Thread: AW: Could turn on -O2 in AIX

> > but, there is eighter an optimizer bug or a code bug in nabstime.c
> > leading to regression failure in abstime, tinterval and horology.
> > The rest all passes.
> > Does anybody see similar behavior ?
> 
> IIRC, the same regression tests fail on Linux/Alpha with 7.0.2, even
> at -O0.  I had assumed it was a 64-bit issue (probably time_t 
> is 64-bit on that platform), but maybe not.  I haven't had time to dig into the
> Alpha situation yet; can you poke at it on your machine and 
> narrow down the problem?

The macro AbsoluteTimeIsReal does not work somehow.

Andreas

Zeugswetter Andreas SB <ZeugswetterA@wien.spardat.at> writes:
> The macro AbsoluteTimeIsReal does not work somehow.

Hm.  That expands to
(((int) time) < ((int) 0x7FFFFFFC) && \ ((int) time) > ((int) 0x80000001))

On a machine where int is 32 bits, the second constant *ought* to be
treated as -2147483647, but I wonder if your compiler is doing something
bizarre with it --- say, forcing the comparison to be done as unsigned
rather than signed.
        regards, tom lane