Grr. I've been following this thread with displeasure since it started.
It sounds like a nasty intermittent problem. :-( Hopefully someone can
reproduce the problem (or get hit with it often-enough to make
diagnostics possable).

The original problem I heard about was that sometimes on a serial
terminal the output just seems to stop for a while, and then all of a
sudden it will start up with a flood of characters. Nothing else on the
machine seems stalled at this point.

Sounds like either we're dropping an interrupt, or there's a bug
in our handling of block-finished events.

Has anyone found rhyme or reason to when characters start back up?
Also, was any form of flow control on at the time (I gather not, but
want to make sure).

Unfortunately we're within a month of having a new serial driver, and
I've been staring at it for the last few months instead of ser.c. I
doubt it will fix the error, but I'd rather debug it than debug the
soon-to-go-away ser.c driver.

> > In other words, is NetBSD crashing or is the serial interface getting
> > blocked somehow (or not being serviced properly or being intentionally
> > paused)?
> Oh, its definitely an issue with the serial driver.  The whole time, you can
> telnet into the machine and do anything you want as normal.  When its forzen,
> the believes there is still me logged into the serial prot (it won't let me
> kill the shell in fact) .... it is as if the characters are being suddenly
> buffered up or put somewhere.  The system at a topical level seems to be
> convinced it is still connected.

Technically it's the serial driver+tty system. Transmitted data is always
buffered; normally, though, the data flows out the buffers fairly
quickly.

take care,

Bill