MIPS/Linux: Single-stepping running away

MIPS/Linux: Single-stepping running away

[Maciej W. Rozycki]

Hello,

 I've found two bugs in gdb that lead to single-stepping over branches 
failing under specific conditions -- the code tests register numbers 
instead of their values for a few variants of branches.  Here is an 
obvious fix.  It applies against the head.

2004-10-07  Maciej W. Rozycki  <macro@mips.com>

	* mips-tdep.c (mips32_next_pc): Check the register's contents,
	not its number for BLEZ/BLEZL and BGTZ/BGTZL.

 Please apply.

  Maciej

gdb-mips32_next_pc.patch
diff -up --recursive --new-file src.macro/gdb/mips-tdep.c src/gdb/mips-tdep.c
--- src.macro/gdb/mips-tdep.c	2004-09-28 14:04:51.000000000 +0000
+++ src/gdb/mips-tdep.c	2004-10-07 16:26:14.000000000 +0000
@@ -1273,7 +1273,7 @@ mips32_next_pc (CORE_ADDR pc)
 	    pc += 8;
 	  break;
 	case 6:		/* BLEZ, BLEZL */
-	  if (read_signed_register (itype_rs (inst) <= 0))
+	  if (read_signed_register (itype_rs (inst)) <= 0)
 	    pc += mips32_relative_offset (inst) + 4;
 	  else
 	    pc += 8;
@@ -1281,7 +1281,7 @@ mips32_next_pc (CORE_ADDR pc)
 	case 7:
 	default:
 	greater_branch:	/* BGTZ, BGTZL */
-	  if (read_signed_register (itype_rs (inst) > 0))
+	  if (read_signed_register (itype_rs (inst)) > 0)
 	    pc += mips32_relative_offset (inst) + 4;
 	  else
 	    pc += 8;

Re: MIPS/Linux: Single-stepping running away

[Daniel Jacobowitz]

On Thu, Oct 07, 2004 at 06:12:10PM +0100, Maciej W. Rozycki wrote:
> Hello,
> 
>  I've found two bugs in gdb that lead to single-stepping over branches 
> failing under specific conditions -- the code tests register numbers 
> instead of their values for a few variants of branches.  Here is an 
> obvious fix.  It applies against the head.
> 
> 2004-10-07  Maciej W. Rozycki  <macro@mips.com>
> 
> 	* mips-tdep.c (mips32_next_pc): Check the register's contents,
> 	not its number for BLEZ/BLEZL and BGTZ/BGTZL.
> 
>  Please apply.

I made the foolish mistake of mixing this fix into some larger MIPS
unwinder changes many months ago, which I've never had time to
disentangle.  Sorry 'bout that.

This patch is indeed obvious; I'll check it in for you.

-- 
Daniel Jacobowitz