[Bug fortran/102430] [9/10/11/12 Regression] ICE in fold_convert_loc, at fold-const.c:2554 since r7-4447-gb4c3a85be9658537

["jakub at gcc dot gnu.org" <gcc-bugzilla@gcc.gnu.org>]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=102430

Jakub Jelinek <jakub at gcc dot gnu.org> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
                 CC|                            |burnus at gcc dot gnu.org

--- Comment #2 from Jakub Jelinek <jakub at gcc dot gnu.org> ---
The OMP_CLAUSE_LINEAR_ARRAY support is there only for simd constructs (and
those are even simpler because we mostly punt on optimizing them by forcing vf
of 1).
The above mentioned commit was the "Partial OpenMP 4.5 fortran support" patch,
which enabled linear clause on do but apparently didn't actually add support
for array or allocatable linear.  The support for linear on worksharing-loops
is done  in omp-expand.c (in 3 places expand_omp_for_generic,
expand_omp_for_static_nochunk, expand_omp_for_static_chunk) and needs to be
added to those.  Whether it can use omp_clause_linear_ctor langhook or not is
unclear (or whether it needs to be done somehow during omp-low.c instead of
omp-expand.c).
So, for release branches I think the best would be to sorry on
OMP_CLAUSE_LINEAR_ARRAY clauses on OMP_FOR (e.g. in the FE).