[[!img Logo]

HomeAboutCommunityDownloadDocumentationPlanet

[[!format rawhtml """

!html


"""]]

PulseAudio 3.0 Release Notes

We're, back with another shiny PulseAudio release! While the 3.0 release was a little delayed, it brings a number of important improvements, and bug fixes. A summary of changes follows.

Notable Changes

  • ALSA Use Case Manager (UCM) support
  • Runtime editable LADSPA filter parameters
  • Out-of-the-box support for Bluetooth sources
  • ARM NEON optimisations
  • Configurable device latency offset
  • Adhere to the XDG Base Directory Specification
  • Various ALSA changes
  • Lots of infrastructure improvements

Packaging

  • Bluetooth support requires now "sbc", a library for the SBC codec. The codec used to be included within PulseAudio, but it has now been split off into a separate library. It's available at http://www.bluez.org.
  • Support for the "socket API" of BlueZ has been dropped in favour of the D-Bus based "media API". Due to this change, the minimum supported version of BlueZ is now 4.99. Also, make sure that you don't have "Disable=Media" in /etc/bluetooth/audio.conf. And due to a bug in BlueZ, it's probably necessary to have "Disable=Socket", otherwise there will be problems with the A2DP profile.
  • Support for HAL has been removed. This shouldn't affect anyone, but if it does, please configure PulseAudio to use udev instead. module-hal-detect still exists for maintaining configuration file compatibility, but all it does is to load module-udev-detect. module-hal-detect may get completely removed in 4.0.

Changes in More Detail

ALSA Use Case Manager (UCM) Support

The audio hardware on desktop and laptop computers is usually standard enough for PulseAudio to handle using a common set of configuration files. The situation is different on e.g. phones and tablets. Practically all of those devices need separate configuration files for describing the hardware so that PulseAudio can properly use all basic features of the hardware. The configuration could be shipped as PulseAudio configuration files, but the ALSA folks have been working on a system that allows applications (such as PulseAudio) to use the hardware without needing any extra configuration in the application. The system is called UCM, and PulseAudio now has support for it.

Runtime Editable LADSPA Filter Parameters

The LADSPA module now exposes a basic D-Bus interface for changing the filter parameters on the fly. Previously the parameters could not be changed after loading the filter.

Out-of-the-box support for Bluetooth sources

PulseAudio can be used in a Bluetooth headset role, for example to connect a laptop to a mobile phone and pretend that the laptop is a headset. It's often desirable in that case to loop back the audio from the phone to the laptop's sound card. That is now done automatically by module-bluetooth-policy, which is loaded by default. Users do need to enable Source support in their BlueZ configuration, though.

ARM NEON optimisations

Optimisations were added for sample format conversion between S16LE and floating point formats using the ARM NEON instructions. Support for these is detected at compile-time (based on FPU flags) as well as run-time (based on /proc/cpuinfo). As part of this effort, the groundwork has been laid for adding more NEON optimisations in the future.

Configurable Device Latency Offset

Accurate latency reporting is important for e.g. "lip sync" in video playback. PulseAudio relies on the audio hardware to provide accurate information about the audio delay. If that information is not accurate for some reason, it is now possible to configure an offset to be applied to each latency report, thanks to Damir Jelić's Google Summer of Code work. For example, if you're watching a video and you use a Bluetooth headset for audio output, the lip sync might be a bit off, because PulseAudio doesn't currently have proper support for querying the latency for Bluetooth devices. With the upcoming version of pavucontrol, you will be able to fix the synchronization problem by adjusting the latency offset of the Bluetooth headset.

Adhere to the XDG Base Directory Specification

The location of configuration files has been moved from ~/.pulse to ~/.config/pulse (or if $XDG_CONFIG_HOME is set, then use that). If ~/.pulse exists, however, it will still be used so that the user configuration is not lost when updating PulseAudio. The authentication cookie has also been moved from ~/.pulse-cookie to ~/.config/pulse/cookie.

The location of runtime files (i.e. files that don't need to be kept across reboots, e.g. sockets) has been moved from a random directory under /tmp to $XDG_RUNTIME_DIR/pulse. In case $XDG_RUNTIME_DIR is not set, the old scheme is still used as a fallback.

Various ALSA Changes

A regression in 2.0, that could cause some machines not to have "Speaker" or "Internal Mic" ports, has been mostly fixed in 3.0 - when used together with Linux 3.6 or higher. Also, a workaround for older kernels is applied to certain most common machines.

Pierre-Louis Bossart introduced a change to improve accuracy of timestamp queries, and thus timer-based scheduling, by querying a number of ALSA timing parameters atomically.

We have added icon name property to ALSA ports, which could help UI makers to display better per-port icons.

Also, there has been a few other changes, such as removing the troublesome lfe-on-mono port, and adding more mixer control names to better stay in sync with the kernel.

Improvements for Automatic Testing

Deng Zhengrong improved PulseAudio's automatic testing support as part of his Google Summer of Code work. The improvements include support for test coverage reporting with gcov, changing the existing tests to use the "check" framework, and making it possible to launch a PulseAudio daemon for testing purposes while simultaneously having the normal daemon running.

Unloading Modules by Name

pactl and pacmd now support unloading modules by name. Previously unloading could only be done by the module index, which was a bit inconvenient, because usually you don't know the index without somehow looking it up first.

Logging Improvements

In addition to the automatic testing improvements, GSoC student Deng worked also on PulseAudio's logging facilities. It's now possible to change the log target of a running daemon with pacmd, using the set-log-target command. Also, a new log target type was added: "newfile". The "newfile" target is like the old "file" target, with the difference that if the given file name already exists, the file is not overwritten, but instead a new file is created with a numbered suffix.

Bluetooth Code Refactoring

There has been a lot of refactoring work done on the Bluetooth modules, as preparation for the upcoming BlueZ 5 support and for making the code more pleasant to work with in general. These are not directly user-visible changes, but we'd like to take the opportunity here to thank Mikel Astiz anyway for the great work he has done.

Next Steps

The development continues as always, and the 4.0 release is targeted to happen in April (so far we haven't been very good at maintaining a 4 month release cycle, though, so take that with a grain of salt). The report from PulseConf 2012 offers some clues about what might be coming next.

git shortlog

[[!format txt """ Andika Triwidada (1): i18n: Add Indonesian translation.

Arti Trivedi Bora (4): modules: Use pa_streq instead of strcmp. pulsecore: Use pa_streq instead of strcmp. tests: use pa_streq instead of strcmp daemon: use pa_streq instead of strcmp

Arun Raghavan (68): sink-input,source-output: Avoid unneccessary rate updates role-cork: Fix a minor leak core-util: Add a pa_split_in_place() string utility function core-util: Make pa_make_secure_dir() act like mkdir -p auth: Create cookie directory if it doesn't exist pactl: Print sink-input/source-output corked status glib: Stop using g_source_get_current_time() core-util: Fix permissions handling while creating directories Revert "role-cork: Fix a minor leak" role-cork: Fix incorrect check at deinitialisation time Revert "Revert "role-cork: Fix a minor leak"" role-cork: Fix another minor leak alsa: Allow channel count probe on open by mapping alsa: Add a proplist to mappings alsa: Add separate sinks/sources for UCM modifiers if needed i18n: Remove module-hal-detect reference in POTFILES echo-cancel: Print what AEC engine is being used build: Avoid libstdc++ dep for module-echo-cancel if possible i18n: module-coreaudio-device now has some translatable strings build: Fix distcheck failure on libwebrtc-util tests: Factor out Orc test code into cpu-test tests: Make cpu-test less verbose build: Merge bluez pkg-config checks into one Revert "build: Merge bluez pkg-config checks into one" stream: Allow record streams to start muted core: Separate ARM CPU detection from initialisation tests: Factor out ARM svolume test into cpu-test tests: Minor cpu-test reorganisation core: Document ARM-optimised svolume code a bit core: Fix a litte-endian bug in ARM svolume code tests: Add a copyright header to cpu-test tests: Factor out core sconv test code in cpu-test tests: Reorganise cpu-test to reuse code tests: Add a basic sanity test to sconv cpu-test tests: Run sconv tests with multiple alignments build-sys: Add volume code to libpulsecommon sconv: Fix NEON sconv rounding code tests: Allow off-by-one error in sconv test tests: Increase sconv cpu-test timeout tests: Print average outer-loop iteration time in cpu-test tests: Minor cpu-test fixes for non-NEON builds core: Fix warning on non-win32 builds tests: Run svolume test for various sample alignments tests: Make cpu-test less verbose tests: Run svolume on different channel counts tests: Fix a cpu-test debug message build-sys: Drop -Wvla from compiler flags tests: Minor alignment adjustment fix for cpu-test svolume: Fix ARM alignment issues Revert "core: adjust playing_for and underrun_for at rewind" build-sys: Bump soname i18n: Fix POTFILES for poll changes build-sys: Document libpulsecommon vs. libpulse duplication alsa: Drop verbosity on UCM message Revert "tests: modify alsa-time-test to use 'check' framework" introspect: Minor documentation fix man: Correction for how sample rate switching is disabled sink, source: Prevent unnecessary rate update attempts modules: Micro-optimisation for rewind_requested paths build-sys: Bump BlueZ dependency to 4.99 build-sys: Bump soname build-sys: Drop ChangeLog generation source-output: Fix volume fixup for rate update sink-input, source-output: Check rate update success for passthrough core: Remove bad free() call alsa: Try to support non-standard rates in alsa-sink/source build-sys: Bump soname build-sys: Bump soname

Chan-yeol Park (1): bluetooth: Remove ipc.[ch] files in the bluetooth module

Colin Walters (1): git-version-gen: Honor GIT_DESCRIBE_FOR_BUILD environment variable

David Henningsson (22): once: Fix race causing pa_once to sometimes run twice alsa-mixer: Add special profiles for some laptops missing speaker and/or internal mic alsa-mixer: Add Phantom Jack support alsa-mixer: Always turn "Inverted Internal Mic" off alsa-mixer: Add "Front Headphone" jack alsa-mixer: Document "state.plugged" and "state.unplugged" alsa-mixer: Add "Front Headphone Jack" (fixup) alsa-mixer: Add "Headphone Mic" support for 3-pin ASUS netbooks resampler: Fix volume on downmix to mono alsa-mixer: Add "iec958-stereo-input" to well known path names flist: Increase default list size to 256 alsa-mixer: Cache failure to open inputs/output mappings alsa-mixer: Remove analog-output-lfe-on-mono cli: Output asterisk when default sink/source is found alsa-sink/source: Warn for scheduling delays alsa-mixer: Don't let "Mic Jack Mode" alone create a "Line In" path alsa-mixer: Add a few more machines to internal mic whitelist alsa-mixer: Add "CLFE" and "Bass Speaker" names alsa-mixer: Prefer "Digital Input Source:Digital Mic 1" alsa udev quirks: Add some more Dell devices to speaker whitelist alsa-mixer: Add Dell Inspiron One 2020 to mic whitelist alsa-mixer: Add device.icon-name property for some common ports

Deng Zhengrong (48): cli: Add set-log-target command for pacmd daemon: use pa_streq instead of plain strcmp x11: fix the wrong parameter sequence in pax11publish build: Add gcov coverage support pacmd: add help info for 'set-log-target' xen: add the HAVE_CONFIG_H macro guard add a new log target that enables to create new log file if it exists bluetooth: add a parenthesis around pa_streq() man: document option set-log-target core: add more verbose error info build-sys: add check test framework tests: modify mix-test to use new 'check' test framework tests: add cpu test tests: modify mainloop-test to use new 'check' framework tests: modify utf8-test to use new 'check' test framework tests: modify strlist-test to use new 'check' framework build: add a target to ease the creation of coverage files tests: enable to test standalone pulseaudio daemon tests: modify volume_test to use new 'check' framework tests: modify usergroup-test to use 'check' test framework tests: modify format-test to use 'check' framework tests: modify get-binary-name-test to use 'check' framework tests: modify thread-test to use 'check' framework tests: modify thread-mainloop-test to use 'check' framework tests: modify alsa-time-test to use 'check' framework tests: modify asyncmsgq-test to new 'check' framework tests: modify asyncq-test to use 'check' framework tests: modify channelmap-tets to use 'check' framework tests: modify cpulimit-test to use 'check' framework tests: modify queue-test to use 'check' framework tests: modify connect-stress to use 'check' framework tests: modify memblock-test to use 'check' framework tests: modify proplist-test to use 'check' framework tests: modify memblockq-test to use 'check' framework tests: modify hook-list-test to use 'check' framework tests: modify extended-test to use 'check' framework tests: modify sync-playback to use 'check' framework tests: modify smoother-test to use 'check' framework tests: modify interpol-test to use 'check' framework tests: modify sigbus-test to use 'check' framework tests: modify sig2str-test to use 'check' framework tests: modify rtpoll-test to use 'check' framework tests: modify lock-autospawn-test to use 'check' framework tests: modify once-test to use 'check' framework tests: modify ipacl-test to use 'check' framework tests: fix the wrong library path in check-daemon tests: make 'check' optional build: fix Mac OS X configure process

Eero Nurkkala (2): alsa-sink: add missing header 'signal.h' alsa-source: add missing header 'signal.h'

Feng Wei (3): alsa: Integrate UCM basic functions alsa: Add UCM jack detection alsa: Catch role matched streams to enable/disable modifier

Flavio Ceolin (4): sink-input: Remove redundant check in pa_sink_input_request_rewind(). modargs: New function: pa_modargs_get_value_double(). pulse: Fix for volume documentation utils: Adding a function to get volume from string

Frédéric Dalleau (6): loopback: Enable routing on loopback streams bluetooth: module-bluetooth-policy initial commit pacmd: Display inputs and outputs PASSTHROUGH flag loopback: Cork sink-input if source is suspended loopback: Cork source-output if sink is suspended resampler: Fix crash if 'auto' resampler chooses ffmpeg with variable rate

Frédéric Danis (2): bluetooth: Fix crash on disconnection bluetooth: Fix bluetooth.nrec property not updated

Harsh Prateek Bora (2): modules: Use PA_IDXSET_FOREACH wherever applicable. pulsecore: Use PA_IDXSET_FOREACH wherever applicable.

Ismo Puustinen (2): ladspa: D-Bus interface for setting algorithm parameters on-the-fly. ladspa: Added a python script for testing.

Jan Alexander Steffens (heftig) (1): alsa-mixer: Actually install analog-input-headphone-mic.conf

Jarkko Nikula (2): alsa: move pa_alsa_setting_select close to pa_alsa_path_select alsa: Merge pa_alsa_setting_select with pa_alsa_path_select

Jarkko Suontausta (1): bluetooth: Release transport when the pa_rtpoll_run loop finishes.

Juho Hämäläinen (1): stream-restore: Add missing method handler argument.

Lennart Poettering (14): util: XDG_SESSION_COOKIE is unsuitable as session ID proplist: document new meaning of the session ID util: /etc/machine-id should be tried first, the D-Bus only as fallback for legacy systems util: use the return value of gethosid() as fallback, not the address of the function build-sys: readd stub makefiles to subdirectories to make building with emacs easier build-sys: remove HAL support, it's obsolete since years rtkit: update drop-in files context: get rid of really old runtime dir logic, i.e. break compat with >4y-old PA util: hook up pa_get_runtime_dir() with XDG_RUNTIME_DIR core-util: move configuration home directory from ~/,pulse to ~/.config/pulse to follow XDG basedir spec man: update man pages to only refer to the new place for the configuration files core-util: when searching for configuration files, honour XDG basedir spec auth: move cookie file to ~/.config/pulse/cookie gnome: start PA early in the gnome session

Luiz Augusto von Dentz (2): bluetooth: Remove built-in/static SBC codec bluetooth: Don't force any profile on discovery module

Marc-Antoine Perennou (1): udev: Don't use deprecated udev_get*path() functions

Martin-Éric Racine (1): manpage, finnish translation: fix spelling errors

Matthijs Kooijman (1): equalizer: Don't cleanup u->sink in sink_input_kill_cb yet

Mihai Moldovan (1): core-util: use the generic PATH_MAX variant of pa_realpath on Mac OS X

Mikel Astiz (67): bluetooth: Don't use the old socket IPC mechanism with BlueZ bluetooth: Refactor property parsing code bluetooth: Remove library for IPC to BlueZ bluetooth: Minor style fixes bluetooth: Consider different input and output MTU bluetooth: Avoid duplicating profile argument twice bluetooth: Replace deprecated ListAdapters() bluetooth: Replace deprecated ListDevices() bluetooth: Remove minor unnecessary check bluetooth: Minor style fix bluetooth: Fix missing state checks for a2dp_source bluetooth: Fix bluetooth.protocol property bluetooth: Trivial style fix bluetooth: Generalize module-bluetooth-policy bluetooth: Support HFGW in module-bluetooth-policy loopback: Disable adjust timer when suspended bluetooth: Remove return value of bt_transport_config() bluetooth: Remove return value of setup_stream() bluetooth: Refactor parsing of signal PropertyChanged bluetooth: Refactor code to helper function bluetooth: Fix wrongly set "phone" role for HFGW bluetooth: Fix using garbage memory bluetooth: Fix check if transport exists before acquire sink, source: Support creating suspended sinks and sources bluetooth: Provide dummy set_port callbacks bluetooth: Config MTU transport after acquire bluetooth: Support port availability flag bluetooth: Set profile even if transport not acquired bluetooth: Do not acquire transport during profile change bluetooth: Acquire transport when becomes available bluetooth: Release transport when not available bluetooth: Do not switch to HFGW automatically bluetooth: Let suspend-on-idle request audio in headset bluetooth: Add port availability transition policies bluetooth: Ignore Device.DisconnectRequested bluetooth: Trivial function rename bluetooth: Fix potential assertion failure bluetooth: Don't find device if set profile is off bluetooth: Release transport in stop_thread() bluetooth: Unlink sink-sources in stop_thread() bluetooth: Remove stream moving code bluetooth: Use assertions when setting off profile bluetooth: Check return value of init_profile() bluetooth: Check return value of start_thread() bluetooth: Remove const qualifier for transports bluetooth: Add hook to tell transport was removed bluetooth: Set to off if transport removed bluetooth: Set to off instead of failing module load bluetooth: Hold transport pointer while profile set bluetooth: Remove const qualifier for device bluetooth: Add hook to tell device was removed bluetooth: Self unload module-bluetooth-device bluetooth: Hold device pointer while module loaded bluetooth: Refactor code to create card profiles bluetooth: Refactor code to create profile ports card: Support adding profiles dynamically card: Support adding ports dynamically bluetooth: Add hook to announce late UUIDs bluetooth: Rename former device_is_audio() bluetooth: Run the discovery hook only when necessary bluetooth: Merge headset ports into one bluetooth: Disable profile auto-switch policy for headsets conf: Load bluetooth-policy module by default bluetooth: Trivially refactor to call setup_stream() directly bluetooth: Do not setup stream before thread starts bluetooth: Request headset audio during profile switch bluetooth: Fix unacquired transports during sink resume

Niels Ole Salscheider (3): virtual-surround: Add silence to hrir if necessary. virtual-surround: Limit the number of hrir samples. virtual-surround: check if resampled memblock is not equal to input

Paul Menzel (1): Fix spelling of separated: s, sepera, separa, g

Peter Meerwald (13): pipe: whitespace and log output cleanup pipe: check return value of mkfifo() memblock: Fix typos. build: Fix static linking core: Set volumes const in pa_do_volume_func_t modules: Add null/dummy echo canceller daemon: Fix redundant redeclaration warning svolume_arm: Fix a const warning. echo-cancel: Fix false warning in webrtc AEC. rtp: Fix warning using pa_assert_not_reached() core: Add ARM NEON optimized sample conversion code tests: Fix test description in cpu-test tests: Implement test code for ARM NEON sconv s16_to_float

Pierre-Louis Bossart (1): alsa: get avail, delay, timestamps in a single kernel call

Sjoerd Simons (2): build-sys: webrtc-utils needs to be installed before module-echo-cancel build-sys: Correct bluez support error if sbc is missing

Sjors Gielen (1): osx: Add a single "On" profile to coreaudio devices. Fixes crash on OS X.

Tanu Kaskinen (51): device-port: Create the profiles hashmap at initialization. device-port: Remove an out-of-date comment. native: Don't save device, volume or mute of new streams. sink, source: Fix setting the latency offset when the sink/source is unlinked. Fix a copy-paste error in PROTOCOL. pulse: Use more intuitive indexing with port infos in introspect.c. proplist: Change proplist_name_valid() to be public function pa_proplist_key_valid(). tagstruct: Allow NULL proplist with pa_tagstruct_get_proplist(). conf-parser: Pass parser state in a struct instead of function parameters. conf-parser: Pass parser state in a struct also for parse callbacks. conf-parser: Add support for parsing property lists. alsa-mixer: Add support for defining port property lists in the path configuration files. native: Send the actual port proplists with card info. pactl: Print card port properties with the "list" command. card: Don't crash if someone gives NULL name to pa_card_set_profile(). sink, source: Always create a hashmap for ports. card: Ensure that there's always at least one profile. Assume that the profiles hashmap of ports is always non-NULL. Assume that the ports hashmap of cards is always non-NULL. card-restore: Handle reading NULL profile name from the database. alsa-mixer: Implement a new path option: "mute-during-activation". build-sys: Link utf8-test to libpulsecommon. Add comments referring to bug #53709. memblock: Add pa_memblock_acquire_chunk(). object: Get rid of "warning: cast increases required alignment of target type" sink-input: Fix comment: s/push/peek/ sink-input: Add a comment in pa_sink_input_request_rewind(). sink: Remove an incorrect FIXME comment. bluetooth: Remove commented out code. .gitignore: Add cpu-test. rtp: Fix rtp_port reading. card: Store a pa_card pointer in pa_card_profile. build: Add a2dp-codecs.h to libbluetooth-util sources. memblockq: Fix the order of setting minreq and prebuf. resampler: Make sure that there are no overflows when multiplying potentially big numbers. loopback: Use the real sample spec once it's known. loopback: Don't fix the source output format/rate/channels. virtual-surround-sink: Fix setting max_request and max_rewind. combine: Keep the timer active in the null mode only when running. match: Use the SINK_INPUT_FIXATE hook instead of NEW. build: Add .gitignore files to EXTRA_DIST. device-restore: When restoring volume, print the restored volume to the log. build: Add PROTOCOL to EXTRA_DIST. pulse: Fix hole handling in pa_stream_peek(). mainloop: Change wakeup_requested type from pa_bool_t to pa_atomic_t. mainloop: Don't care about the mainloop state variable when waking up the mainloop. sink: Process rewind requests also when suspended. bluetooth: Add a pa_bluetooth_discovery pointer to pa_bluetooth_device. bluetooth: Ignore Device.Connected bluetooth: Add helper pa_bluetooth_device_any_audio_connected() bluetooth: Unload device module when no audio profiles connected

Thomas Martitz (7): pacat: Enable binary mode on Windows. core: Transparently handle non-blocking sockets on Windows pacat: Replace read(), write() with pa_* equivalent. core: Slightly more helpful pa_cstrerror for unknown errors gccmacro: Disable printf-like format checking on mingw32 compilers. core: Proper poll() emulation to fix pacat and friends on Windows core-util: Don't error out on existing runtime directory.

Uoti Urpala (2): sink-input: Fix underrun_for calculation when resampling. core: adjust playing_for and underrun_for at rewind

Wieland Hoffmann (1): man pulse-daemon.conf: Correct typoes

poljar (8): pacmd: Added --help and --version descriptions to the man page. native: Use foreach to iterate trough modules. pactl: Add unloading modules by name. pacmd: Add unloading modules by name. bluetooth: Add ports to the bluetooth sink/source device-port: Add a latency variable to the port struct sink, source: Add a latency offset which is inherited from the port pacmd: Add functions to handle the latency offset

poljar (Damir Jelic) (2): introspect: Add functions to handle the latency offset. pactl: Add set-latency-offset command.

poljar (Damir Jelić) (5): device-port: Change the latency offset type to a signed int. conf-parser: Initialize the state to zero immediately. device-port: Send a subscription event when the offset changes. man: Add latency offset documentation to the cli syntax. card-restore: Add the ability to save and restore the latency offset. """]]