Hi Brian: That's an interesting idea which I'll have to look into. I thought the flush ioctl was being used but I don't know for sure. I'll see what I can find. The old espeak never had the problem but it used a different interface to the audio subsystem. I did a quick look yesterday but didn't immediately find anything. Espeak-ng has been a pita to speakup ever since it was swapped into espeakup. The thing is that it does have a lot better alternative language capabilities with the work that has been put into it, so I don't think going backward is a useful solution. On Wed, 2 Mar 2022, Brian Buhrow wrote:
hello. Just out of curiosity, is there a way to get the audio system to flush its buffer before it finishes speaking the current buffer? I know when I worked on getting eflite working on my systems many years ago, the speech engine was calling the drain ioctl rather than the flush ioctl when the stop speech command was issued to it. Changing the flite engine so it could distinguish between drain and flush fixed the problem and made things very responsive without having to tweak buffer sizes. -Brian
