[pulseaudio-discuss] [PATCH v2] Add a news file

[David Henningsson]

On 05/30/2013 11:54 AM, Tanu Kaskinen wrote:
> Changes in v2:
>   * Renamed "new optimizations" to "performance improvements".
>   * Added the s16 <-> s32 conversion note.
>   * Added the default resampler change note.
>   * Added the remixing note.
> ---
>   NEWS | 122 +++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++
>   1 file changed, 122 insertions(+)
>   create mode 100644 NEWS
>
> diff --git a/NEWS b/NEWS
> new file mode 100644
> index 0000000..6091bf5
> --- /dev/null
> +++ b/NEWS
> @@ -0,0 +1,122 @@
> +PulseAudio 4.0, not yet released:
> +
> +  Utility changes:
> +
> +   * Shell completion is now available for most of the utility
> +     programs. Bash and zsh are supported.
> +
> +   * pactl now supports toggling mute. To be clear, setting the mute
> +     to on or off was already supported before, this change adds the
> +     possibility to give "toggle" as a parameter to the set-*-mute
> +     commands.
> +
> +   * New pactl commands: set-default-sink and set-default-source.
> +
> +  Module changes:
> +
> +   * New module: module-role-ducking. The module lowers the volume of
> +     less important streams when a more important stream appears, and
> +     raises the volume back up once the important stream has finished
> +     (this is called "ducking"). The decision whether a stream has
> +     high or low priority is made based on the stream role (the
> +     media.role property). By default, "music" and "video" streams are
> +     ducked, and "phone" streams trigger the ducking. This module is
> +     not loaded in the default configuration.
> +
> +   * New module: module-remap-source. The module creates a virtual
> +     source on top of another source, with a different channel map.
> +     This allows swapping left and right channels, for example.
> +
> +   * module-udev-detect: New module argument: use_ucm. PulseAudio uses
> +     ALSA UCM configuration by default if it's available. This module
> +     argument can be used to disable UCM.
> +
> +   * module-udev-detect: New module argument: tsched_buffer_size.
> +     Buffer size in bytes when timer based scheduling is enabled.
> +
> +   * module-rtp-send: The "destination" argument was renamed to
> +     "destination_ip".
> +
> +   * module-rtp-send: New module argument: source_ip. On a multi-homed
> +     system, it may be desirable to use RTP only on a specific
> +     interface. The interface can be selected by specifying the source
> +     IP address.
> +
> +  Hardware support changes:
> +
> +   * Both outputs of Native Instruments Traktor Audio 2 can now be
> +     used simultaneously without any manual configuration.
> +
> +   * Various improvements for Intel HDA based sound cards.

This sounds very vague. What is this referring to?


> +
> +   * Capture support fixed for Logitech B530 USB headset.
> +
> +  libpulse changes:
> +
> +   * New function: pa_operation_set_state_callback(). It allows an
> +     application to get a callback from pa_operation when it's
> +     cancelled (perhaps due to disconnection).
> +
> +   * PA_ENCODING_MPEG2_AAC_IEC61937 added to the pa_encoding_t
> +     enumeration. This means that it's now possible to play MPEG-2 AAC
> +     streams in the passthrough mode to devices that support it.
> +
> +   * HDMI / DisplayPort monitor name is now available to clients. If
> +     supported by HDMI / DisplayPort hardware and drivers, the name of
> +     the connected monitor, receiver etc. will now show up as a port
> +     property (with key PA_PROP_DEVICE_PRODUCT_NAME). Note that the
> +     name might become available slightly after the port becomes
> +     available (this is due to hardware/driver limitations). A card
> +     change event will be sent out when this information changes.
> +
> +   * Removed function: pa_format_info_free2(). The
> +     pa_format_info_free2 symbol was never part of the public API, but
> +     was still exported in the ABI in the previous version of
> +     PulseAudio. It is now completely dropped. Since it was never
> +     exposed in any header files, the only programs affected should be
> +     test tools specifically looking for missing symbols.
> +
> +  Dependency changes:
> +
> +   * The D-Bus version requirement was bumped to 1.4.12.
> +
> +   * The alsa-lib version requirement was bumped to 1.0.24.
> +
> +   * The GTK+ version requirement was bumped to 3.0.
> +
> +   * The libsystemd-daemon dependency was dropped.
> +
> +  Performance improvements:
> +
> +   * Faster C code for mixing 16-bit streams with 1 or 2 channels.
> +
> +   * ARM NEON implementation for mixing 16-bit streams with 2
> +     channels.
> +
> +   * Faster software volume for 16-bit streams on 64-bit platforms.
> +
> +   * Faster format conversion between 16-bit and 32-bit (integer)
> +     audio.
> +
> +   * The default resampler was changed from speex-float-3 to
> +     speex-float-1, so the resampling is now much lighter on CPU by
> +     default than before, with the tradeoff that the audio quality
> +     is worse (usually not noticeably so, though).
> +
> +  Bug fix highlights:
> +
> +   * Quicker drains: in previous versions of PulseAudio, it could
> +     sometimes take up to a few seconds after a stream finished
> +     playing until pa_stream_drain() actually returned. This has now
> +     been fixed, which also means that programs such as paplay will
> +     end quicker after having finished playback.
> +
> +   * Issues with device handover to JACK were fixed. JACK should be
> +     now able to reliably acquire the sound card while PulseAudio is
> +     running.
> +
> +   * When converting between different channel maps, the remixing
> +     can't result in clipping anymore. Streams that earlier could
> +     potentially get clipped are now a bit quieter.

Btw, do you think the low-latency commits are worth mentioning? E g,

  * A bug caused PulseAudio to request data too late in low-latency 
scenarios (unless minreq was specified by the client), causing dropouts. 
If the total latency requested is below 80 ms, we now default to 
requesting more data four times as often as the requested latency.

(feel free to rephrase)

> +
> +  Contributions from: (TODO)
>

Looks good otherwise, thanks for writing it up!


-- 
David Henningsson, Canonical Ltd.
https://launchpad.net/~diwic