Re: [PATCH v8 0/12] MR1948: Draft: windows.media.speech: Recognition session work the second.

[Bernhard Kölbl (＠besentv)]

On Tue Jan 24 12:26:24 2023 +0000, Jinoh Kang wrote:
> - Syntax: Redundant parentheses.
> - CSE: the common subexpression (`frames_available *
>
> impl->capture_wfx->nBlockAlign`) can be moved into its own variable.
> 3. Since we're fetching audio packets in a loop, there is no guarantee
> that the buffer won't overflow.
> How about:
>                 SIZE_T packet_size = frames_available * impl->capture_wfx->nBlockAlign;
>                 if (tmp_buf_offset + packet_size > tmp_buf_size)
>                 {
>                     /* Defer processing until the next iteration of the
> worker loop. */
>
>
>
>                  IAudioCaptureClient_ReleaseBuffer(impl->capture_client, 0);
>                     SetEvent(impl->audio_buf_event);
>                     break;
>                 }
>                 memcpy(tmp_buf + tmp_buf_offset, audio_buf, packet_size);
>                 tmp_buf_offset += packet_size;
>                 IAudioCaptureClient_ReleaseBuffer(impl->capture_client, frames_available);
>
> and double (or triple) `tmp_buf_size` before `malloc()`.
> (The `SetEvent` can be replaced with a traker boolean flag that lets you
> bypass the wait entirely if you so prefer.)

Should I really double or triple the size?