[Bug fortran/126485] New: Regression caused by r16-8455 / `64e03b96dfe`: Fortran: fix resolution of generic interface with TYPE(C_PTR) [PR66973]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=126485

            Bug ID: 126485
           Summary: Regression caused by r16-8455 / `64e03b96dfe`:
                    Fortran: fix resolution of generic interface with
                    TYPE(C_PTR) [PR66973]
           Product: gcc
           Version: 16.1.0
            Status: UNCONFIRMED
          Severity: normal
          Priority: P3
         Component: fortran
          Assignee: unassigned at gcc dot gnu.org
          Reporter: henri at henrimenke dot de
  Target Milestone: ---

Created attachment 65169
  --> https://gcc.gnu.org/bugzilla/attachment.cgi?id=65169&action=edit
Minimal (probably insufficient) patch

The reproducer needs two files:


! mod_a.f90
module mod_a
  use iso_c_binding
  implicit none
end module mod_a

! bug.f90
subroutine s1(x)
  use mod_a          ! C_LOC reached indirectly, via a .mod
  real, target :: x
  call foo(c_loc(x)) ! first call: interface is synthesised here
end subroutine s1

subroutine s2(y)
  use mod_a
  real, target :: y
  call foo(c_loc(y)) ! second call: checked against it -> bogus error
end subroutine s2


Compiling the two gives a bogus error about a dummy argument '_formal_0' that
doesn't exist. The same code compiled correctly with GCC 15.1. Since PR66973
was backported to GCC 15, I think GCC 15.2 is affected as well.


$ gfortran --version
GNU Fortran (GCC) 16.1.0
[...]
$ gfortran -c mod_a.f90 
$ gfortran -c bug.f90 
bug.f90:10:11:

   10 |   call foo(c_loc(y)) ! second call: checked against it -> bogus error
      |           1
Error: ISO_C_BINDING function actual argument at (1) requires dummy argument
‘_formal_0’ to have a matching type from ISO_C_BINDING