[patch, fortran] Further fixes for PR 49479

[Thomas Koenig]

Hi Tobias,

> OK. Though, I find the ChangeLog entry a bit misleading: There is a
> (zero-sized) return array.

I have changed the entry to

	* generated/m4/ifunction.m4:  Always call internal_malloc_size
	even when there is a zero-sized return array

on commit.

Thanks for the review!

	Thomas