Thanks for the correction; I was remembering the bad old days when min_time was a time_t.

So...in the present day, the "dayoff" code is, as you noted, a no-op on systems where longs are 32 bits.
It is still needed on systems where longs are 64 bits. And since "sizeof" can't be used in "#if" directives,
conditionalizing (to get rid of the dead code warning) isn't as easy as we might like.

        @dashdashado


On Tue, Mar 5, 2013 at 10:11 PM, Christos Zoulas <christos@zoulas.com> wrote:
On Mar 5, 10:05pm, arthurdavidolson@gmail.com (Arthur David Olson) wrote:
-- Subject: Re: [tz] dead code in zic

| While the code in question can surely be improved, it wasn't/isn't a no-op
| on systems where both time_t's and long's were/are 32-bit entities.

It might not have been when zic_t was a time_t (if it ever was), but it is
a no-op now because zic_t is int_fast64_t. If time_t is 32 or 64 is not
relevant to the particular comparison.

| The code in question dates back to when "long" was the longest integer type
| available with all compilers ("long long" was not universal then).

Yes, but things have changed and the code can be made more portable and
work better across a wider set of platforms.

christos