[tz] localtime.c issue

[Andy Heninger]

Whatever the code ends up being, it would be helpful if it produced no
compiler warnings with the warning level cranked up to the maximum, on
either gcc or clang.

  -- Andy



On Mon, Jul 8, 2013 at 2:49 PM, Alan Barrett <apb at cequrux.com> wrote:

> On Mon, 08 Jul 2013, Paul_Koning at Dell.com wrote:
>
>> Perhaps this will work without either a compiler warning or potential
>>> loss of precision:
>>>
>>>           seconds = tdays * SECSPERDAY + (time_t)0.5;
>>>
>>> --apb (Alan Barrett)
>>>
>>
>> I rather doubt it.  Casting 0.5 to an integer type (as time_t is) ends up
>> adding 0 or 1, but not 0.5.
>>
>
> The whole point of adding 0.5 before converting from time_t to
> int_fast32_t is to round up if time_t happens to be a floating point type,
> and to do nothing if time_t happens to be an integral type.  You can't
> assume that time_t is an integral type.
>
>
>  How about (tdays * SECSPERDAY * 2 + 1) / 2;   ?
>>
>
> Multiplying by 2 might overflow the available range.  (I have not checked
> whether there are reasons why it would always be safe.)
> --apb (Alan Barrett)
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://mm.icann.org/pipermail/tz/attachments/20130708/b1daad59/attachment.html>