X3J11

[Arthur David Olson]

> > . . .the big problem here is that code like the Rationale's
> > 
> >	when = *localtime(now);
> >	when.tm_hour += 1;
> >	deadline = mktime(when);
> > 
> >	printf("Loop will finish: %s\n", asctime(when));
> >	while (difftime(time(0), deadline) > 0) whatever();
> > 
> > is *not* "a paradigm for continuing some loop for an hour."
> > The code may continue for two hours (or not at all)
> > if it is executed just before a daylight saving time transition.

> How?  deadline is a time_t (GMT) and so is time(0).

(I hope I don't end up with egg all over my face on this one!)

First, assume that it's 1:30:00 a.m. on a "spring forward" day when we execute
the
		when = *localtime(now);
statement.  We next do a
		when.tm_hour += 1;
which sets when.tm_hour to 2.  We then execute the
		deadline = mktime(when);
statement to answer the question "At what elapsed time in seconds since 1970
will my wall clock read 2:30 a.m.?" Since mktime knows we're springing forward,
it returns -1 as a way of saying "There is no such time today."
This means that the test in the
	while (difftime(time(0), deadline) > 0) whatever();
line will *always* succeed--so "whatever" will never be called.
(Side questions:  does this say anything about whether time_t should be signed?
Does it say anything about extra checking difftime should do?)

Now assume that it's 1:30:00 a.m. (for the first time :-) on a "fall back" day
when we execute the
		when = *localtime(now);
statement.  Again we
		when.tm_hour += 1;
		deadline = mktime(when);
to learn when our local wall clock will read 2:30 a.m.  But that won't be until
two hours from "now", so the loop will run twice as long as we might expect.

				--ado