[tz] dead code in zic

Paul Eggert eggert at cs.ucla.edu
Wed Mar 6 05:18:31 UTC 2013


On 03/05/2013 06:47 PM, Christos Zoulas wrote:
> As I mentioned before if you change the const
> variable declaration to a #define, you'll get the warning in all
> gcc 32 bit platforms (or if you use a gcc circa 4.1 which had code
> to do the constant folding more aggressively).

Yes, older GCCs are particularly bad about that; if you ask them
to warn, they issue all sorts of bogus warnings.  So it's better
to not ask them to warn.  Even newer GCCs mess up sometimes; and
even then it's often better to ignore them.

> if the code is a no-op on all 32 bit platforms it should probably be
> improved!

So far, all the suggestions for improvement would complicate
the code, which will make it harder for humans to understand.
Often in such cases it's better to keep the code simple, and to
ignore or disable the bogus warnings.

> What I am asking here, is if dayoff
> is mistyped (long instead of zic_t) and 'long oadd(long, long)'
> should be 'zic_t oadd(zic_t, long)'.

I'd go further.  I expect the zic.c code would benefit
by a uniform substitution of int_fast64_t for long -- but one would
have to do considerably more than just a simple string substitution.


More information about the tz mailing list