Re: Re: Call for port testing on fmgr changes -- Results!

On Sun, 25 Jun 2000, Tom Lane wrote:

> Tatsuo Ishii <t-ishii@sra.co.jp> writes:
> > Tamotsu Nakagawa has posted a fix for this to a local mail list in
> > Japan. Can someone comment on this? According to him, with the patch
> > now only the geometry test fails.
>
> >  void
> > -abstime2tm(AbsoluteTime time, int *tzp, struct tm * tm, char *tzn)
> > +abstime2tm(AbsoluteTime _time, int *tzp, struct tm * tm, char *tzn)
> >  {
> > +       time_t time = (time_t) _time;
> >  #ifdef USE_POSIX_TIME
> >         struct tm  *tx;
>
> Hmm, that makes all kinds of sense if time_t is not the same size as
> AbsoluteTime --- which wouldn't surprise me at all on a 64-bit system.
> time_t *ought* to be 64-bits on such a machine.  The casts in that
> routine,
>         tx = localtime((time_t *) &time);
> are obviously bogus if so.  Can anyone with an Alpha comment?

    On Linux/Alpha, time_t is indeed 64 bit (i.e.
printf("%d\n",sizeof(time_t)) results in an '8'). If AbsoluteTime is only
32 bit, then that would definitely cause a problem.

> What surprises me more is the implication that this is the only place
> that makes such a bogus assumption about the size of time_t.  I'd have
> guessed there are more places...

    I will apply the above patch, check for other instances of time_t
size assumptions, and see what the result is this afternoon. TTYL.

---------------------------------------------------------------------------
|   "For to me to live is Christ, and to die is gain."                    |
|                                            --- Philippians 1:21 (KJV)   |
---------------------------------------------------------------------------
|   Ryan Kirkpatrick  |  Boulder, Colorado  |  http://www.rkirkpat.net/   |
---------------------------------------------------------------------------