Re: RFC on reduce intrinsic function, dead code?

[Jerry D]

On 7/16/26 1:53 PM, Paul Richard Thomas wrote:
My version of the REDUCE documentation was overly close to the
standard and so violated copyright. I will have another go.
Paul


I managed to come up with a test case that exercises the error message that was 
dead code. It is a runtime check.

Jerry

On Thu, 16 Jul 2026 at 19:07, Jerry D <jvdelis...@gmail.com> wrote:

On 7/15/26 10:43 PM, Thomas Koenig wrote:
Hi Jerry,

While looking at other things I noticed what I think is dead code.

libgfortran/intrinsics/reduce.c

    bool masked = false;

which is never changed.

Farther down in the function:

        if (masked && (ext != GFC_DESCRIPTOR_EXTENT (mask, i)))
      {
        int mext = (int)GFC_DESCRIPTOR_EXTENT (mask, i);
        runtime_error ("shape mismatch between ARRAY and MASK in the REDUCE "
               "intrinsic (%" FLM_Z "d/%d)", ext, mext);
      }

I wonder if this is supposed have been mask_present?

Very probably, following the logic. A (mismatching) test
case should confirm that.

BTW, I just noticed that REDUCE seems to be missing from the
documentation.  That should be easy to add (and also to
backport to all branches where it has been implemented).

Best regards

      Thomas


I made that simple and obvious change and no regressions, so we are not
exercising that check. I will have to look at our test cases to see if I can add
a check into one of them.

Regards,

Jerry