[cfarm-users] What are people testing 32-bit builds on?

[Jonathan Wakely]

On Fri, 15 May 2026 at 12:09, Joel Sherrill via cfarm-users
<cfarm-users at lists.tetaneutral.net> wrote:
>
> On 5/15/2026 2:49 AM, Peter Gutmann via cfarm-users wrote:
> > An additional note for people targeting embedded, a number of RTOS vendors
> > have solved the time_t problem by making it unsigned, so remember to use
> > ( time_t ) casts everywhere you check values, e.g. comparing to -1.
>
> Making it unsigned is wrong per POSIX since at least Issue 6. This page
> has a 2004 copyright so it has been that way a long time.
>
> https://pubs.opengroup.org/onlinepubs/009695399/basedefs/sys/types.h.html

Where does it say signed integer type? It says integer type of
floating-point type, but an unsigned integer type is an integer type.
Dates before the epoch are not defined in POSIX, so there's not really
any need to represent negative time_t values (you would need them if
you use time_t for durations rather than time points, but there's a
reason that difftime returns double, not time_t).

Anyway, POSIX has changed, so what it said in 2004 no longer binds
implementations.

In POSIX.1-2001 it was allowed to be a floating-point type, but since
2008 it's not.

The current POSIX standard just says it's an integer type of at least
64 bits. Before that it just said an integer type.