Dear all,

here's a simple and obvious patch for a rejects-valid case when
we pass a NULL() actual to an optional dummy for variants where
there is no MOLD argument and it is also not required.

The testcase is an extended version of PR55978 comment#16
and cross-checked with Intel and NAG.

Regtested on x86_64-pc-linux-gnu.

I intend to commit soon unless there are objections.

Thanks,
Harald