[Bug ada/20515] New: "stdcall" imports are not handled correctly

[Bug ada/20515] New: "stdcall" imports are not handled correctly

[sbellon at sbellon dot de]

When using a GCC cross-compiler from GNU/Linux to Windows with Ada support, the
link names of pragma Import with Convention "stdcall" are not mangled correctly
and thus linking to the Windows API does not work.

A small example is:

bellonsn@pscube$ cat time.adb
with Ada.Calendar;
procedure Time is
   T : Ada.Calendar.Time;
begin
   T := Ada.Calendar.Clock;
end Time;
bellonsn@pscube$ i586-mingw32msvc-gnatmake -f time
i586-mingw32msvc-gcc -c time.adb
i586-mingw32msvc-gnatbind -x time.ali
i586-mingw32msvc-gnatlink time.ali
Warning: resolving _Sleep by linking to _Sleep@4
Use --enable-stdcall-fixup to disable these warnings
Use --disable-stdcall-fixup to disable these fixups
/usr/lib/gcc/i586-mingw32msvc/3.4.2/adalib/libgnat.a(s-osprim.o): In function `s
ystem__os_primitives__clock':
/tmp/bellonsn-mingw32/mingw32-3.4.2.20040916.1/build_dir/objs/gcc/ada/rts/s-ospr
im.adb:147: undefined reference to `_QueryPerformanceCounter'
/tmp/bellonsn-mingw32/mingw32-3.4.2.20040916.1/build_dir/objs/gcc/ada/rts/s-ospr
im.adb:151: undefined reference to `_GetSystemTimeAsFileTime'
... and so on ...

Danny Smith of MinGW suggested a small patch which works for me:

--- gcc-3.4.2-20040916-1/gcc/ada/decl.c 2005-03-17 07:11:51.208188883 +0100
+++ gcc-3.4.2-20040916-1/gcc/ada/decl.c.patched 2005-03-17 07:11:40.675851093 +0100
@@ -3509,7 +3509,7 @@
        if (list_length (gnu_return_list) == 1)
          gnu_return_type = TREE_TYPE (TREE_PURPOSE (gnu_return_list));

-#ifdef _WIN32
+#if TARGET_IS_PE_COFF
        if (Convention (gnat_entity) == Convention_Stdcall)
          {
            struct attrib *attr

I hope you can apply this patch.

-- 
           Summary: "stdcall" imports are not handled correctly
           Product: gcc
           Version: 3.4.2
            Status: UNCONFIRMED
          Severity: normal
          Priority: P2
         Component: ada
        AssignedTo: unassigned at gcc dot gnu dot org
        ReportedBy: sbellon at sbellon dot de
                CC: gcc-bugs at gcc dot gnu dot org
  GCC host triplet: i686-pc-linux-gnu
GCC target triplet: i686-pc-mingw32


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=20515

[Bug ada/20515] "stdcall" imports are not handled correctly

[charlet at adacore dot com]

------- Additional Comments From charlet at adacore dot com  2005-03-17 09:58 -------
Subject: Re:  New: "stdcall" imports are not handled correctly

Patch looks correct (certainly better) to me, investigating.

Arno


-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=20515

[Bug ada/20515] "stdcall" imports are not handled correctly

[dannysmith at users dot sourceforge dot net]

-- 
           What    |Removed                     |Added
----------------------------------------------------------------------------
                 CC|                            |dannysmith at users dot
                   |                            |sourceforge dot net


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=20515

[Bug ada/20515] "stdcall" imports are not handled correctly

[dannysmith at users dot sourceforge dot net]

------- Additional Comments From dannysmith at users dot sourceforge dot net  2005-03-18 04:23 -------
The IS_TARGET_PE_COFF hack works on 3.4.x bit won't on trunk.

Here is a cleaner patch  against trunk that replaces the preprocessor tests 
with runtime tests. I have tested with nativr mingw bootstrap but not with 
cross-compile build.


	* decl. c (gnat_to_gnu_entity): Remove unnecessary _WIN32
	preprocessor guard.
	(create_concat_name): Replace _WIN32 preprocessor guard
	with runtime chcek on TARGET_DLLIMPORT_DECL_ATTRIBUTES.

Index: ada/decl.c
===================================================================
RCS file: /cvs/gcc/gcc/gcc/ada/decl.c,v
retrieving revision 1.65
diff -c -3 -p -r1.65 decl.c
*** ada/decl.c	10 Feb 2005 13:53:20 -0000	1.65
--- ada/decl.c	18 Mar 2005 04:15:54 -0000
*************** gnat_to_gnu_entity (Entity_Id gnat_entit
*** 3642,3648 ****
  	if (list_length (gnu_return_list) == 1)
  	  gnu_return_type = TREE_TYPE (TREE_PURPOSE (gnu_return_list));
  
- #ifdef _WIN32
  	if (Convention (gnat_entity) == Convention_Stdcall)
  	  {
  	    struct attrib *attr
--- 3642,3647 ----
*************** gnat_to_gnu_entity (Entity_Id gnat_entit
*** 3655,3661 ****
  	    attr->error_point = gnat_entity;
  	    attr_list = attr;
  	  }
- #endif
  
  	/* Both lists ware built in reverse.  */
  	gnu_param_list = nreverse (gnu_param_list);
--- 3654,3659 ----
*************** create_concat_name (Entity_Id gnat_entit
*** 6576,6602 ****
  
    Get_External_Name_With_Suffix (gnat_entity, fp);
  
- #ifdef _WIN32
    /* A variable using the Stdcall convention (meaning we are running
       on a Windows box) live in a DLL. Here we adjust its name to use
       the jump-table, the _imp__NAME contains the address for the NAME
       variable. */
!   {
!     Entity_Kind kind = Ekind (gnat_entity);
!     const char *prefix = "_imp__";
!     int plen = strlen (prefix);
! 
!     if ((kind == E_Variable || kind == E_Constant)
! 	&& Convention (gnat_entity) == Convention_Stdcall)
!       {
! 	int k;
! 	for (k = 0; k <= Name_Len; k++)
! 	  Name_Buffer [Name_Len - k + plen] = Name_Buffer [Name_Len - k];
! 	strncpy (Name_Buffer, prefix, plen);
!       }
!   }
! #endif
! 
    return get_identifier (Name_Buffer);
  }
  
--- 6574,6596 ----
  
    Get_External_Name_With_Suffix (gnat_entity, fp);
  
    /* A variable using the Stdcall convention (meaning we are running
       on a Windows box) live in a DLL. Here we adjust its name to use
       the jump-table, the _imp__NAME contains the address for the NAME
       variable. */
!   if (TARGET_DLLIMPORT_DECL_ATTRIBUTES
!       && (Ekind (gnat_entity) == E_Variable
!           || Ekind (gnat_entity) == E_Constant)
!       && Convention (gnat_entity) == Convention_Stdcall)
!     {
!       const char *prefix = "_imp__";
!       int plen = strlen (prefix);
!       int k;
!       for (k = 0; k <= Name_Len; k++)
! 	Name_Buffer [Name_Len - k + plen] = Name_Buffer [Name_Len - k];
!       strncpy (Name_Buffer, prefix, plen);
!     }
!   
    return get_identifier (Name_Buffer);
  }
  

-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=20515

[Bug ada/20515] "stdcall" imports are not handled correctly

[p dot obry at wanadoo dot fr]

------- Additional Comments From p dot obry at wanadoo dot fr  2005-03-19 13:08 -------
(In reply to comment #2)
> The IS_TARGET_PE_COFF hack works on 3.4.x bit won't on trunk.
> 
> Here is a cleaner patch  against trunk that replaces the preprocessor tests 
> with runtime tests. I have tested with nativr mingw bootstrap but not with 
> cross-compile build.

The first part of the patch is not correct. On UNIX machines, we want the
Stdcall calling convention applied to a function or procedure be equivalent
to a C calling convention. The first part of the pacth break that.

> Index: ada/decl.c
> ===================================================================
> RCS file: /cvs/gcc/gcc/gcc/ada/decl.c,v
> retrieving revision 1.65
> diff -c -3 -p -r1.65 decl.c
> *** ada/decl.c	10 Feb 2005 13:53:20 -0000	1.65
> --- ada/decl.c	18 Mar 2005 04:15:54 -0000
> *************** gnat_to_gnu_entity (Entity_Id gnat_entit
> *** 3642,3648 ****
>   	if (list_length (gnu_return_list) == 1)
>   	  gnu_return_type = TREE_TYPE (TREE_PURPOSE (gnu_return_list));
>   
> - #ifdef _WIN32
>   	if (Convention (gnat_entity) == Convention_Stdcall)
>   	  {
>   	    struct attrib *attr
> --- 3642,3647 ----

The other part looks fine to me.

Pascal.

-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=20515

[Bug ada/20515] "stdcall" imports are not handled correctly

[p dot obry at wanadoo dot fr]

------- Additional Comments From p dot obry at wanadoo dot fr  2005-03-19 14:52 -------
I had a closer look.

In GCC 3.4.x we have TARGET_IS_PE_COFF. So changing _WIN32 by
TARGET_IS_PE_COFF is ok.

In GCC 4.x we have:

#define TARGET_DLLIMPORT_DECL_ATTRIBUTES 1

So instead of using TARGET_IS_PE_COFF we can use TARGET_DLLIMPORT_DECL_ATTRIBUTES.

So I propose to keep the #ifdef as it is today and
use TARGET_IS_PE_COFF on 3.4 and TARGET_DLLIMPORT_DECL_ATTRIBUTES
on 4.0. Does this looks fine to you Danny ?

Pascal.

-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=20515

[Bug ada/20515] "stdcall" imports are not handled correctly

[dannysmith at users dot sourceforge dot net]

------- Additional Comments From dannysmith at users dot sourceforge dot net  2005-03-19 20:54 -------
In reply to comment #3

>From gnat_ugn.texi:

@findex Stdcall
@cindex Convention Stdcall
@item Stdcall
This is relevant only to NT/Win95 implementations of GNAT,
and specifies that the Stdcall calling sequence will be used, as defined
by the NT API.

In all other languages "stdcall" also applies to all i386 targets, although the 
symbol decoration part is specific to windows

Danny

-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=20515

[Bug ada/20515] "stdcall" imports are not handled correctly

[p dot obry at wanadoo dot fr]

------- Additional Comments From p dot obry at wanadoo dot fr  2005-04-06 18:14 -------
(In reply to comment #5)

Ok, the documentation is missing the part where Stdcall is equivalent to
C on UNIX. This is a feature implemented some time ago. Will fix the doc.

So, what about the proposal on #4 ?

Pascal.


-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=20515

[Bug ada/20515] "stdcall" imports are not handled correctly

[dannysmith at users dot sourceforge dot net]

------- Additional Comments From dannysmith at users dot sourceforge dot net  2005-04-07 08:29 -------
"So, what about the proposal on #4 ?"

OK with me.
Danny


-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=20515

[Bug ada/20515] "stdcall" imports are not handled correctly

[cvs-commit at gcc dot gnu dot org]

------- Additional Comments From cvs-commit at gcc dot gnu dot org  2005-06-16 08:57 -------
Subject: Bug 20515

CVSROOT:	/cvs/gcc
Module name:	gcc
Changes by:	charlet@gcc.gnu.org	2005-06-16 08:56:49

Modified files:
	gcc/ada        : gigi.h trans.c decl.c 

Log message:
	2005-06-10  Eric Botcazou  <ebotcazou@adacore.com>
	Olivier Hainque  <hainque@adacore.com>
	Richard Kenner  <kenner@vlsi1.ultra.nyu.edu>
	Pascal Obry  <obry@adacore.com>
	
	* gigi.h: (build_allocator): Add arg IGNORE_INIT_TYPE.
	
	* trans.c (call_to_gnu): Issue a warning for users of Starlet when
	making a temporary around a procedure call because of non-addressable
	actual parameter.
	(process_freeze_entity): If entity is a private type, capture size
	information that may have been computed for the full view.
	(tree_transform, case N_Allocator): If have initializing expression,
	check type for Has_Constrained_Partial_View and pass that to
	build_allocator.
	(tree_transform, case N_Return_Statement): Pass extra arg to
	build_allocator.
	
	* decl.c (annotate_value): Remove early return if -gnatR is not
	specified.
	(gnat_to_gnu_field): Don't make a packable type for a component clause
	if the position is byte aligned, the field is aliased, and the clause
	size isn't a multiple of the packable alignment. It serves no useful
	purpose packing-wise and would be rejected later on.
	(gnat_to_gnu_entity, case object): Pass extra arg to build_allocator.
	
	PR ada/20515
	(gnat_to_gnu_entity): Remove use of macro _WIN32 which is wrong in the
	context of cross compilers. We use TARGET_DLLIMPORT_DECL_ATTRIBUTES
	instead.
	(create_concat_name): Idem.

Patches:
http://gcc.gnu.org/cgi-bin/cvsweb.cgi/gcc/gcc/ada/gigi.h.diff?cvsroot=gcc&r1=1.35&r2=1.36
http://gcc.gnu.org/cgi-bin/cvsweb.cgi/gcc/gcc/ada/trans.c.diff?cvsroot=gcc&r1=1.97&r2=1.98
http://gcc.gnu.org/cgi-bin/cvsweb.cgi/gcc/gcc/ada/decl.c.diff?cvsroot=gcc&r1=1.71&r2=1.72



-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=20515

[Bug ada/20515] "stdcall" imports are not handled correctly

[charlet at gcc dot gnu dot org]

------- Additional Comments From charlet at gcc dot gnu dot org  2005-06-16 09:11 -------
Fixed on mainline.

Arno

-- 
           What    |Removed                     |Added
----------------------------------------------------------------------------
             Status|UNCONFIRMED                 |RESOLVED
         Resolution|                            |FIXED
   Target Milestone|---                         |4.1.0


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=20515

[Bug ada/20515] "stdcall" imports are not handled correctly

[dannysmith at users dot sourceforge dot net]

------- Additional Comments From dannysmith at users dot sourceforge dot net  2005-06-25 12:20 -------
The patch that was committed to fix this is wrong.

#ifdef TARGET_DLLIMPORT_DECL_ATTRIBUTES is always true. It is defined to 0 for 
non-dll targets in defaults.h.

Why not make this a runtime switch as suggested earlier, rather than a 
preprocessor switch? 

Danny


-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=20515

[Bug ada/20515] "stdcall" imports are not handled correctly

[p dot obry at wanadoo dot fr]

------- Additional Comments From p dot obry at wanadoo dot fr  2005-06-27 18:50 -------
Danny,

(In reply to comment #10)
> The patch that was committed to fix this is wrong.
> 
> #ifdef TARGET_DLLIMPORT_DECL_ATTRIBUTES is always true. It is defined to 0 for 
> non-dll targets in defaults.h.

I see! Would #if instead of #ifdef fix the problem. I'm not a cross expert, so
this is why I got this wrong.

> Why not make this a runtime switch as suggested earlier, rather than a 
> preprocessor switch? 

Maybe it is better to have this checked statically if possible.

What do you think ?

Pascal.

-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=20515