Add unsafe readn and writen

[anas2660]

While actual hardware mostly only supports interleaved buffers, it can make sense to have available in some cases.
For example, using pipewire with pipewire-alsa and capturing data from external jack clients using non-interleaved buffers.

[diwic]

Thanks. So then we need readn and writen.

Can we make it safer somehow? And do we need all the mut even on the write side?

[anas2660]

Can we make it safer somehow?

We would need to know the number of channels so we can check the length of the slice.

Then you could make a fully safe version by taking a slice of slices, and individually check that each inner slice is at least frames in length. You would then have to allocate an array of pointers from that on the stack, since it does not transmute.

So it is doable. I just didn't know how to get the number of channels in a sane way in there.

And do we need all the mut even on the write side?

The alsa-sys crate takes mut pointers currently

extern "C" {
    pub fn snd_pcm_readn(
        pcm: *mut snd_pcm_t,
        bufs: *mut *mut ::std::os::raw::c_void,
        size: snd_pcm_uframes_t,
    ) -> snd_pcm_sframes_t;
}

[diwic]

It does not make sense for writei to take *const and writen to take *mut. This is probably a mistake on the alsa-lib side?

writen(&self, bufs: &[*const S], frames: usize) would make more sense, or am I missing something?

[anas2660]

yeah my mistake, it is this definition from the alsa-sys side

extern "C" {
    pub fn snd_pcm_writen(
        pcm: *mut snd_pcm_t,
        bufs: *mut *mut ::std::os::raw::c_void,
        size: snd_pcm_uframes_t,
    ) -> snd_pcm_sframes_t;
}

So yes it is a mistake from that side. They are not marked as const from the lib-alsa side either.

snd_pcm_sframes_t snd_pcm_writen(snd_pcm_t *pcm, void **bufs, snd_pcm_uframes_t size);

We could just ignore it and risk wonky codegen. I don't think the writen will ever write to the buffers so probably fine in that aspect and also the implicit __restrict__ on mut references probably doesn't matter in this case. Up to you, I just exposed what was there in alsa-sys, but I can change it if you want.

[diwic]

Yeah, change the writen function signature. Then just test both functions locally somehow so you know that they work like we expect them to. Thanks!

[anas2660]

Seems ok

[diwic]

Thanks!