[RFA/ARM] Fix recognition of Neon instruction shapes.

[Richard Earnshaw]

On 07/12/11 13:56, Matthew Gretton-Dann wrote:
> The attached patch fixes how Neon instruction shapes are recognised.
> 
> Currently neon_select_shape will recognise the following instruction as
> NS_DF instead of rejecting it:
>     vcvt.f64.s32 d0, s0, #11
> 
> This is because the recognition checks that the first operands match the
> 'shape' for NS_DF, but doesn't check that there are no operands
> following.
> 
> Tested on arm-none-eabi.  This shows regressions which are fixed by the 
> previously posted patch at 
> http://sourceware.org/ml/binutils/2011-12/msg00079.html.
> 
> OK?
> 
> gas/ChangeLog:
> 
> 2011-12-07  Matthew Gretton-Dann  <matthew.gretton-dann@arm.com>
> 
> 	* config/tc-arm.c (ARM_IT_MAX_OPERANDS): New define.
> 	(arm_it): Use ARM_IT_MAX_OPERANDS.
> 	(neon_select_shape): Ensure we have matched all	operands.
> 
> gas/testsuite/ChangeLog:
> 
> 2011-12-07  Matthew Gretton-Dann  <matthew.gretton-dann@arm.com>
> 
> 	* gas/arm/neon-suffix-bad.l: Add testcase.
> 	* gas/arm/neon-suffix-bad.s: Likewise.
> 
> Thanks,
> 
> Matt
> 
> 

OK

R.