On Sat, 16 May 2015 14:58:14 +0200, Jan Kratochvil wrote:
> On Fri, 15 May 2015 19:04:03 +0200, Pedro Alves wrote:
> > On 05/13/2015 09:17 PM, Jan Kratochvil wrote:
> > > +  if (TYPE_CODE (gdb_type) != TYPE_CODE_PTR)
> > > +    error (_("Invalid type code %d of symbol \"%s\" "
> > > +	     "in compiled module \"%s\"."),
> > > +	   TYPE_CODE (gdb_type_from_ptr), COMPILE_I_EXPR_VAL,
> > > +	   objfile_name (objfile));
> > > +  
> > > +  switch (TYPE_CODE (gdb_type_from_ptr))
> > > +    {
> > > +    case TYPE_CODE_ARRAY:
> > > +      retval = gdb_type_from_ptr;
> > > +      gdb_type_from_ptr = TYPE_TARGET_TYPE (gdb_type_from_ptr);
> > > +      break;
> > > +    case TYPE_CODE_FUNC:
> > > +      retval = gdb_type_from_ptr;
> > 
> > AFAIC, retval is always gdb_type_from_ptr, and could be
> > moved out of the switch.
> 
> It was written this way as it has semantical logic - how the type is
> determined depends on TYPE_CODE_* of that object.  Just accidentally in this
> case it is the same.  I do not do performance micro-optimizations which
> compiler will do anyway.  But I have changed it.

I see I made a thinko in that optimization, therefore checked-in this fix-up:
	4d18dfad9edf822df205edc2c1fe3fe9f1e467b8


Jan