[mips patch RFA] get disasm ABI from section, not symbols.

[cgd at broadcom dot com]

The patch below causes the disassembler to get the ABI from the
disassembly data's section's owner, rather from the symbols associated
with the disassembly data.

Doing the latter, you can get into a weird situation like:

                op reg, reg
        foo:
                op reg, reg

and, if 'foo' is the first symbol in the file, the result of
disassembling those two instructions may be different (the former will
not have the ABI set properly, the latter will).


This seems to DTRT at least for the case discussed in the thread including:

        http://sources.redhat.com/ml/binutils/2003-06/msg00865.html

and it doesn't break binutils/ld/gas make/"make check" (for the big
list of architectures for which i typically run that 8-).

Thinking that use of 'section->owner' is icky, i looked around, and
tehre are other similar uses.  So I don't feel so bad.  8-)



chris
--
2003-06-30  Chris Demetriou  <cgd@broadcom.com>

	* mips-dis.c (set_default_mips_dis_options): Get BFD from
	the disassembler_info's section, rather than from the
	disassembler_info's symbols pointer.

Index: mips-dis.c
===================================================================
RCS file: /cvs/src/src/opcodes/mips-dis.c,v
retrieving revision 1.43
diff -u -p -r1.43 mips-dis.c
--- mips-dis.c	8 Apr 2003 07:14:47 -0000	1.43
+++ mips-dis.c	30 Jun 2003 22:03:16 -0000
@@ -495,11 +495,11 @@ set_default_mips_dis_options (info)
   mips_hwr_names = mips_hwr_names_numeric;
 
   /* If an ELF "newabi" binary, use the n32/(n)64 GPR names.  */
-  if (info->flavour == bfd_target_elf_flavour && info->symbols != NULL)
+  if (info->flavour == bfd_target_elf_flavour && info->section != NULL)
     {
       Elf_Internal_Ehdr *header;
 
-      header = elf_elfheader (bfd_asymbol_bfd (*(info->symbols)));
+      header = elf_elfheader (info->section->owner);
       if (is_newabi (header))
 	mips_gpr_names = mips_gpr_names_newabi;
     }