original audio content doesn't have a separate LFE channel. Previously PulseAudio supported doing that by simply feeding the average of all input channels to the LFE channel, but it's often better to filter out any high frequencies going to the subwoofer. That's now supported, and enabled by default.
The default value of "enable-lfe-remixing" in daemon.conf has changed from "false" to "true", and there's a new daemon.conf option for configuring the crossover frequency: "lfe-crossover-freq". The crossover filtering can be disabled by setting "lfe-crossover-freq" to zero, and if "enable-lfe-remixing" is "true", then the old LFE synthesis algorithm is used.
New libsoxr based resamplers
There are three new libsoxr based resamplers available: soxr-mq, soxr-hq and soxr-vhq. Quality-wise all should be perfect (no audible distortions), but the hq and vhq variants can be used if "even more perfect" quality is desired. The libsoxr developers recommend the hq variant for bit depths up to 16 bits, and the vhq variant for higher bit depths (the internal processing happens with higher precision in the vhq variant). The statement that even the mq variant is "perfect" is based on Alexander E. Patrakov's measurements using a psychoacoustic model for determining whether any produced distortions are actually audible (they're not). The model is described in this paper: http://www.mp3-tech.org/programmer/docs/6_Heusdens.pdf
The mq and hq variants have better performance than speex-float-1 (the current default resampler). We would make soxr-mq the default resampler, if it wasn't for one significant drawback: the resamplers usually introduce up to around 20 ms of extra latency, and in some rare cases even more.
More discussion about the merits and shortcomings of the new resamplers can be found in this mailing list thread: http://thread.gmane.org/gmane.comp.audio.pulseaudio.general/22158
Here are some benchmarking results: http://lastique.github.io/src_test/
Socket activation support for TCP
The previous release introduced support for systemd socket activation for unix sockets. Now the support has been extended to cover TCP sockets as well. This allows PulseAudio to be started on-demand on headless servers, for example.
The "srbchannel" IPC mechanism enabled by default
The "srbchannel" IPC mechanism, which improves efficiency particularly on low latency streams, was introduced in the previous release, but it was too buggy to be enabled by default at that time. Bugs have been fixed since, and now the feature is enabled by default.
More flexible jack detection support when using UCM
Systems that use ALSA's UCM (use case manager) for describing the hardware can now configure the jack kcontrol for a device using the "JackControl" value. Previously, jack detection with UCM relied on implicit heuristics for the jack kcontrol name, which didn't always work well enough. Also, the "JackHWMute" value for UCM devices is now supported, so PulseAudio understands when plugging in a jack for one device makes another device unavailable.
Exiting due to SIGTERM isn't considered a failure
When PulseAudio receives SIGTERM, the process now exits with return value 0. Previously the return value was 1, signaling a failure, which didn't make much sense. If you have e.g. scripts or systemd service definitions that are interested in the process return value, they may need some modification.
Better support for Creative SoundBlaster Omni Surround 5.1
The Creative SoundBlaster Omni Surround 5.1 USB sound card sets up its microphone ALSA device in a way that PulseAudio doesn't support with the default configuration. The card now has its own configuration file in PulseAudio to make the microphone work.
git shortlog
Ahmed S. Darwish (1): pulse: Document client libraries logging behavior Alexander E. Patrakov (4): Make pa_socket_server_new static module-loopback: don't use 0 for custom source output message id echo-cancel: fix the obviously-wrong "buffer+=buffer" logic loopback: Fix the obviously-wrong "buffer+=buffer" logic Andrey Semashev (4): Added libsoxr resampler backend. Enabled libsoxr resampler backend. Added libsoxr detection and optional build of soxr resampler backend. Added documentation for soxr resampling methods. Andrey Zholos (1): detect: Fix OSS device detection with long names Arun Raghavan (17): sink-input: Clarify the meaning of rewrite and flush a bit in rewinds udev: Deal with unavailable bus property protocol-native: Fix source latency calculation in ADJUST_LATENCY mode rescue-streams: Document a cryptic looking condition rescue-streams: Fix a couple of typos stream: Support extended API with PULSE_LATENCY_MSEC if possible sink,source: Add a helper function to check whether this is a filter always-sink: Ignore filter sinks echo-cancel: Handle underlying sink going away better when autoloaded equalizer: Handle underlying sink going away better when autoloaded build-sys: Fix distcheck for systemd user unit dir build-sys: Make echo-cancel use speex cflags/libs conditionally build-sys: Fix up setting the bash completion dir alsa-mixer: Drop redundant conditional frees alsa: Clarify potentially confusing device port data field lfe-filter: Deal with empty input chunks Update NEWS for 7.0 release Cédric Valmary (1): i18n: add Occitan translation David Henningsson (29): alsa-util: Make two of the warnings "debug" instead of "error" pstream: Don't split (non-SHM) memblocks protocol-native: Re-enable srbchannel pstream: Remove unnecessary if condition thread-mq: Make pa_thread_mq_done more robust lfe-filter: Import code from the Chrome OS audio server lfe-filter: Enable LFE filter in the resampler lfe-filter: Cleanup and refactor memblock: Change pa_memblock_new_malloced to an inline function lfe-filter: Add rewind support resampler: Make some basic functions for rewinding resampler: Allow disabling the LFE filter by setting crossover to 0 resampler: Rename "lfe_filter_required" to "lfe_remixed" module: Add hook dynarray module-*-restore: use pa_module_hook_connect module-suspend-on-idle: use pa_module_hook_connect module-alsa-card: use pa_module_hook_connect module-filter-*: use pa_module_hook_connect module-switch-*: use pa_module_hook_connect alsa-mixer: Use separate profiles for Multichannel Input/Output alsa-mixer: Add paths/ports for multichannel profile core-util, pactl: Make one localised and one non-localised version of pa_yes_no alsa-mixer: Make line out path unavailable when "Front Headphone" is plugged in lfe-filter-test: Use S16NE format echo-cancel: Add config.h in adrian-aec.c stream: Check sample spec validity before abiding PULSE_LATENCY_MSEC module: Warn about modules loaded after unload alsa-mixer: Add "Front Line Out" and "Rear Line Out" module: Always remove freed module from modules_pending_unload David Herrmann (1): core-util: Fix set_nice() to use private bus connections Deepak Srivastava (3): padsp: Avoid possible deadlock. Unlocked i->mainloop before returning from default case. patcl: Error msg from pactl not quite accurate for 'mute' commands Removed exclamation marks from user-visible messages. Dušan Kazik (2): i18n: update Slovak translation i18n: Update Slovak translation Felipe Sateler (5): Add a .travis.yml for Travis CI man: fix reference to --daemonize option Fix return code check of pa_append_to_home_dir util: pa_get_fqdn: always free addrinfo sink, source: Do not dereference freed memory when freeing the next events Hui Wang (5): lfe-filter: change the crossover frequency as a parameter tests: adding lfe-filter-test daemon-conf: enable the lfe remixing by default tests: Remove a redundant mempool_new from resampler-test tests: add tolerant variation for comparing the rewind result Jaska Uimonen (1): socket-server: add tcp support for systemd socket activation Juho Hämäläinen (17): dbus: Use hooks for card profile events dbus: Use hooks for client proplist changes dbus: Use hooks for sink and source events dbus: Use hooks for sink-input and source-output events dbus: Use state changed hook for sink-input and source-output rate changes core: Add hooks for default sink and source changes namereg: Fire hooks for default sink and source changes dbus: Use hooks for put and unlink dbus: Use hooks for default sink and source changes core: Add core hooks for module changes module: Fire hooks for new, proplist changed and removed core, core-scache: Fire hooks for new, changed and removed events dbus: Use hooks for sample cache new and removed events dbus: Use hook for sample cache proplist changes dbus: Use hooks for module new and removed events dbus: Use hooks for module proplist changes Revert "Warn on loading module-dbus-protocol" Manish Sogi (2): xen: Resource leak in local function alloc_gref() - alloc_fd and dev_fd pacmd: Placeholders describing command line options should not have spaces Michael Cree (1): tests: Fix test-suite failure on Alpha Milo Casagrande (1): Update Italian translation. Muhammet Kara (1): i18n: Updated Turkish translation Nazar Mokrynskyi (1): alsa-mixer: Add support for Creative SoundBlaster Omni Surround 5.1 USB sound card Paul Menzel (1): lfe-filter: Use length modifier `z` for type `size_t` Peter Meerwald (29): context: pa_context_connect() with PA_CONTEXT_NOFAIL should return 0 when waiting for PA on D-Dus man: Describe optional prefix {string} in server address man: Drop separate paplay man page, move info to pacat man page man: Install man page symlinks to pacat for paplay, parec, parecord, pamon pacat: State purpose of program depending how it was invoked alsa-util: No logging when sound card only supports non-interleaved sample format alsa: Fix spelling of officially tagstruct: Distinguish pa_tagstruct_new() use cases tagstruct: Replace dynamic flag with type tagstruct: Get rid of pa_tagstruct_free_data() tagstruct: Add type _APPENDED tagstruct: Use flist to potentially save calls to malloc()/free() packet: Hide internals of pa_packet, introduce pa_packet_data() packet: Make pa_packet_new() create fixed-size packets packet: Introduce pa_packet_new_data() to copy data into a newly created packet packet: Use flist to save calls to malloc()/free() pacat: Fix comment wording man: Fix typos in pulse-cli-syntax.5 man: Document .ifexists CLI meta command daemon: Exit with code 0 on SIGINT and SIGTERM pulse: Document pa_mainloop_quit()'s retval parameter core: Don't export pa_rtpoll_quit(), only used internally biquad: Make use of M_SQRT2 constant biquad: Fix warning, gamma shadows global declaration padsp: Fix wrong condition discovered by -Wlogical-not-parentheses warning core: Work around -Wlogical-not-parentheses warnings daemon: No need to check optarg, -p requires argument core: Add missing return on protocol error modules: Fix entry leak in module-card-restore Pierre Ossman (1): core: make sure we use the correct win32 socket errno:s Piotr Drąg (1): i18n: Update Polish translation René J. V. Bertin (1): coreaudio: Change error type to OSStatus Sagar Nageshmurthy (1): Fix: Prevent calling pa_rtpoll_free() for a NULL rtpoll Shawn Walker (1): modules: Fix struct namespace collision on Solaris Sukesh Adiga (1): cpu-arm: Fix memory leak in pa_cpu_get_arm_flags() Takashi Sakamoto (1): udev: use ID_MODEL/ID_VENDOR to give friendly name for FireWire devices Tanu Kaskinen (33): socket-server: pa_socket_server_new() can't fail, so don't check its return value dynarray: Add pa_dynarray_last() dynarray: Add pa_dynarray_remove_by_index() dynarray: Add pa_dynarray_remove_by_data() dynarray: Add PA_DYNARRAY_FOREACH tagstruct: Refactor writing/reading basic types alsa: Don't access pa_sink/source_new_data after done() has been called core-util: Make number parsing stricter ucm: Add support for "JackControl" .gitignore: Add lfe-filter-test sink, source: Fix a volume change leak pactl: Fix relative volume parsing core-util: Filter out not-a-numbers in pa_atod() core-util: Make pa_parse_volume() more strict oss: Use the correct fd when interacting with the mixer build-sys: add winerrno.h to libpulsecommon_SOURCES combine-sink: Fix unsafe message handling alsa-mixer: Add descriptions for analog-stereo-input and analog-stereo-output mappings build-sys: make error message visible alsa: Add pa_alsa_jack_new() and pa_alsa_jack_free() ucm: Don't create separate input and output jacks alsa: Add associations between jacks, UCM devices and UCM ports alsa: Move UCM port availability updating to the mixer code ucm: Create only one jack object per kcontrol ucm: Add support for "JackHWMute" build-sys: bump sonames update NEWS for 7.0 alsa-mixer: sb-omni-surround-5.1.conf: rename analog-stereo to analog-stereo-output alsa-mixer: sb-omni-surround-5.1.conf: remove analog-surround-40 mapping build-sys: check strtod_l instead of strtof_l core-util: include xlocale.h when using strtod_l() build-sys: bump sonames build-sys: add sb-omni-surround-5.1.conf to Makefile.am Ville Skyttä (1): build-sys: Install bash completion to where bash-completion.pc says Wim Taymans (1): Add korean translation Yuri Chornoivan (1): i18n: Update Ukrainian translation Zavadovsky Yan (1): build-sys: fix order of libraries installation