Patch: Only use strict COFF for djgpp

[H . J . Lu]

On Mon, Aug 21, 2000 at 05:07:45PM -0700, Nick Clifton wrote:
>   If you could resubmit a patch with these changes I would be happy to
>   approve it.
> 

Here is the patch.

Thanks.


H.J.
---
2000-08-22  H.J. Lu  <hjl@gnu.org>

	* config.in (STRICTCOFF): New for strict COFF.

	* configure.in: Define STRICTCOFF for i386-*-msdosdjgpp*,
	i386-*-go32* and i386-go32-rtems*.
	* configure: Rebuilt.

	* config/obj-coff.c (obj_coff_endef): Follow the historical
	behavior if STRICTCOFF is not defined.

	* doc/internals.texi: Document STRICTCOFF.

Index: config.in
===================================================================
RCS file: /work/cvs/gnu/binutils/gas/config.in,v
retrieving revision 1.1.1.1
diff -u -p -r1.1.1.1 config.in
--- config.in	1999/06/03 18:01:57	1.1.1.1
+++ config.in	2000/08/19 17:28:05
@@ -166,6 +166,9 @@
 /* Using cgen code? */
 #undef USING_CGEN
 
+/* Using strict COFF? */
+#undef STRICTCOFF
+
 /* Using i386 COFF? */
 #undef I386COFF
 
Index: configure.in
===================================================================
RCS file: /work/cvs/gnu/binutils/gas/configure.in,v
retrieving revision 1.31
diff -u -p -r1.31 configure.in
--- configure.in	2000/08/16 02:37:52	1.31
+++ configure.in	2000/08/19 17:28:14
@@ -251,7 +251,9 @@ changequote([,])dnl
       i386-*-sco3.2*)       fmt=coff ;;
       i386-*-vsta)          fmt=aout ;;
       i386-*-msdosdjgpp* | i386-*-go32* | i386-go32-rtems*)
-			    fmt=coff em=go32;;
+			    fmt=coff em=go32
+			    AC_DEFINE(STRICTCOFF, 1, [Using strict COFF?])
+			    ;;
       i386-*-rtemself*)     fmt=elf ;;
       i386-*-rtems*)        fmt=coff ;;
       i386-*-gnu*)          fmt=elf ;;
Index: config/obj-coff.c
===================================================================
RCS file: /work/cvs/gnu/binutils/gas/config/obj-coff.c,v
retrieving revision 1.1.1.21
diff -u -p -r1.1.1.21 obj-coff.c
--- config/obj-coff.c	2000/08/16 00:53:47	1.1.1.21
+++ config/obj-coff.c	2000/08/22 15:11:38
@@ -732,13 +732,49 @@ obj_coff_endef (ignore)
     case C_ARG:
     case C_REGPARM:
     case C_FIELD:
+
+    /* According to the COFF documentation:
+
+       http://osr5doc.sco.com:1996/topics/COFF_SectNumFld.html
+
+       A special section number (-2) marks symbolic debugging symbols,
+       including structure/union/enumeration tag names, typedefs, and
+       the name of the file. A section number of -1 indicates that the 
+       symbol has a value but is not relocatable. Examples of
+       absolute-valued symbols include automatic and register variables, 
+       function arguments, and .eos symbols. 
+
+       But from Ian Lance Taylor:
+
+       http://sources.redhat.com/ml/binutils/2000-08/msg00202.html
+
+       the actual tools all marked them as section -1. So the GNU COFF
+       assembler follows historical COFF assemblers.
+
+       However, it causes problems for djgpp
+
+       http://sources.redhat.com/ml/binutils/2000-08/msg00210.html
+
+       By defining STRICTCOFF, a COFF port can make the assembler to
+       follow the documented behavior. */
+#ifdef STRICTCOFF
     case C_MOS:
     case C_MOE:
     case C_MOU:
     case C_EOS:
+#endif
       SF_SET_DEBUG (def_symbol_in_progress);
       S_SET_SEGMENT (def_symbol_in_progress, absolute_section);
       break;
+
+#ifndef STRICTCOFF
+    case C_MOS:
+    case C_MOE:
+    case C_MOU:
+    case C_EOS:
+      S_SET_SEGMENT (def_symbol_in_progress, absolute_section);
+      break;
+#endif
 
     case C_EXT:
     case C_WEAKEXT:
Index: doc/internals.texi
===================================================================
RCS file: /work/cvs/gnu/binutils/gas/doc/internals.texi,v
retrieving revision 1.1.1.12
diff -u -p -r1.1.1.12 internals.texi
--- doc/internals.texi	2000/07/09 23:38:11	1.1.1.12
+++ doc/internals.texi	2000/08/22 15:17:05
@@ -773,6 +773,12 @@ variable.  Normally the default value of
 configuration process will create a file named @file{targ-env.h} in the build
 directory which includes @file{te-@var{em}.h}.
 
+There is a special case for COFF. For historical reason, the GNU COFF
+assembler doesn't follow the documented behavior on certain debug symbols for
+the compatibility with other COFF assemblers. A port can define
+@code{STRICTCOFF} in the configure script to make the GNU COFF assembler
+to follow the documented behavior.
+
 Porting GAS to a new CPU requires writing the @file{tc-@var{CPU}} files.
 Porting GAS to a new object file format requires writing the
 @file{obj-@var{fmt}} files.  There is sometimes some interaction between these