GCC Gfortran 15.1.0 new bug with -Wall flag

Readme.md

GCC Gfortran 15.1.0 has a new bug that is fixed in Gfortran 15.2.0 release on any operating system (Windows, macOS, ...) that is triggered by -Wall

gfortran -Wall wall.f90 && ./a.out

wall.f90:38:12:

38 | call nexact(fprime) | 1 Error: Interface mismatch in dummy procedure 'fprime' at (1): Shape mismatch in argument '_formal_0'

This occurs regardless of the implied dimension(:) or specific (even dimension(1)) dimension of the parameter.

However, doing

gfortran wall.f90 && ./a.out

works and prints the expected "ok".

Working Fortran compilers

• Flang 20.1.6: flang wall.f90 works ok--note that Flang doesn't have a -Wall like flag yet.
• Intel oneAPI 2024.0, 2025.1: ifx -warn wall.f90 works OK

Working GFortran versions with gfortran -Wall wall.f90 include: 8.5.0, 9.3.0, 10.2.0, 11.2.0, 12.2.0, 13.2.0, 14.2.0

GCC's self-test for this issue: https://gcc.gnu.org/cgit/gcc/tree/gcc/testsuite/gfortran.dg/interface_60.f90?id=e7a2b8b76ae0c8f1e49c780aa82ebb5f0325f515

wall.f90

module mismatch

implicit none

private
public :: nexact, ofun

abstract interface
  integer function ofun(parms)
    integer, dimension(1), intent(in) :: parms
  end function
end interface

contains

integer function nexact(fprime)
procedure(ofun), pointer, intent(in) :: fprime
integer, dimension(1) :: parms

parms = [1]

nexact = fprime(parms)
end function

end module


program main
use mismatch, only:  ofun, nexact
implicit none

procedure(ofun), pointer :: fprime
integer :: i

fprime=>rpoly

i = nexact(fprime)

print '(a,i0)', 'expected 1, got ', i

contains

integer function rpoly(parms)
  integer, dimension(1), intent(in) :: parms
  rpoly = parms(1)
end function

end program