[tz] problems with unsigned 32-bit time_t coming up

[Alan Barrett]

On Mon, 27 May 2013, Christos Zoulas wrote:
>There have been two major bugs I can remember relating to NetBSD's
>64 bit time_t.
[...]
>2. Postgres assumed the sizeof(time_t) == sizeof(long), and when systems
>   upgraded to NetBSD-6 and recompiled postgres (because the compatibility
>   code just works fine), it exposed that postgres bug and none of the
>   databases would load.

Let me expand a little about the compatibility code.

Up to and including NetBSD-5.x, NetBSD on 32-bit platforms used 
32-bit time_t (the same size as long).  NetBSD-6.0 was the first 
release in which NetBSD on 32-bit platforms used 64-bit time_t 
(but 32-bit long).  The kernel and libc provide compatibility 
interfaces that allow applications that were compiled against an 
older version of NetBSD to continue to work (with 32-bit time_t in 
the application) when run under a newer kernel (with 64-bit time_t 
in the kernel).

Because of the compatibility code, latent bugs in applications 
that were compiled on an older version of NetBSD (that had 32-bit 
time_t) may be hidden until the application is recompiled under 
the newer version of NetBSD (with 64-bit time_t).

--apb (Alan Barrett)