Testsuite entry for GNATS PR optimization/2463 - a regression

[Toon Moene <toon at moene dot indiv dot nluug dot nl>]

This PR consists of a Fortran program used to sort a rank-2 array.

I simplified the example as follows:

      REAL DAT(2,5)
      DO I = 1, 5
         DAT(1,I) = I*1.6356-NINT(I*1.6356)
         DAT(2,I) = I
      ENDDO
      DO I = 1, 4
         DO J = I+1, 5
            IF (DAT(1,J) - DAT(1,I) .LT. 0.0) THEN
               DO K = 1, 2
                  TMP = DAT(K,I)
                  DAT(K,I) = DAT(K,J)
                  DAT(K,J) = TMP
               ENDDO
            ENDIF
         ENDDO
      ENDDO
      DO I = 1, 4
         IF (DAT(1,I) .GT. DAT(1,I+1)) CALL ABORT
      ENDDO
      END

This succeeds when compiling with 2.95.3, fails with gcc version 3.0
20010428 (prerelease) and succeeds again with gcc version 3.1 20010429
(experimental).  So it is a regression (it's wrong already for a long
time - the PR is against Red Hat's 2.96 compiler).

I plan to install this on the branch and the trunk as
g77.f-torture/execute/20010430.f - is that OK ?  Should it be installed
XFAIL (on i?86) on the branch ?

TIA,

-- 
Toon Moene - mailto:toon@moene.indiv.nluug.nl - phoneto: +31 346 214290
Saturnushof 14, 3738 XG  Maartensdijk, The Netherlands
Maintainer, GNU Fortran 77: http://gcc.gnu.org/onlinedocs/g77_news.html
Join GNU Fortran 95: http://g95.sourceforge.net/ (under construction)