[Patch, fortran] PR44945 - [4.6 Regression] Wrong decl for module vars and PR45077

[Paul Richard Thomas]

The comment before trans-decl.c:gfc_get_module_backend_decl describes
the main features of this patch.  The key functional change is that a
gsymbol is created for modules that are compiled separately and
symbols added to its namespace, in order to access the same
backend_decl for all use associations.

The testcases really only check the fix for pr45077, although the
mechanism that fixes pr44945 is tested.  This is because PR44945 does
not cause a failure for all platforms, or perhaps none at all, as far
as Tobias and I can determine.  The patch ensures that only one
declaration is used in all circumstances.

Bootstraps and regtests on FC9/x86_64 - OK for trunk?

Paul

2011-02-20  Paul Thomas  <pault@gcc.gnu.org>

	PR fortran/45077
	PR fortran/44945
	* trans-types.c (gfc_get_derived_type): Remove code that looks
	for decls in gsym and add call to gfc_get_module_backend_decl.
	* trans.h : Add prototype for gfc_get_module_backend_decl.
	* trans-decl.c (gfc_get_module_backend_decl): New function.
	(gfc_get_symbol_decl): Call it.

2011-02-20  Paul Thomas  <pault@gcc.gnu.org>

	PR fortran/45077
	PR fortran/44945
	* gfortran.dg/whole_file_28.f90 : New test.
	* gfortran.dg/whole_file_29.f90 : New test.
-------------- next part --------------
A non-text attachment was scrubbed...
Name: submit.diff
Type: text/x-patch
Size: 7888 bytes
Desc: not available
URL: <http://gcc.gnu.org/pipermail/gcc-patches/attachments/20110220/331cca70/attachment.bin>