Re: PR30308, infinite recursion in i386_intel_simplify

[Alan Modra <amodra@gmail.com>]

On Thu, Feb 15, 2024 at 10:23:16AM +0100, Jan Beulich wrote:
> On 15.02.2024 02:28, Alan Modra wrote:
> > This patch exposes the symbol "resolving" flag for use in
> > i386_intel_simplify, not only preventing infinite recursion on the
> > testcase in the PR but also more complicated cases like:
> > 
> >  .intel_syntax
> >  b = a
> >  a = b
> >  mov eax, [a]
> 
> Thanks for addressing this. I wonder though: Wouldn't such circular
> equates better be rejected when they're created?

I wouldn't be inclined to do that.  If the symbols are unused then
they don't cause any problem.

> > +/* Return whether a symbol is being resolved.  */
> > +
> > +int
> > +symbol_resolving_p (symbolS *s)
> > +{
> > +  return s->flags.resolving;
> > +}
> 
> Mind me asking that such predicate (and alike) functions, when introduced
> anew, take pointer-to-const right away?

Yes, that would have been a good idea.

-- 
Alan Modra
Australia Development Lab, IBM