Re: COLS should be type size_t

[William McBrine]

On 7/25/06, Gurganus, Brant L <address@hidden> wrote:

> Currently COLS is an int which can default to being signed or unsigned on
> different machines.

Eh? You might be thinking of char. An int is signed by definition.
(Specifically, it must be able to represent values from -32767 to
32767.)

In any case, all plausible values for COLS are well within the
positive part of the minimum range for int.

> As far as I can tell, though I don't have significant experience with
> ncurses, there shouldn't be a reason for these to be negative and
> since they are size of something, they should use size_t which,
> on most machines, should be unsigned.

It's not the size of something _in bytes_, which is what size_t is
for, and which is the only appropriate use of size_t. Counting screen
columns in size_t would be a gross misuse of size_t, IMHO.