Re: [patch, fortran, committed] Fix off-by-one in IBCLR/IBSET arg checking.

[Brooks Moses <brooks dot moses at codesourcery dot com>]

Steve Kargl wrote:
> On Wed, Jan 10, 2007 at 05:31:00PM -0500, Jack Howarth wrote:
> > On Tue, Jan 09, 2007 at 11:46:25PM -0800, Brooks Moses wrote:
> > > ----------------------------------------------------------------
> > > 2007-01-09  Brooks Moses  <brooks.moses@codesourcery.com>
> > >
> > > 	* simplify.c (gfc_simplify_ibclr): Fix POS comparison.
> > > 	(gfc_simplify_ibset): Same.
> > >
> > > ----------------------------------------------------------------
> > >
> > > The bits in an N-bit integer are numbered from 0 to N-1; thus, this 
> > > error should be thrown when the POS argument is equal to N, as well as 
> > > when it's greater than N.
> > >
> > > Committed to trunk as obvious.
> > Shouldn't this be applied under the obvious rule to gcc 4.2
> > branch as well?
>
> No.  Brooks is fixing a oversight in his previous patch on trunk.

Only partly an oversight -- it's a pre-existing bug, though I should 
have included the fix with the other patch since it's in the same code. 
 (My patch did convert its manifestation from a "Result out of range" 
error to an ICE, but it's a bug in either case.)

Since the other patch is a bug-fix, I was intending to apply it to 4.2 
after a regression check and the customary one-week delay; I'll include 
this in with it then.

- Brooks