[PATCH] Multi-arch i386 instruction printing

[PATCH] Multi-arch i386 instruction printing

[Mark Kettenis]

This patch multi-arch's i386 instruction printing.  In doing so I took
a somewhat different approach than what's done for other targets, by
always overriding the disassemble_info fields that the i386
instruction printer cares about.

Seems to work fine, and should also be able to handle x86_64.

Checked in.

Index: ChangeLog
from  Mark Kettenis  <kettenis@gnu.org>

	* i386-tdep.c (gdb_print_insn_i386): Removed.
	(i386_print_insn): New function.
	(i386_gdbarch_init): Set print_insn to i386_print_insns.
	(_initialize_i386_tdep): Don't initialize tm_print_insn and
	tm_print_insn_info.

Index: i386-tdep.c
===================================================================
RCS file: /cvs/src/src/gdb/i386-tdep.c,v
retrieving revision 1.86
diff -u -p -r1.86 i386-tdep.c
--- i386-tdep.c 6 Sep 2002 21:28:18 -0000 1.86
+++ i386-tdep.c 15 Sep 2002 13:47:53 -0000
@@ -1288,15 +1288,17 @@ i386_pc_in_sigtramp (CORE_ADDR pc, char 
    deals with switching between those.  */
 
 static int
-gdb_print_insn_i386 (bfd_vma memaddr, disassemble_info *info)
+i386_print_insn (bfd_vma pc, disassemble_info *info)
 {
-  if (disassembly_flavor == att_flavor)
-    return print_insn_i386_att (memaddr, info);
-  else if (disassembly_flavor == intel_flavor)
-    return print_insn_i386_intel (memaddr, info);
-  /* Never reached -- disassembly_flavour is always either att_flavor
-     or intel_flavor.  */
-  internal_error (__FILE__, __LINE__, "failed internal consistency check");
+  gdb_assert (disassembly_flavor == att_flavor
+	      || disassembly_flavor == intel_flavor);
+
+  /* FIXME: kettenis/20020915: Until disassembler_options is properly
+     constified, cast to prevent a compiler warning.  */
+  info->disassembler_options = (char *) disassembly_flavor;
+  info->mach = gdbarch_bfd_arch_info (current_gdbarch)->mach;
+
+  return print_insn_i386 (pc, info);
 }
 \f
 
@@ -1569,6 +1571,8 @@ i386_gdbarch_init (struct gdbarch_info i
   set_gdbarch_pseudo_register_read (gdbarch, i386_pseudo_register_read);
   set_gdbarch_pseudo_register_write (gdbarch, i386_pseudo_register_write);
 
+  set_gdbarch_print_insn (gdbarch, i386_print_insn);
+
   /* Hook in ABI-specific overrides, if they have been registered.  */
   gdbarch_init_osabi (info, gdbarch, osabi);
 
@@ -1600,9 +1604,6 @@ _initialize_i386_tdep (void)
 {
   register_gdbarch_init (bfd_arch_i386, i386_gdbarch_init);
 
-  tm_print_insn = gdb_print_insn_i386;
-  tm_print_insn_info.mach = bfd_lookup_arch (bfd_arch_i386, 0)->mach;
-
   /* Add the variable that controls the disassembly flavor.  */
   {
     struct cmd_list_element *new_cmd;
@@ -1623,7 +1624,7 @@ and the default value is \"att\".",
     struct cmd_list_element *new_cmd;
 
     new_cmd = add_set_enum_cmd ("struct-convention", no_class,
-				 valid_conventions,
+				valid_conventions,
 				&struct_convention, "\
 Set the convention for returning small structs, valid values \
 are \"default\", \"pcc\" and \"reg\", and the default value is \"default\".",

Re: [PATCH] Multi-arch i386 instruction printing

[Michal Ludvig]

[-- Attachment #1: Type: text/plain, Size: 537 bytes --]

Mark Kettenis wrote:
> This patch multi-arch's i386 instruction printing.  In doing so I took
> a somewhat different approach than what's done for other targets, by
> always overriding the disassemble_info fields that the i386
> instruction printer cares about.
> 
> Seems to work fine, and should also be able to handle x86_64.

Here's the same for x86-64 - just for easier merge when you'll do the 
multiarch gdb ;-)

Can I commit it?

Michal Ludvig
-- 
* SuSE CR, s.r.o     * mludvig@suse.cz
* +420 2 9654 5373   * http://www.suse.cz

[-- Attachment #2: x86_64_print_insn.diff --]
[-- Type: text/plain, Size: 2111 bytes --]

2002-09-16  Michal Ludvig  <mludvig@suse.cz>

	* x86-64-tdep.c (gdb_print_insn_x86_64): Removed.
	(86_64_print_insn): Added.
	(x86_64_gdbarch_init): Set print_insn to i386_print_insns.
	(_initialize_x86_64_tdep): Don't initialize tm_print_insn and
	tm_print_insn_info.

Index: x86-64-tdep.c
===================================================================
RCS file: /cvs/src/src/gdb/x86-64-tdep.c,v
retrieving revision 1.27
diff -u -p -r1.27 x86-64-tdep.c
--- x86-64-tdep.c	12 Sep 2002 08:39:26 -0000	1.27
+++ x86-64-tdep.c	16 Sep 2002 15:19:22 -0000
@@ -820,19 +820,14 @@ x86_64_register_number (const char *name
 \f
 
 
-/* We have two flavours of disassembly.  The machinery on this page
-   deals with switching between those.  */
-
+/* This is the same as i386_print_insn().  */
 static int
-gdb_print_insn_x86_64 (bfd_vma memaddr, disassemble_info * info)
+x86_64_print_insn (bfd_vma pc, disassemble_info *info)
 {
-  if (disassembly_flavour == att_flavour)
-    return print_insn_i386_att (memaddr, info);
-  else if (disassembly_flavour == intel_flavour)
-    return print_insn_i386_intel (memaddr, info);
-  /* Never reached -- disassembly_flavour is always either att_flavour
-     or intel_flavour.  */
-  internal_error (__FILE__, __LINE__, "failed internal consistency check");
+  info->disassembler_options = (char *) disassembly_flavour;
+  info->mach = gdbarch_bfd_arch_info (current_gdbarch)->mach;
+
+  return print_insn_i386 (pc, info);
 }
 \f
 
@@ -1021,6 +1016,8 @@ x86_64_gdbarch_init (struct gdbarch_info
 
   set_gdbarch_read_fp (gdbarch, cfi_read_fp);
 
+  set_gdbarch_print_insn (gdbarch, x86_64_print_insn);
+  
 /* Discard from the stack the innermost frame, restoring all registers.  */
   set_gdbarch_pop_frame (gdbarch, x86_64_pop_frame);
 
@@ -1142,9 +1139,6 @@ _initialize_x86_64_tdep (void)
 	offset += x86_64_register_info_table[i].size;
       }
   }
-
-  tm_print_insn = gdb_print_insn_x86_64;
-  tm_print_insn_info.mach = bfd_mach_x86_64;
 
   /* Add the variable that controls the disassembly flavour.  */
   {

Re: [PATCH] Multi-arch i386 instruction printing

[Andrew Cagney]

> +  gdb_assert (disassembly_flavor == att_flavor
> +	      || disassembly_flavor == intel_flavor);
> +
> +  /* FIXME: kettenis/20020915: Until disassembler_options is properly
> +     constified, cast to prevent a compiler warning.  */
> +  info->disassembler_options = (char *) disassembly_flavor;
> +  info->mach = gdbarch_bfd_arch_info (current_gdbarch)->mach;


Just create a ``change-request'' bug report.  Something else I get to 
persue with the binutils folks (unless you want the task :-).

Andrew

Re: [PATCH] Multi-arch i386 instruction printing

[Michal Ludvig]

Mark Kettenis wrote:
> This patch multi-arch's i386 instruction printing.  In doing so I took
> a somewhat different approach than what's done for other targets, by
> always overriding the disassemble_info fields that the i386
> instruction printer cares about.
> 
> Seems to work fine, and should also be able to handle x86_64.

I propose once again an introdction of i386-common.[ch] (as I already 
did in eflags register printing thread).

IMHO i386_print_insn() is a perfect candidate to go there, while it's 
the same on all i386-platforms as well as on x86-64 arch. I believe that 
more functions finally could go there. What do you think about it?

Michal Ludvig
-- 
* SuSE CR, s.r.o     * mludvig@suse.cz
* +420 2 9654 5373   * http://www.suse.cz