Re: DecTalk External (decext) testers wanted!

[Okash Khawaja <okash.khawaja@gmail.com>]

Hi,

On Mon, May 01, 2017 at 11:31:57AM +0200, Samuel Thibault wrote:
> Hello,
> 
> John Covici, on dim. 30 avril 2017 04:18:12 -0400, wrote:
> > If there is a lot of output, it takes a very long time -- 30 seconds
> > or more to shut up when I hit the enter key on the numpad.
> 
> Okash, this may be related with the serial port layer buffering data.
Right, that would explain it but I thought send_xchar would prioritise
SYNTH_CLEAR over buffered data.
> I
> realize that we haven't made the synth_flush functions call the
> flush_buffer tty ops. We need to do so, otherwise even if we send the
> clear character, that's only after the processing of pending data.
> 
> So in the synth_flush functions, one needs to first call flush_buffer()
> and then send the clear character.
> 
> Yes, flush_buffer is yet another method to be added to the spk_io_ops
> structure.
Cool, I'll make the changes and share them here as a patch.

> In the serialio case, you can just let it do nothing. One
> could want to flush the UART 16550 buffer but that's not the point of
> your patch anyway :)
Sure, may be something to keep in mind for later.

Thanks,
Okash