jack@rlgvax.UUCP (Jack Waugh) (11/06/83)
According to rlgvax!jds:
The System III tty driver uses a coding convention to associate
flag bits with flag words. If shown the flag bit "OPOST", you
immediately know that it pertains to the "t_oflag" flag word of
the tty structure. This is just great except that the convention
is not used 100% of the time.
The following is the example of what can happen if you come up
with a nice little coding convention but don't stick to it:
tttimeo(tp)
register struct tty *tp;
{
tp->t_state &= ~TACT;
if (tp->t_iflag&ICANON || tp->t_cc[VTIME] == 0)
return;
if (tp->t_rawq.c_cc == 0)
return;
if (tp->t_state&RTO) {
tp->t_delct = 1;
if (tp->t_state&IASLP) {
tp->t_state &= ~IASLP;
wakeup((caddr_t)&tp->t_rawq);
}
} else {
tp->t_state |= RTO|TACT;
timeout(tttimeo, tp, tp->t_cc[VTIME]*(HZ/10));
}
}
When you see the "ICANON" flag word, you immediately assume (as
did person who wrote the above function) that it belongs to the
"t_iflag" word.
Thus, if you want to check for canonicalization of the input
stream, you simply do the following:
if (tp->t_iflag&ICANON)
/*
* We're processing input characters.
*/
Wrong! The "ICANON" flag is one of those little exceptions -- it
is a flag bit for "t_lflag". Thus, the above is equivalent to:
if (tp->t_iflag&BRKINT)
/*
* We're processing input characters.
*/
Which is not what you want to do at all!
The upshot of this is that when you do a:
stty -icanon brkint min '^f' time '^b'
It isn't going to work like the book tells you it will.
This is a bug in Bell's S3, meaning it is probably a bug in
EVERYONE's port of S3.
It has been fixed in System 5.