In article <20020830174728.G2322@night-porter.duskware.de>,
Martin Husemann <martin@duskware.de> wrote:

I think I like this :-)

christos

>So inspired by David's sugestions I made two patches.
>
>The first restructures the timeout handling and uses callouts instead of
>timestamps. It resets a callout every time a byte is received and stops it
>when we reach inpustate 0 again. If the callout fires, something is wrong
>and the reset thread is woken up.
>
>This works, but the timing granularity is higher (on my i386 test notebook
>with hz=100 the threshold moves from 25ms to 30ms). I don't think this is a
>problem.
>
>After this worked fine, I did a much simpler version that just replaces
>the call to microtime with a read from mono_time and adjusts the timeout
>threshold to the new granularity.
>
>I prefer the mono_time version.
>
>Any comments? Should I commit this?
>
>Martin