5.8 KiB
5.8 KiB
TO DO
General
- Resolve the issues
- Some of bugs are found when adding tests. Search FIXME to find them.
- Remove
#[allow(non_camel_case_types)]
,#![allow(unused_assignments)]
,#![allow(unused_must_use)]
- Use
ErrorChain
- Centralize the error log in one place
- Support
enumerate_devices
with in-out type? - Monitor
kAudioDevicePropertyDeviceIsAlive
for output device. - Create a wrapper for
CFArrayCreateMutable
like what we do forCFMutableDictionaryRef
- Create a wrapper for property listener’s callback
- Use
Option<AggregateDevice>
rather thanAggregateDevice
foraggregate_device
inCoreStreamData
Type of stream
- Use
Option<device_info>
rather thandevice_info
for{input, output}_device
inCoreStreamData
- Use
Option<StreamParams>
rather thanStreamParams
for{input, output}_stream_params
inCoreStreamData
- Same as
{input, output}_desc
,{input, output}_hw_rate
, ...etc - It would much clearer if we have a
struct X
to wrap all the above stuff and useinput_x
andoutput_x
inCoreStreamData
Separate the stream implementation from the interface
The goal is to separate the audio stream into two parts(modules). One is inner, the other is outer.
- The outer stream implements the cubeb interface, based on the inner stream.
- The inner stream implements the stream operations based on the CoreAudio APIs.
Now the outer stream is named
AudioUnitStream
, the inner stream is namedCoreStreamData
.
The problem now is that we don't have a clear boundry of the data ownership between the outer stream and inner stream. They access the data owned by the other.
audiounit_property_listener_callback
is tied to outer stream but the event listeners are in inner streamaudiounit_input_callback
,audiounit_output_callback
are registered by the inner stream but the main logic are tied to outer stream
Callback separation
- Create static callbacks in inner stream
- Render inner stream's callbacks to outer stream's callbacks
Reinitialization
If the outer stream and the inner stream are separate properly, when we need to reinitialize the stream, we can just drop the inner stream and create a new one. It's easier than the current implementation.
Aggregate device
Usage policy
- BMO 1563475: Only use aggregate device when the mic is a input-only and the speaker is output-only device.
- Test if we should do drift compensation.
- Add a test for
should_use_aggregate_device
- Create a dummy stream and check
- Check again after reinit
- Input only: expect false
- Output only: expect false
- Duplex
- Default input and output are different and they are mic-only and speaker-only: expect true
- Otherwise: expect false
Get sub devices
- A better pattern for
AggregateDevice::get_sub_devices
Set sub devices
- We will add overlapping devices between
input_sub_devices
andoutput_sub_devices
.- if they are same device
- if either one of them or both of them are aggregate devices
Setting master device
- Check if the first subdevice of the default output device is in the list of sub devices list of the aggregate device
- Check the
name: CFStringRef
of the master device is notNULL
Mixer
- Don't force output device to mono or stereo when the output device has one or two channel
- unless the output devicv is Bose QC35, mark 1 and 2.
Interface to system types and APIs
- Check if we need
AudioDeviceID
andAudioObjectID
at the same time - Create wrapper for
AudioObjectGetPropertyData(Size)
with qualifier info - Create wrapper for
CF
related types - Create wrapper struct for
AudioObjectId
- Add
get_data
,get_data_size
,set_data
- Add
- Create wrapper struct for
AudioUnit
- Implement
get_data
,set_data
- Implement
- Create wrapper for
audio_unit_{add, remove}_property_listener
,audio_object_{add, remove}_property_listener
and their callbacks- Add/Remove listener with generic
*mut T
data, fire their callback with generic*mut T
data
- Add/Remove listener with generic
Cubeb Interface
current_device
should be the in-use device of the current stream rather than default input and output device.- Implement
From
trait forenum cubeb_device_type
so we can usedevtype.into()
to getffi::CUBEB_DEVICE_TYPE_*
. - Implement
to_owned
inStreamParamsRef
- Check the passed parameters like what cubeb.c does!
- Check the input
StreamParams
parameters properly, or we will set a invalid format intoAudioUnit
. - For example, for a duplex stream, the format of the input stream and output stream should be same. Using different stream formats will cause memory corruption since our resampler assumes the types (short or float) of input stream (buffer) and output stream (buffer) are same (The resampler will use the format of the input stream if it exists, otherwise it uses the format of the output stream).
- In fact, we should check all the parameters properly so we can make sure we don't mess up the streams/devices settings!
- Check the input
Test
- Rewrite some tests under cubeb/test/* in Rust as part of the integration tests
- Add tests for capturing/recording, output, duplex streams
- Update the manual tests
- Those tests are created in the middle of the development. Thay might be not outdated now.
- Add in-out support for
TestDevicePlugger