[Patch, fortran] PR31609 - module that calls a contained function with an ENTRY point

[Patch, fortran] PR31609 - module that calls a contained function with an ENTRY point

[Paul Thomas]

[-- Attachment #1: Type: text/plain, Size: 992 bytes --]

:ADDPATCH fortran:

Note that this problem has already been fixed by Jerry - this is an 
extension, introduced in comment #8.

I had thought that it had been fixed with PR31474 but, no, I was wrong. 
This latter fixed cases where the function itself was declared as a 
module procedure.  This new one occurred where the function was not a 
module procedure but one of its alternate entries was.  The symbol for 
the entry was winding up in the module namespace, rather than the 
function namespace.  This patch corrects this in resolve_entries in 
similar manner to the function case of PR31474.  The testcase is based 
on that in comment #8.

Regtested on Cygwin_NT/amd64 - OK for trunk?

Paul

2007-07-31  Paul Thomas  <pault@gcc.gnu.org>

    PR fortran/31609
    * resolve.c (resolve_entries): Entries declared to be module
    procedures must point to the function namespace.

2007-07-31  Paul Thomas  <pault@gcc.gnu.org>

    PR fortran/31609
    * gfortran.dg/entry_12.f90: New test.


[-- Attachment #2: pr31609.diff --]
[-- Type: text/x-patch, Size: 1978 bytes --]

Index: /svn/trunk/gcc/fortran/resolve.c
===================================================================
*** /svn/trunk/gcc/fortran/resolve.c	(revision 127061)
--- /svn/trunk/gcc/fortran/resolve.c	(working copy)
*************** resolve_entries (gfc_namespace *ns)
*** 431,436 ****
--- 431,445 ----
        && ns->parent && ns->parent->proc_name->attr.flavor == FL_MODULE)
      el->sym->ns = ns;
  
+   /* Do the same for entries where the master is not a module
+      procedure.  These are retained in the module namespace because
+      of the module procedure declaration.  */
+   for (el = el->next; el; el = el->next)
+     if (el->sym->ns->proc_name->attr.flavor == FL_MODULE
+ 	  && el->sym->attr.mod_proc)
+       el->sym->ns = ns;
+   el = ns->entries;
+ 
    /* Add an entry statement for it.  */
    c = gfc_get_code ();
    c->op = EXEC_ENTRY;
Index: /svn/trunk/gcc/testsuite/gfortran.dg/entry_12.f90
===================================================================
*** /svn/trunk/gcc/testsuite/gfortran.dg/entry_12.f90	(revision 0)
--- /svn/trunk/gcc/testsuite/gfortran.dg/entry_12.f90	(revision 0)
***************
*** 0 ****
--- 1,30 ----
+ ! { dg-do run }
+ ! Tests the fix for pr31609, where module procedure entries found
+ ! themselves in the wrong namespace.  This test checks that all
+ ! combinations of generic and specific calls work correctly.
+ !
+ ! Contributed by Paul Thomas <pault@gcc.gnu.org> as comment #8 to the pr.
+ !
+ MODULE ksbin1_aux_mod
+   interface foo
+     module procedure j
+   end interface
+   interface bar
+     module procedure k
+   end interface
+   interface foobar
+     module procedure j, k
+   end interface
+   CONTAINS
+     FUNCTION j () 
+     j = 1
+     return
+     ENTRY k (i) 
+     k = 2
+     END FUNCTION j
+ END MODULE ksbin1_aux_mod
+ 
+     use ksbin1_aux_mod
+     if (any ((/foo (), bar (99), foobar (), foobar (99), j (), k (99)/) .ne. &
+              (/1, 2, 1, 2, 1, 2/))) Call abort ()
+ end

Re: [Patch, fortran] PR31609 - module that calls a contained function with an ENTRY point

[Tobias Burnus]

:REVIEWMAIL:

Paul Thomas schrieb:
> Regtested on Cygwin_NT/amd64 - OK for trunk?
OK. Thanks for the patch.

Tobias