"Duncan Kinder" <dckinder@mountain.net> writes:
> Set forth below are three regression test failures encountered on RedHat
> 6.0. All involved differences between expected and actual dates. The
> actual date is the date the test was run:
I see no date problems there. Are you mistaking the diff headers for
data differences?
Anyway, the exp/pow problems are probably the result of a known
configuration glitch: if finite() is a function rather than a macro
on your machine, then 6.5.* doesn't do overflow testing correctly for
these two functions. The polygon diffs look like just the usual
cross-machine differences in floating-point roundoff error, and the
random diff is, well, random. (Run it again and it probably won't
fail.)
In short, nothing to worry about here.
regards, tom lane