termios and non-blocking write()

[Joel Sherrill]

Daniel Néri wrote:
> 
> [De-lurking...]
> 
> "Aaron J. Grier" <aaron at frye.com> writes:
> 
> > also, where did 100ms granularity for VTIME come from?
> 
> I'm guessing that the people who came up with the API never imagined
> it would be used in a real-time environment...
> 
> > I'm considering breaking it, and curious if there are any other
> > implementations of termios that have done this "nicely."
> 
> I had this problem on NetBSD a while ago where I needed a 30ms
> VTIME. I ended up hacking the kernel to use a spare c_cc[] entry as a
> flag, which if non-zero, provides microsecond VTIME granularity. I'm
> not sure if I'd call it "nice", but at least it doesn't break other
> parts of the userland that don't know about this "extension".

That doesn't sound like a particularly bad idea.  It needs to
be run by Eric Norum as he is the author of that code and has
been considering replacing it with actual BSD code.  

The limitation is ultimately the clock tick quantum which is the
basis for all timing.

> Best wishes,
>    --Daniel
> 
> --
> Daniel Neri
> dne at mayonnaise.net

-- 
Joel Sherrill, Ph.D.             Director of Research & Development
joel at OARcorp.com                 On-Line Applications Research
Ask me about RTEMS: a free RTOS  Huntsville AL 35805
Support Available                (256) 722-9985