gfortran: loop bind(thread) or bind(teams) inside parallel or teams with default(none) is rejected ("'i' not specified in enclosing 'parallel'")

! # Finding (GCC 16.2): a loop construct in a parallel or teams region with default(none) is rejected for its own loop variable; the
!   reproducer is the catalog program omp52-clause-bind-thread
!
! - toolchains: gcc_g90 (GCC 16.2.0, AMD Instinct MI250X gfx90a, Tioga) and gcc_sm80 (GCC 16.2.0,
!   NVIDIA A100 sm_80, wingtip); separate memory only; -O2; OMP_TARGET_OFFLOAD=mandatory
! - build: gfortran -fopenmp -foffload=amdgcn-amdhsa -foffload-options=amdgcn-amdhsa=-march=gfx90a -O2 (gcc_g90);
!   -foffload=nvptx-none -foffload-options=nvptx-none=-march=sm_80 -foffload-options=nvptx-none=-latomic -O2 (gcc_sm80)
! - failure: rejected: not specified in enclosing
! - observed: `Error: 'i' not specified in enclosing 'parallel'` (`... enclosing 'teams'` for bind(teams) in a teams region with
!   default(none)), on both GPUs.
! - expected: compiles; `RESULT ran_on_host=0 check=pass`
! - rule: i is the loop iteration variable of the loop construct's associated loop, so it has a predetermined attribute and
!   default(none) does not apply to it: "The loop iteration variable in any associated loop of a loop construct is lastprivate" (5.2
!   §5.1.1 p.97 l.18), and in Fortran "Loop iteration variables inside parallel, teams, or task generating constructs are private in
!   the innermost such construct that encloses the loop" (5.2 §5.1.1 p.97 l.22-23). The loop construct and its bind clause: 5.2
!   §11.7 p.257, §11.7.1 p.258.
! - GCC 16.2 documentation (libgomp.texi of the GCC 16.2.0 release tarball, "OpenMP Implementation Status"): "loop construct": Y
!   (OpenMP 5.0 table). The rejection is not documented.
! - other families: NVHPC 25.11/26.9 compile and run omp52-clause-bind-thread correctly; AFAR and CCE compile it and fail at run
!   time (fortran-loop-bind-thread-in-parallel.f90). CCE compiles and runs the bind(teams) program; AFAR and NVHPC reject it.
! - gate and confirmer: 12 cells (6 gcc_g90, 6 gcc_sm80), each confirmed in the gate's four environments (a-d;
!   results/gcc/gate-tioga/ and gate-wingtip/ verdicts.jsonl); confirmer verdicts (results/gcc/confirm/batch.gcc_*.jsonl):
!   compiler_bug (R11-compile-rejection) 8, unsupported (R11a-newer-form-unclaimed) 4
! - catalog programs (6): omp50-clause-bind, omp50-clause-bind-thread, omp52-clause-bind, omp52-clause-bind-thread,
!   omp60-clause-bind, omp60-clause-bind-thread
! - legality: the catalog programs are legal by construction (offload.engine.gen, c_gen and acc_gen
!   check every program against the catalog rules and the offload scope as it is built; no
!   compiler). GCC is a compiler under test only: never a legality witness, here or for another
!   family's case (catalog/machines.yaml under_test_only).
! Generated by compliance.gen (OpenMP 5.2, free-form Fortran). Do not edit.
! id: omp52-clause-bind-thread
! primary atom: omp.clause.bind.thread
! legal: yes
! observable: yes
! ran_on_host: inside the target region of interest
! evidence: construct
! needs: none
! std: f2008
! derivation (check=pass): loop bind(thread) in a parallel region: the binding thread set is the encountering thread, and each
! logical iteration is executed once per loop region by exactly one thread of that set (5.2 §11.7 p.257 and §11.7.1 p.258), so every
! thread runs all 64 iterations and writes hit(i, tid + 1) once each (distinct elements, so the concurrent iterations do not race):
! tot(t) = 64 for every thread; the team size is implementation defined and only threads 0 .. 1023 record (hit has 1024 columns), so
! all(tot(1:min(nthr, 1024)) == n) holds for any team size, and nthr >= 1; thread 0 writes b = ref; the check accepts
program gen_prog
  use omp_lib, only: omp_get_num_threads, omp_get_thread_num, omp_is_initial_device
  implicit none
  integer, parameter :: n = 64
  integer, allocatable, target :: a(:), b(:)
  integer :: ref(n)
  integer :: i, host_flag
  logical :: ok
  integer :: hit(n, 1024), tot(1024), k
  integer :: tid, nthr
  allocate(a(n), b(n))
  do i = 1, n
    a(i) = i
    b(i) = 0
    ref(i) = 2*i + 1
  end do
  host_flag = 1
  hit = 0
  nthr = -1
  !$omp target map(to: a) map(from: b) map(tofrom: hit, nthr) map(tofrom: host_flag)
    host_flag = merge(1, 0, omp_is_initial_device())
    !$omp parallel default(none) private(tid) shared(a, b, hit, nthr)
      tid = omp_get_thread_num()
      if (tid == 0) nthr = omp_get_num_threads()
      !$omp loop bind(thread)
        do i = 1, n
          if (tid == 0) b(i) = 2*a(i) + 1
          if (tid < 1024) hit(i, tid + 1) = hit(i, tid + 1) + 1
        end do
      !$omp end loop
    !$omp end parallel
  !$omp end target
  do k = 1, 1024
    tot(k) = 0
    do i = 1, n
      tot(k) = tot(k) + hit(i, k)
    end do
  end do
  ok = all(b == ref) .and. (nthr >= 1) .and. all(tot(1:min(nthr, 1024)) == n)
  print '(a,i0,a,a)', 'RESULT ran_on_host=', host_flag, ' check=', merge('pass', 'fail', ok)
  deallocate(a, b)
end program gen_prog