Add timeradd() and timersub() to <sys/time.h>. [PSARC/2008/368 FastTrack timeout 06/16/2008]

[Don Cragun]

>Date: Mon, 09 Jun 2008 15:06:14 -0400
>From: James Carlson <james.d.carlson at sun.com>
>
>Joseph Kowalski writes:
>> 
>> Maybe this code review discussion should be taken elsewhere?
>
>Indeed.  The architectural bit was just the suggestion that if the
>implementation can be fixed so that it doesn't make lint whine, it
>probably should be.  If it can't be fixed right now, then I'm less
>certain what ought to be in the man page.  (Is that really the right
>place to document bugs in lint ... ?)

Dan has agreed to use an #ifdef __lint (to preserve standards namespace
it is __lint; not lint) to get rid of the lint issue, and has asked me
to remove the warning paragraph about lint from the man page.  I have
updated the man page in the materials directory as requested.

For the record, Dan and I discussed changing these macros to avoid the
side effects issue.  I insisted that we document that there is an
issue, but did not (and still do not) believe that the side effect
issue is a problem because all known implementations of these macros on
other OSes have the same issue.  Note that although undocumented,
timerclear(), timercmp(), and timerisset() have been present in our
<sys/time.h> header with the side effects for more than 14 years.  This
case is not changing those macros in the header at all; it is just
documenting their existence.

 - Don