[Patch, fortran] PR28873 and PR20067 - Cannot resolve subroutine calls when modules are used in different scopes

["THOMAS Paul Richard 169137" <Paul dot Richard dot THOMAS at cea dot fr>]

:ADDPATCH fortran:

This is a fix for yet another regression of which I am the author.  This time, in fixing
PR28201, I broke the legitimate use of different modules to carry different specific
procedures for a single generic interface.

The fix turned out to be easy; it turns out that the problem in PR28201 was not the use
association but, rather, that the search for specific procedures in the parent namespace
was returning a null symbol and this was producing the ICE.  Thus, the correct fix here
was to refresh the symbol before the final ditch attempt to find a matching intrinsic. At
the same time, the subsequent error message was changed to reflect what actually is
happening in resolve_generic_s (and resolve_generic_f), thereby fixing PR20067.

At the same time, I have restructured the search in resolve_generic_s and resolve_specific_s
to give them the same form as their function counterparts:

for (;;)
  {
    resolve as appropriate
    if MATCH_YES => return SUCCESS
    else if MATCH_ERROR => return FAILURE
[generic:]
    if no parent namespace => break
    look for same name symbol in parent namespace
    if symbol is NULL => break
    [if !generic => goto generic]
  }

This has no effect on the functionality but does improve maintainability.

The additional testcase is a slightly extended version of that offered by the reporter.

Regtested on Cygwin_NT/PIV:  OK for trunk and 4.1?

Paul

2006-08-29  Paul Thomas  <pault@gcc.gnu.org>

	PR fortran/28873
	REGRESSION FIX
	PR fortran/20067
	* resolve.c (resolve_generic_f): Make error message more
	comprehensible.
	(resolve_generic_s): Restructure search for specific procedures
	to be similar to resolve_generic_f and change to similar error
	message.  Ensure that symbol reference is refreshed, in case
	the search produces a NULL.
	(resolve_specific_s): Restructure search, as above and as
	resolve_specific_f. Ensure that symbol reference is refreshed,
	in case the search produces a NULL.

2006-08-29  Paul Thomas  <pault@gcc.gnu.org>

	PR fortran/20067
	* gfortran.dg/used_types_5.f90: Change error message.

	PR fortran/28873
	* gfortran.dg/used_types_6.f90: New test.

Attachment: pr28873.diff
Description: pr28873.diff

Attachment: commit0829.msg
Description: commit0829.msg