mktime(3) return value

[Paul Goyette <paul%whooppee.com@localhost>]

While looking at more of our old src/regress/lib/libc regression tests, 
I noticed that mktime(3) doesn't seem to work correectly.

The following code snippet demonstrates the problem

        {
                time_t rc;
                struct tm tms;

                (void)memset(&tms, 0, sizeof(tms));
                tms.tm_year = ~0;

                errno = 0;
                rc = mktime(&tms);
        }

Given the rather absurd value of tms.tm_year, you would certainly expect 
an error.  mktime(3) does indeed set errno to EOVERFLOW, however it does 
set the return value to (time_t)-1 as is documented in the man page:

        The mktime() function returns the specified time since the Epoch
        as a time_t type value.  If the time cannot be represented, then
        mktime() returns (time_t)-1 setting the global variable errno to
        indicate the error.

(If it matters, the actual return value from mktime is 2,054,384,896 
which is 0x7a736d00)

So, my question is simple:

Is the man page wrong about returning (time_t)-1?  Or is the return 
value actually a correct representation of tm_year = ~0 ?


-------------------------------------------------------------------------
| Paul Goyette     | PGP Key fingerprint:     | E-mail addresses:       |
| Customer Service | FA29 0E3B 35AF E8AE 6651 | paul at whooppee.com    |
| Network Engineer | 0786 F758 55DE 53BA 7731 | pgoyette at juniper.net |
| Kernel Developer |                          | pgoyette at netbsd.org  |
-------------------------------------------------------------------------