With my previous patch, after an underrun occurred sometimes polling the
alsa FDs would always return POLLIN immediately, but avail_frames would
be zero. Eventually avail_frames would become nonzero and audio output
would work, but poll would still return immediately afterwards.
Calling `stream.prepare()` when this happens seems to fix the issue after
a few loops.
1) warning: use of deprecated item 'std::error::Error::description': use the Display impl or to_string()
2) warning: unnecessary `unsafe` block
3) warning: field is never read: `device_id`
Currently we are yet to implement input stream support for CPAL's
webaudio host. Instead of panicking, we should return an
error, None or empty iterator in order to let the user write well
behaved cross-platform apps and notify the user accordingly rather than
crashing.
The `supported_stream_configs` method now returns the range of
configurations that are required to be supported for
`BaseAudioContext.createBuffer()` as mentioned here:
https://developer.mozilla.org/en-US/docs/Web/API/BaseAudioContext/createBuffer
That is, valid stream configurations are now considered to be any
configuration that has:
- 1 <= channel_count <= 32 and
- 8khz <= sample_rate <= 96khz
- sample_format == f32
Closes#410.
Closes#411.
This rebases #372, addressing the recent changes introduced by #397, #395, and #371 in the process.
TODO:
- [ ] Complete implementation of `callback` and `playback` timestamps in the output stream callback.
I think this is mostly complete, however there are two major TODOs:
- There seems to be a bug where timestamps are not reported when the
target device is `pulse`. This causes a `panic!` as the trigger
timestamp should always precede the timestamp when retrieved while the
stream is running.
- We need to specify the timestamp type as MONOTONIC_RAW, however the
`alsa` and `alsa-sys` crates do not yet expose the necessary
`set_tstamp_type` function.
Using an unsigned representation was problematic as some platforms
provide no exact "capture" instant. As a result, we need to calculate
the instant of capture by subtracting the buffer duration from the
callback duration. In some cases this would occasionally cause the
unsigned representation to underflow unnecessarily.