[Patch, Fortran] PR fortran/35820 memory leak while compiling nested forall constructs.

[Paul Richard Thomas]

Mikael,

This looks like the right thing to do, from my notes on the PR.

> The patch adds a (recursive) function to count iterators.
>
> It is tested with valgrind on both reduced and initial testcase.
> The regression test passes on all but c_by_val_1.f and value_4.f90 (both
> failing as well on trunk).

Sorry, I don't understand - do these fail without the patch too?

> I would like some advices about the testcase as I don't know how to add
> a valgrind check.

I'll give your patch a whirl tonight.  I think that we'll have to take
the view that if there are memory problems, the testcase will fail on
some platform or another!

Thanks

Paul