Select case restrictions

From: Gordon Sande (g.sande_at_worldnet.att.net)
Date: 12/17/04 

Date: Fri, 17 Dec 2004 15:10:42 GMT

In the fragment

  select case ( expression )
   case ( value )
    ....
  end select

the restrictions listed in F95 Handbook are that both
expression and value must be of the same discrete type,
that is either integer, logical or character and that in
the case of character must be of the same kind but may be
of differing length.

gfortran is diagnosing an error for differing kinds
of integers. Either this is an error, my guess, or
many other compilers have been letting this extension
through without comment.

Do I get to report one error to gfortran or an error
to many others?

In the spirit of equal time I have reported the problem
with g95's driver that does not handle "g95 a.f90 b.f90"
on MacOsX as it dumps b.f90 into ld.

Compile times for xlf, gftn and g95 seem to be 5, 5, 50
of real time with xlf using only 1 or 2 of user and system
time with much disk chatter while the other two are processor
limited. Execution times seem to be 4, 5, 15. Namelist on gftn
does not accept array elements so testbed programs using
namelist do not work. Phooey! Timings do not mean much as
I expect there is a fair amount of internal checking still
enabled.

Finding g95 and gftn is more like a urban scavenger hunt
that one would like. Fink is supposed to be the easy route
for MacOsX but it seems "not fully functional". There is
a g95 package on www.g95.org and and sourceforge has a
gftn under the guise of g95v4.0 (ask google).