[patch, Fortran] Fix PR 60526, variable name has already been declared as a type

[Thomas Koenig]

Am 14.02.2016 um 15:38 schrieb H.J. Lu:

> It breaks bootstrap on x86:
>
> ../../../src-trunk/libgfortran/intrinsics/selected_int_kind.f90:28:40:
>
>     integer :: _gfortran_selected_int_kind

I have fixed this in two parts:

a) reverted the patch (r233411).  I still managed to catch the revision
    immediately following r233410.  Thanks H.J. for the prompt report!

b) committed a fixed patch (r233413)

The problem was that there is no upcase equivalent for _, so the
test for a symbol with an upcase first letter found the symbol itself.
That fix was obvios, see attached patch.

I tested the new patch by regression-testing and by rebuilding
libgfortran.

I was unable to write a test case because I could not find a
set of options to allow a leading underscore in a function name.

So, should we do something differently?  There are only seven
non-generated *.f90 files in libgfortran.  The chances of
breaking bootstrap this way are relatively low, and patch
reversion is easy enough, so I don't think we should regularly
rebuild libgfortran for this.

Regards

	Thomas


2016-02-14  Thomas Koenig  <tkoenig@gcc.gnu.org>

         PR fortran/60526
         * decl.c (build_sym):  If the name has already been defined as a
         type, it has a symtree with an upper case letter at the beginning.
         If such a symtree exists, issue an error and exit.  Don't do
         this if there is no corresponding upper case letter.


2016-02-14  Thomas Koenig  <tkoenig@gcc.gnu.org>

         PR fortran/60526
         * gfortran.dg/type_decl_4.f90:  Reinstated.

-------------- next part --------------
A non-text attachment was scrubbed...
Name: p5.diff
Type: text/x-patch
Size: 1246 bytes
Desc: not available
URL: <http://gcc.gnu.org/pipermail/fortran/attachments/20160214/066efcdc/attachment.bin>