[patch,fortran] Fix Bug 102430 - [13/14/15/16 Regression] ICE in fold_convert_loc, at fold-const.c
Jerry D <[email protected]>
| Newsgroups | gmane.comp.gcc.patches,gmane.comp.gcc.fortran |
|---|---|
| Message-ID | <[email protected]> |
Hello all, The attached patch resolves this by issueing a sorry error. This fix is also relatively simple. I will also commit this one later today. Regression tested on x86_64. I dont think this one is worth backporting later. Let me know if anyone thinks otherwise. Regards, Jerry --- commit 396a6f28d770b27844cb7a340249d78f3f391649 (HEAD -> master) Author: Christopher Albert <[email protected]> Date: Sun Dec 21 00:33:11 2025 +0100 fortran: Reject array/allocatable LINEAR on DO [PR102430] The middle-end does not implement array/allocatable LINEAR for OpenMP worksharing loops, which can ICE during OpenMP expansion. Diagnose this case in the Fortran front end with a sorry message instead. PR fortran/102430 gcc/fortran/ChangeLog: * openmp.cc (resolve_omp_clauses): Reject array/allocatable LINEAR on worksharing-loop constructs. gcc/testsuite/ChangeLog: * gfortran.dg/gomp/pr102430.f90: New test. Signed-off-by: Christopher Albert <[email protected]>
pr102430.diff
(text/x-patch, 2.7 KB)
From 5424fef298bff8b69a351dde8ab2a6b7c46af9ab Mon Sep 17 00:00:00 2001 From: Christopher Albert <[email protected]> Date: Sun, 21 Dec 2025 00:33:11 +0100 Subject: [PATCH] fortran: Reject array/allocatable LINEAR on DO [PR102430] The middle-end does not implement array/allocatable LINEAR for OpenMP worksharing loops, which can ICE during OpenMP expansion. Diagnose this case in the Fortran front end with a sorry message instead. PR fortran/102430 gcc/fortran/ChangeLog: * openmp.cc (resolve_omp_clauses): Reject array/allocatable LINEAR on worksharing-loop constructs. gcc/testsuite/ChangeLog: * gfortran.dg/gomp/pr102430.f90: New test. Signed-off-by: Christopher Albert <[email protected]> --- gcc/fortran/openmp.cc | 35 +++++++++++++++++++++ gcc/testsuite/gfortran.dg/gomp/pr102430.f90 | 11 +++++++ 2 files changed, 46 insertions(+) create mode 100644 gcc/testsuite/gfortran.dg/gomp/pr102430.f90 diff --git a/gcc/fortran/openmp.cc b/gcc/fortran/openmp.cc index 76c601952a7..71bbce13758 100644 --- a/gcc/fortran/openmp.cc +++ b/gcc/fortran/openmp.cc @@ -10435,6 +10435,41 @@ resolve_omp_clauses (gfc_code *code, gfc_omp_clauses *omp_clauses, } break; case OMP_LIST_LINEAR: + if (code) + { + bool is_worksharing_for = false; + switch (code->op) + { + case EXEC_OMP_DO: + case EXEC_OMP_PARALLEL_DO: + case EXEC_OMP_DISTRIBUTE_PARALLEL_DO: + case EXEC_OMP_TARGET_PARALLEL_DO: + case EXEC_OMP_TEAMS_DISTRIBUTE_PARALLEL_DO: + case EXEC_OMP_TARGET_TEAMS_DISTRIBUTE_PARALLEL_DO: + is_worksharing_for = true; + break; + default: + break; + } + + if (is_worksharing_for + && (n->sym->attr.dimension + || n->sym->attr.allocatable)) + { + if (n->sym->attr.allocatable) + gfc_error ("Sorry, ALLOCATABLE object %qs in " + "LINEAR clause on worksharing-loop " + "construct at %L is not yet supported", + n->sym->name, &n->where); + else + gfc_error ("Sorry, array %qs in LINEAR clause " + "on worksharing-loop construct at %L " + "is not yet supported", + n->sym->name, &n->where); + break; + } + } + if (code && n->u.linear.op != OMP_LINEAR_DEFAULT && n->u.linear.op != linear_op) diff --git a/gcc/testsuite/gfortran.dg/gomp/pr102430.f90 b/gcc/testsuite/gfortran.dg/gomp/pr102430.f90 new file mode 100644 index 00000000000..73c8c3f1dec --- /dev/null +++ b/gcc/testsuite/gfortran.dg/gomp/pr102430.f90 @@ -0,0 +1,11 @@ +! { dg-do compile } +! { dg-options "-fopenmp" } +! PR fortran/102430 + +program p + integer :: a(2) + !$omp parallel do linear(a) ! { dg-error "Sorry, array" } + do i = 1, 8 + a = a + 1 + end do +end program p -- 2.52.0