[PATCH] fortran: allow character in conditional expression

[Tobias Burnus]

Hi all,

(mostly my fault that the original email didn't went to the mailing lists.)

Yuao Ma wrote:
> On Thu, Oct 16, 2025 at 5:47 PM Tobias Burnus <tburnus@baylibre.com> wrote:
>>>>> +  /* TODO: support array for conditional expressions  */
>>>> Shouldn't this be, e.g.,
>>>> - TODO: support arrays in conditional expressions
>>>> - TODO: conditional expressions with arrays
>>>> - TODO: array support for/in conditional expressions
>> Thus, I think this tiny change and the gfc_is_proc_ptr_comp is all
>> what's required for this patch, i.e. it should be quick.
> Done.
* * *
>>>> * * *
>>>>
>>>> Otherwise, LGTM.
>>>> Thanks for the patch!
>> * * *
>>>> PS: I was a bit unsure whether the string-length evaluation
>>>> properly handles se.pre/se.post, but it does:
>> i.e. no action required – unless you want to add a testcase for it.
>>
> I checked this program locally today. The original tree dump behaves
> as expected. The key seems to be that the pre/post block is properly
> handled in the cond-expr condgen part; we just need to propagate the
> string_length attribute. It looks like it works as expected.
> As a side question, is ALLOCATE unnecessary in Fortran for this
> particular program?

I assume that relates to:

   integer, allocatable :: f
   f = x

If so: Fortran (re)allocates the left-hand as needed,
i.e. if 'f' is unallocated (which it is by construction)
it will be allocated - to the same array shape and
character length as the RHS. As in:
   f = "hello world"

If it is already allocated - suitable for the RHS, the
allocation remains as it was.

In this case, 'x' is unallocated – but it is also not
called.

(That relates to:)

> implicit none (type, external)
> integer, allocatable :: x
>
> print *, (allocated (x) ? "abcded"(1:f()) : "bar")
> contains
> function f()
>    integer, allocatable :: f
>    f = x
> end
> end
* * *

> The updated character patch is attached. I'll commit this over the
> weekend if there are no objections.

Sorry for the delay. Yes, LGTM.

Thanks,

Tobias