i'm just asking questions, because i don't have the experience the author
clearly has.
i'm looking at this comment
/*
* if we encounter a long run of continuous read
* errors, do something drastic so that our caller
* doesn't just spin its wheels forever.
*/
long run is defined to be 10000 for the lifetime of all serial ports. i'm
thinking
of making this 5 in a row per call, and checking for a few more cant continue
type messages. was there a particular device that might spit out hundreds of
consecutive read timeouts (that's the only error this works for) before
returning
something? could the same thing happen on write?
- erik
