https://gcc.gnu.org/bugzilla/show_bug.cgi?id=92929
--- Comment #6 from jules at gcc dot gnu.org --- Apologies for breakage. This part of the patch was originally from the og9 patch supporting Fortran polymorphic class pointers posted at https://gcc.gnu.org/ml/gcc-patches/2019-07/msg00752.html. The rationale was as follows: [...] OpenACC "enter data" and "exit data" now have GOMP_MAP_POINTER and GOMP_MAP_PSET mappings removed during gimplification. In some circumstances, passing an array to a function/subroutine and then doing an "enter data" on it could leave dangling references to the function's stack, although the actual array data is defined outside the function. In any case, the pointer/pointer-set mappings don't seem to be necessary for OpenACC "enter data". I observed the described problem with a large test program, and unfortunately did not manage to come up with a minimised test case at the time. Someone more familiar with Fortran might be able to do so more easily than me! In terms of backwards compatibility, we can't do anything about the "parasitic binding" to a function's expired stack frame as generated by an older version of the compiler in this case, I don't think. That's most likely going to manifest as an intermittent crash.
