[PATCH 20/20] MIPS/GAS: Remove dead format specifier code

[PATCH 20/20] MIPS/GAS: Remove dead format specifier code

[Maciej W. Rozycki]

Hi,

 This change removes unreachable code -- at this point c may only be one 
of: 'b', 'd', 's', 't', 'r', 'v', 'w', 'E', 'K', 'x', 'z', 'U' or 'g' as 
selected by the switch case this piece of code is contained within.  This 
observation also invalidates the comment included -- given the limited set 
of choices as quoted above c cannot contain "other operands like 'i'."

2010-12-02  Maciej W. Rozycki  <macro@codesourcery.com>

	gas/
	* config/tc-mips.c (mips_ip): Remove dead format specifier code.

 OK to apply?

  Maciej

binutils-gas-mips-ip-reg-dead.diff
Index: binutils-fsf-trunk-quilt/gas/config/tc-mips.c
===================================================================
--- binutils-fsf-trunk-quilt.orig/gas/config/tc-mips.c	2010-12-01 21:05:59.000000000 +0000
+++ binutils-fsf-trunk-quilt/gas/config/tc-mips.c	2010-12-01 21:05:59.000000000 +0000
@@ -9506,7 +9506,6 @@ mips_ip (char *str, struct mips_cl_insn 
 		      INSERT_OPERAND (RS, *ip, regno);
 		      break;
 		    case 'd':
-		    case 'G':
 		    case 'K':
 		    case 'g':
 		      INSERT_OPERAND (RD, *ip, regno);
@@ -9535,11 +9534,6 @@ mips_ip (char *str, struct mips_cl_insn 
 			 is $0.  This only matches $0, and is checked
 			 outside the switch.  */
 		      break;
-		    case 'D':
-		      /* Itbl operand; not yet implemented. FIXME ?? */
-		      break;
-		      /* What about all other operands like 'i', which
-			 can be specified in the opcode table? */
 		    }
 		  lastregno = regno;
 		  continue;

Re: [PATCH 20/20] MIPS/GAS: Remove dead format specifier code

[Richard Sandiford]

"Maciej W. Rozycki" <macro@codesourcery.com> writes:
> 	gas/
> 	* config/tc-mips.c (mips_ip): Remove dead format specifier code.

OK, thanks.  I wonder if the 'G' confusion came from someone misinterpreting:

	      if (ok) 
		{
		  lastregno = regno;
		  continue;
		}
	      else
		break;

as fallthrough?

	      if (ok) 
		{
		  lastregno = regno;
		  continue;
		}
	      break;

would be more usual.  Maybe I'll tweak it after the microMIPS stuff
has gone in.

Richard

Re: [PATCH 20/20] MIPS/GAS: Remove dead format specifier code

[Maciej W. Rozycki]

On Tue, 7 Dec 2010, Richard Sandiford wrote:

> > 	gas/
> > 	* config/tc-mips.c (mips_ip): Remove dead format specifier code.
> 
> OK, thanks.  I wonder if the 'G' confusion came from someone misinterpreting:
> 
> 	      if (ok) 
> 		{
> 		  lastregno = regno;
> 		  continue;
> 		}
> 	      else
> 		break;
> 
> as fallthrough?

 Yes, this well could be indeed, good point.  I haven't thought about this 
arrangement being a possible cause.

 Another possibility is simply a remnant from before cooked CP0 register 
names were introduced -- I would expect 'G' to have been handled like 'd' 
back then without dedicated outermost switch statement's case code.

> 	      if (ok) 
> 		{
> 		  lastregno = regno;
> 		  continue;
> 		}
> 	      break;
> 
> would be more usual.  Maybe I'll tweak it after the microMIPS stuff
> has gone in.

 Well, I have actually removed the oddity with the microMIPS patch as the 
case had to be heavily tweaked anyway.  I also removed a couple of similar 
oddities, where a "break" or "continue" was included at the end in a block 
rather than following it within the microMIPS patch itself, but I have 
just noticed a couple escaped me.  I'll clean them up as well as they may 
make one wonder if the closing brace is the end of a (misindented) 
compound selection or iteration statement.

  Maciej