> 1) Your patch is against a fairly old com.c.

   My patch is against the com.c-current that I ftp'd from ftp.netbsd.org
30 minutes before mailing to gnats.

> 2) Your modified comdiag() doesn't reset sc_errors, so you will not
> observe any further overflows.

   You are right.

> 3) There are a few places that can be faster; e.g. by insuring that a
> value is cached in a register.  Also, comeint() can simply go away.

   I'm sure it can be much faster overall; comeint() should be inlined by
gcc anyhow, given the default optimisation level.  I didn't modify it for
speed, but to get it to work.  Indeed, my patch makes things slower.

> 4) Your change to comparam() is slightly wrong; the second arg to
> l_modem() in that case should always be 0.  Otherwise, TS_CARR_ON may
> be incorrectly set when turning off MDMBUF.

   You are right.

> 5) The way you use `comevents' seems more complicated (and thus
> slower) than necessary.

   I think you're more concerned with speed than I am.

> Enclosed is a new set of diffs against the -current com.c.  I can't
> test this very easily, so it would be nice if someone else did.

   I'm sorry, but I'm not in a position to do this at the moment.

   Regards,
   Felix Croes