[patch] Fix PR middle-end/42224

[Richard Guenther]

On Tue, 1 Dec 2009, Richard Guenther wrote:

> On Tue, 1 Dec 2009, Richard Guenther wrote:
> 
> > On Mon, 30 Nov 2009, Ulrich Weigand wrote:
> > 
> > > Hello,
> > > 
> > > this patch fixes PR middle-end/42224.  The problem was a too-strict sanity
> > > check in int_or_pointer_precision, as discussed in the bugzilla report.
> > > 
> > > Tested on s390x-ibm-linux.
> > > OK for mainline?
> > 
> > Ok.
> 
> Btw, the assert for prec != 0 causes PR42217.  I also think that
> int_or_pointer_precision should be split because it is unnecessarily
> expensive to
> 
> tree.h
> 
> static inline
> int_or_pointer_precision (const_tree type)
> {
> #ifndef ENABLE_CHECKING
>   return TYPE_PRECISION (type);
> #else
>   int_or_pointer_precision_1 (type);
> #endif
> }
> 
> with int_or_pointer_precision_1 in tree.c with all the verification
> (but excluding the != 0 precision check).
> 
> Can you followup with these changes?

Or remove int_or_pointer_precision alltogether - it seems to be
a remanent of named-address-space development and certainly was
useful there.

Richard.