[Bug fortran/40898] New: STDCALL mangling problem for strings @8 instead of @4

[Bug fortran/40898] New: STDCALL mangling problem for strings @8 instead of @4

[burnus at gcc dot gnu dot org]

There seems to be a problem with mangling, STDCALL and BIND(C), see
http://groups.google.com/group/comp.lang.fortran/browse_thread/thread/eb7ca487876d9420


-- 
           Summary: STDCALL mangling problem for strings @8 instead of @4
           Product: gcc
           Version: 4.5.0
            Status: UNCONFIRMED
          Keywords: wrong-code
          Severity: normal
          Priority: P3
         Component: fortran
        AssignedTo: unassigned at gcc dot gnu dot org
        ReportedBy: burnus at gcc dot gnu dot org


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=40898

[Bug fortran/40898] STDCALL mangling problem for strings @8 instead of @4

[burnus at gcc dot gnu dot org]

------- Comment #1 from burnus at gcc dot gnu dot org  2009-07-29 07:04 -------
Function declaration

       function LoadLibrary(lpFileName) bind(C,name='LoadLibraryA')
!gcc$ ATTRIBUTES STDCALL :: LoadLibrary
         use ISO_C_BINDING

... and use

hmodule = LoadLibrary('fun.dll'//achar(0))

... and dump

hmodule = loadlibrary (&"fun.dll"[1]{lb: 1 sz: 1});

Reported error:

  undefined reference to `LoadLibraryA@8'

For a 32bit system, one expects a single 4-bye pointer and thus @4. However, I
do not see any reason why it should be 64bit instead of 32bits. It also cannot
be a 64bit Windows as one there does not use STDCALL.


-- 

burnus at gcc dot gnu dot org changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
                 CC|                            |fxcoudert at gcc dot gnu dot
                   |                            |org


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=40898

[Bug fortran/40898] STDCALL mangling problem for strings @8 instead of @4

[burnus at gcc dot gnu dot org]

------- Comment #2 from burnus at gcc dot gnu dot org  2009-07-29 07:27 -------
>From the thread at c.l.f:

"A little testing revealed that the assembly code is correct, only the mangled 
 name is wrong when there is a character argument.  Perhaps the mangled name
 takes into account a spurious hidden LEN parameter while the code
 generator knows there is no such thing."


The issue is: While I corrected the procedure call and the creation of the
procedure itself for BIND(C) [i.e. no hidden-length argument; probably also
returning a "char" instead of a "char, char_len" as return argument], I forgot
to update the declaration of external functions, namely: gfc_get_function_type

That should also help with -fwhole-file issues.


-- 

burnus at gcc dot gnu dot org changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
         AssignedTo|unassigned at gcc dot gnu   |burnus at gcc dot gnu dot
                   |dot org                     |org
             Status|UNCONFIRMED                 |ASSIGNED
     Ever Confirmed|0                           |1
   Last reconfirmed|0000-00-00 00:00:00         |2009-07-29 07:27:13
               date|                            |


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=40898

[Bug fortran/40898] STDCALL mangling problem for strings @8 instead of @4

[fxcoudert at gcc dot gnu dot org]

------- Comment #3 from fxcoudert at gcc dot gnu dot org  2009-07-29 07:32 -------
By using my cross-compiler and setting a breakpoint in
i386_pe_mangle_decl_assembler_name(), I can confirm the analysis (that it's a
question of hidden length parameter).


-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=40898

[Bug fortran/40898] STDCALL mangling problem for strings @8 instead of @4

[burnus at gcc dot gnu dot org]

------- Comment #4 from burnus at gcc dot gnu dot org  2009-07-29 10:20 -------
FX: As you seemingly have a working cross compiler: Can you check the patch
below?

(One does not need to consider the return value as gfc_return_by_reference
deals with that. Only the last change is relevant, the others are just
optimizations for performance. As all the hidden arguments are of the same type
and as we do not care about the dummy name nor have an actual name, we just add
(nstr) arguments of the char-len type.)

Index: trans-types.c
===================================================================
--- trans-types.c       (revision 150203)
+++ trans-types.c       (working copy)
@@ -2249,7 +2310,7 @@ gfc_get_function_type (gfc_symbol * sym)
             Contained procedures could pass by value as these are never
             used without an explicit interface, and cannot be passed as
             actual parameters for a dummy procedure.  */
-         if (arg->ts.type == BT_CHARACTER)
+         if (arg->ts.type == BT_CHARACTER && !sym->attr.is_bind_c)
             nstr++;
          typelist = gfc_chainon_list (typelist, type);
        }


-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=40898

[Bug fortran/40898] STDCALL mangling problem for strings @8 instead of @4

[fxcoudert at gcc dot gnu dot org]

------- Comment #5 from fxcoudert at gcc dot gnu dot org  2009-07-29 10:28 -------
The patch seems to work:

$ cat a.f90 
subroutine foo
   use ISO_C_BINDING 
   implicit none 
   interface 
      function LoadLibrary(lpFileName) bind(C,name='LoadLibraryA') 
         use ISO_C_BINDING 
         implicit none 
         !gcc$ ATTRIBUTES STDCALL :: LoadLibrary 
         integer(C_INTPTR_T) LoadLibrary 
         character(kind=C_CHAR) lpFileName(*) 
      end function LoadLibrary 
   end interface

   integer(c_intptr_t) :: i
   i = LoadLibrary ("file" // achar(0))
end subroutine
$ i586-pc-mingw32-gfortran -c a.f90       
$ i586-pc-mingw32-nm a.o|grep LoadLibraryA
         U _LoadLibraryA@4


-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=40898

[Bug fortran/40898] STDCALL mangling problem for strings @8 instead of @4

[burnus at gcc dot gnu dot org]

------- Comment #6 from burnus at gcc dot gnu dot org  2009-07-29 14:45 -------
Subject: Bug 40898

Author: burnus
Date: Wed Jul 29 14:44:51 2009
New Revision: 150216

URL: http://gcc.gnu.org/viewcvs?root=gcc&view=rev&rev=150216
Log:
2009-07-29  Tobias Burnus  <burnus@net-b.de>

        PR fortran/40898
        * trans-types.c (gfc_get_function_type): Do not add hidden
        string-length argument for BIND(C) procedures.
        * trans-decl.c (create_function_arglist): Skip over nonexisting
        string-length arguments for BIND(C) procedures.


Modified:
    trunk/gcc/fortran/ChangeLog
    trunk/gcc/fortran/trans-decl.c
    trunk/gcc/fortran/trans-types.c


-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=40898

[Bug fortran/40898] STDCALL mangling problem for strings @8 instead of @4

[burnus at gcc dot gnu dot org]

------- Comment #7 from burnus at gcc dot gnu dot org  2009-07-29 14:47 -------
FIXED on the trunk (4.5) - does not affect any branch.

(In reply to comment #5)
> The patch seems to work:

It caused a regression when I tried to regtest it as I missed another
(follow-up) place, cf. http://gcc.gnu.org/ml/fortran/2009-07/msg00256.html


-- 

burnus at gcc dot gnu dot org changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
             Status|ASSIGNED                    |RESOLVED
         Resolution|                            |FIXED


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=40898