Re: [HACKERS] 6.5 Release date

>
> > I have now a regression test for the NUMERIC data
> > type. Should I add it or is it too late?
>
> Add it. I'll help test. We'll ask Mark about it later ;)

    O.K. - it's in place.

    I've  added  2  tests  for  NUMERIC. One that is performed by
    default on target 'runtest'. This one uses 100  digits  after
    the  decimal point.  It causes the entire regression suite to
    take now 2.5 times longer than before.

    The other one is mainly the same test, just with other values
    and  this  time 1000 digits after the decimal point. This one
    is added to the end of  the  regression  suite  if  the  make
    target 'bigtest' is used.

    Both  stress  the  CPU  havily  by  calculating square roots,
    logarithms and power to the wanted precision.  I  don't  have
    any clue how many internal add, subtract, multiply and divide
    cycles they cause - all the complex functions  are  based  on
    Taylor/McLaurin. But it must be some hundreds.

    All expected results have been precalculated with bc(1) using
    a slightly higher precision. So you shouldn't have  to  check
    ALL  the  numerical  results  in the 250K numeric_big test by
    hand - maybe you do it anyway just to be sure that  there  is
    not one single difference of 10^^-1000 :-)


Jan

--

#======================================================================#
# It's easier to get forgiveness for being wrong than for being right. #
# Let's break this rule - forgive me.                                  #
#========================================= wieck@debis.com (Jan Wieck) #