https://gcc.gnu.org/bugzilla/show_bug.cgi?id=106487
Bug ID: 106487
Summary: Calls to ___builtin_nested_func_ptr_created that
cannot be resolved on M1 (Apple silicon)
Product: gcc
Version: 12.1.0
Status: UNCONFIRMED
Severity: normal
Priority: P3
Component: fortran
Assignee: unassigned at gcc dot gnu.org
Reporter: apersaud at lbl dot gov
Target Milestone: ---
I'm trying to compile a program (not public available) on an M1 and am running
in an issue where the linker is complaining about unknown functions
U ___builtin_nested_func_ptr_created
U ___builtin_nested_func_ptr_deleted
I'm using:
> gfortran-12 --version
GNU Fortran (Homebrew GCC 12.1.0_1) 12.1.0
[...]
Since I cannot share the source code. I tried to reduce it to a small program:
"""
MODULE focus
IMPLICIT NONE
PRIVATE
PUBLIC myinit
ABSTRACT INTERFACE
REAL(kind=8) FUNCTION dummyf(x)
REAL(kind=8), INTENT(IN) :: x
END FUNCTION dummyf
end INTERFACE
CONTAINS
SUBROUTINE myfun(fun, x, res)
! return res = fun(x)
IMPLICIT NONE
PROCEDURE(dummyf) :: fun
REAL(kind=8), INTENT(INOUT) :: x
INTEGER, INTENT(OUT) :: res
res = fun(x)
end SUBROUTINE myfun
SUBROUTINE myinit()
IMPLICIT NONE
INTEGER :: its, nfcalls
REAL(kind=8) :: ftv
CALL myfun(internal_func, ftv, its)
CONTAINS
FUNCTION internal_func(fv)
IMPLICIT NONE
REAL(kind=8), INTENT(IN) :: fv
REAL(kind=8) :: internal_func
!internal_func = nfcalls ! create links to
___builtin_nested_func_ptr_created
internal_func = 111 ! this works
END FUNCTION internal_func
END SUBROUTINE myinit
END MODULE focus
"""
The problem seems to be the function definition in the contains block of a
subroutine.
However, if this is not working on the M1, I would have expected some kind of
warning or error message, but the file compiles OK and only fails in a later
linker stage when the two mentioned functions cannot be found.
The command I used to compile was
"gfortran-12 -m64 -O2 -c input.f95"
When I comment out the "internal_func = nfcalls" line and uncomment the
following line, the compiler does not link to these two functions (this only
works with the -O2 flag enabled though, so it probably just inlines the
function calls and therefore there is no CONTAINS block anymore?)
I'm not that familiar with fortran/gfortran...Not sure if this can/should be
fixed, but I thought I report it.
