Can't find gfortan on RHEL 6.4

[ewmoore]

I cannot build Scipy against a current numpy on RHEL 6.4. Running gfortran --version or gfortran -dumpversion both provide the same output:

GNU Fortran (GCC) 4.4.7 20120313 (Red Hat 4.4.7-3)
Copyright (C) 2010 Free Software Foundation, Inc.

GNU Fortran comes with NO WARRANTY, to the extent permitted by law.
You may redistribute copies of GNU Fortran
under the terms of the GNU General Public License.
For more information about these matters, see the file named COPYING

This has to do with the changes made to how gnu fortran compilers are detected in #5333. Going back to 8a2dd06 works.

This isn't the newest RHEL, but it is still in the full support phase.

[rgommers]

Hmm, the whole point of using -dumpversion was to remove the need for parsing random version strings. If there's gfortran versions floating around where -dumpversion is an alias for --version we should probably just revert gh-5333.

[ewmoore]

Any version of gfortran from earlier than this gcc commit from Aug 2010 will produce the same output for --version and -dumpversion. So yes, there absolutely are versions where this is true.

There is always the try harder option. Use the new check if -dumpversion produces only a single line of output, otherwise try the old check against the --version output. This ought to keep the cases fixed by #5333 working and allow older versions to still work.

[rgommers]

Good idea. Done in gh-5774.