Definite portability bug in 4.2 getty

[kre at mulga.UUCP]

To: Steven M Kramer @ Axiom  (posted as a correction of earlier news)

I'm glad that you approve of the 4.2bsd getty (which is one
of my small contributions), however, your bug fixes really just aren't ...

Easiest is your "fix" to the delay setting tables.  There you simply
have not read the code.  If you look in adelay() in subr.c, you
will see that the delay tables are searched in ascending order
of delay values, stopping when a delay greater than, or equal to,
the requested delay is encountered.  To stop such a search, I
have used a '0' as a sentinal value, a fairly common technique
I think.  An alternative would be to use a value bigger than
anything anyone would ever request - but how big is that?
The '0' means "anything bigger than the previous value", and
should obviously result in the longest possible delay, not none!
Adelay() has a special case for a requested delay of 0.

The "portability" bug in the handling of the flag bits is
really just an implementation technique, perhaps not the
best possible, but not really a bug either.

Getty uses the new, 4.2bsd, #defines of CRTBS, CTLECH, etc,
which aren't available on previous systems (really they're
not intended for use outside the kernel on 4.2, but getty
is a pretty special program).  These names are defined
to be (LCRTBS<<16) (etc) - they must have 16 low order zero
bits, and cannot conflict with any of the sg_flags bits
used with TIOCSETP.

If you are going to port getty from 4.2 to some other
system, you should port the #defines as well, I guess
that you have assumed that CRTBS is just a new name
for LCRTBS;  such is not the case, LCRTBS is still
there, and has the same value that it always had.
It hasn't been "moved".

It is true, as you suggest, that the code would be a little
cleaner if the "standard" flags and the "local" flags were
handled by distinct routines - that is while the distinction
between them remains, which it didn't on 4.1c bsd, where the
4.2 getty was developed.  Its quite likely that sometime the
distinction will vanish again, after all, they're all just flags ...


Robert Elz
decvax!mulga!kre

ps: just in case anyone hasn't guessed, I'm suggesting that
you ignore the suggested fixes to getty, and leave it as it is.