Relay-Version: version B 2.10 5/3/83; site utzoo.UUCP
Path: utzoo!watmath!clyde!burl!codas!peora!pesnta!wjvax!brett
From: brett@wjvax.UUCP (Brett Galloway)
Newsgroups: net.unix-wizards
Subject: 4.2 bsd questions (long)
Message-ID: <764@wjvax.wjvax.UUCP>
Date: Fri, 17-Oct-86 13:40:10 EDT
Article-I.D.: wjvax.764
Posted: Fri Oct 17 13:40:10 1986
Date-Received: Wed, 22-Oct-86 03:08:30 EDT
Reply-To: brett@wjvax.UUCP (Brett Galloway)
Organization: Watkins-Johnson Co., San Jose, Calif.
Lines: 57
Keywords: lint, bsd, curses

I have a fairly long list of questions about the 4.2bsd C development
environment distribution.  Some deal with lint; some deal with apparent
bugs; and some deal with programming/use style.  I am posting here because it
seemed the best group for the overall mix of questions -- if you disagree, type
'n' now.

First, I have some questions about the declarations of getpid() and getppid().
In the 4.2bsd manual, getpid(2) and getppid(2) are both declared long.  Yet
everywhere else a pid is used, it is declared int, and getpid(2) and
getppid(2) are declared int in lint (/usr/lib/lint/llib-lc).  Which is
correct?  Which is going to be correct in the future?

Next, about lseek(2).  In the 4.2bsd manual, lseek(2) and its second argument
(the offset) are declared int, yet lint declares them long (and for that
matter, fseek(3) declares the offset to be long).  Which should be used?

Next, about select(2).  Chris Torek noted a while back to use (fd_set *)
instead of (int *) for the fd mask args to select(2).  That's fine.  But
4.2bsd lint doesn't like that.  Is 4.3bsd lint fixed?

Next, about times and time stamps.  Gettimeofday(2) and settimeofday(2)
return a timeval structures in which the seconds field is defined long.
Likewise, time(3C) and ctime(3) use long.  However, ftime(3C) (within the
timeb structure) uses time_t for the same item (the no. of seconds since
the epoch).  Likewise, stat(2) uses time_t for the time stamps.  However, in
lint, all the ctime(3) routines and time(3C) routines use time_t.  Which
is correct, and in which places?

Next, about some obvious bugs in the lint library.  Neither sys_siglist
(see psignal(3)) nor environ(3) (see getenv(3), environ(7)) is defined in
the lint library, which causes lint to bark about items used and not
declared.  Also, the argument to setjmp(3) in the lint library is wrong.

Next, a possible bug in lint itself.  In my experience, if va_alist is used
as the first argument in a function (i.e. the entire argument list is
variable), and /*VARARGS0*/ is used to tell lint that, lint still barks about
mismatches in the first argument.

Finally, some questions about curses.  First, curses fails to define unctrl()
(actually, _unctrl) in its lint library (/usr/lib/lint/llib-lcurses).
Second, lint's resetty() and setty() macros reference putchar() without
casting _flsbuf to (void), causing lint to complain about _flsbuf
returning a value which is always ignored.

====

I would like to write lint-free code, as far as that is possible.  More
importantly, I would like to write correct, portable code.  I would like to
know the status of the bugs above, and of long vs.  time_t (for times) and
long vs. int (for pids) in 4.3bsd, and even SYSV.

Please E-mail any answers; I will summarize.  Thanks in advance.

-- 
-------------
Brett Galloway
{pesnta,twg,ios,qubix,turtlevax,tymix,vecpyr,certes,isi}!wjvax!brett