Your change of 11/19

[Richard Henderson]

On Thu, Jan 04, 2001 at 12:49:17PM -0500, Richard Kenner wrote:
> Right, but we currently get an abort since the assumption is that any
> zero-length field has a null TYPE_MAX_VALUE.

Ah, you're thinking of the MAX < MIN thing.  Which I'm not fond of,
but grudgingly admit has some special significance to Ada.

There's also the later abort for a CONSTRUCTOR with nothing in it.
The assumption being that these useless initializers are stripped
out by the front end.

So, two choices: (1) Strip out the empty initializers in the front end,
or (2) return zero from array_size_for_constructorinstead of aborting.



r~