[PATCH] Fix PR24851 (3rd version)

[Ian Lance Taylor]

Richard Guenther <rguenther@suse.de> writes:

> I'm handling that specially because later on I need a type for the
> index, which I don't have in the both offsets NULL case.  At this
> place I know we can fold to either false or true dependent on the
> comparison code, so I just two arbitrary integer constants doing
> the comparison.  integer_zero_node came handy here, no need of
> fold_convert necessary.

I get it.

Patch is OK after removing the extraneous comment.

Thanks.

Ian