[Patch, Fortran] PR fortran/36746: Fix IMPLICIT typing of derived types

[Daniel Kraft]

Hi,

so here's my second attempt at this bug, this time with a more 
conservative patch as promised :)

All I did is moving the logic that gives a symbol it's IMPLICIT derived 
type if a component reference is found from gfc_match_rvalue to 
gfc_match_varspec, so that an array-reference inbetween does not confuse 
the parser.  In addition I added a new test to check for the behaviour 
we agreed upon in the discussion of my last attempt, it succeeds by default.

I've added the reduced test from the PR to the test-suite and also 
checked that the original test (long) works with my patch.  I'm at the 
moment regression-testing on GNU/Linux-x86-32.  Is this patch ok if 
there are no regressions?

Cheers,
Daniel

-- 
Done:     Arc-Bar-Cav-Sam-Val-Wiz, Dwa-Elf-Gno-Hum-Orc, Law-Neu-Cha, Fem-Mal
To go:    Hea-Kni-Mon-Pri-Ran-Rog-Tou
-------------- next part --------------
An embedded and charset-unspecified text was scrubbed...
Name: patch.changelog
URL: <http://gcc.gnu.org/pipermail/fortran/attachments/20080905/5a0b40b4/attachment.ksh>
-------------- next part --------------
An embedded and charset-unspecified text was scrubbed...
Name: patch.diff
URL: <http://gcc.gnu.org/pipermail/fortran/attachments/20080905/5a0b40b4/attachment-0001.ksh>