Re: int vs. long in test(1)

[Alan Barrett <apb%cequrux.com@localhost>]

On Thu, 19 Jun 2008, James Chacon wrote:
>>> sh(1)'s test (-current/amd64) doesn't deal with big numbers (> MAXINT)
>> I think it would make sense to use intmax_t.  We already do that for
>> the integers in shell arithmatic with the $((...)) syntax.
>
> Isn't that wrong then for 64bit machines where int is 32 and the spec says 
> "signed long" is what should be used here?

I don't see anything in the spec about integer sizes.  I am looking at 
<http://www.opengroup.org/onlinepubs/009695399/utilities/test.html>.

intmax_t, being the largest signed integer type supported by the
compiler, is guaranteed to be at least as large as "signed long", and
may be larger.  Where do you see a spec that requires "signed long" and
forbids anything larger than that?

--apb (Alan Barrett)