Adapted from a test case from c.l.f, from James van Buskirk : $ cat foo.f90 program main integer, parameter :: ipi = 1078530011 real pi parameter(pi=transfer(ipi,pi)) real, parameter :: z = transfer(ipi,z) real, parameter :: y = real(1,kind(y)) write(*,*) pi, z, y end $ gfortran -std=f95 -pedantic foo.f90 $ ./a.out 3.1415927 3.1415927 1.00000000 JvB's explanation: # the second # and third treat transfer as a specification inquiry even thought # that is not permitted even in f03, and the fourth has that pesky # REAL intrinsic, which is not an elemental function of type integer # or character. gfortran 4.2 still generates some error messages: $ gfortran-4.2 foo.f90 foo.f90:4.29: parameter(pi=transfer(ipi,pi)) 1 Error: Parameter 'pi' at (1) has not been declared or is a variable, which does not reduce to a constant expression foo.f90:5.35: real, parameter :: z = transfer(ipi,z) 1 Error: TRANSFER intrinsic not implemented for initialization at (1)
Confirm. Though other compilers have better error messages than 4.2 had. (These are all shown by default, i.e. none of the compilers below accepts it.) NAG f95: Error: PARAMETER PI used before its definition is complete g95: Error: PARAMETER value is used before being defined at (1) ifort: Error: A PARAMETER must not be defined in terms of itself. [PI] (sunf95 and openf95 have also not a good message.) I think for -std=f95 one needs another error. NAG f95 prints: Error: The REAL function is not permitted in an initialisation expression I am also not sure whether I would call the following a regression: > Error: TRANSFER intrinsic not implemented for initialization at (1)
Created attachment 14779 [details] Patch - only for the REAL()/CMPLX() diagnostics
Subject: Bug 34495 Author: burnus Date: Sun Dec 16 22:12:55 2007 New Revision: 130994 URL: http://gcc.gnu.org/viewcvs?root=gcc&view=rev&rev=130994 Log: 2007-12-16 Tobias Burnus <burnus@net-b.de> PR fortran/34495 * intrinsic.c (add_functions): Mark float and sngl as STD_GNU. (gfc_intrinsic_func_interface): Reject REAL, DBLE and CMPLX in initialization expressions for -std=f95. 2007-12-16 Tobias Burnus <burnus@net-b.de> PR fortran/34495 * gfortran.dg/initialization_16.f90: New. Added: trunk/gcc/testsuite/gfortran.dg/initialization_16.f90 Modified: trunk/gcc/fortran/ChangeLog trunk/gcc/fortran/intrinsic.c trunk/gcc/testsuite/ChangeLog
Comment on attachment 14779 [details] Patch - only for the REAL()/CMPLX() diagnostics Still to do: Adding/Re-adding an error for: Error: PARAMETER PI used before its definition is complete
http://groups.google.com/group/comp.lang.fortran/browse_thread/thread/2ad94287d7f8a6f0/ Related for non parameters: real :: pi = transfer(123,pi)
Example why it should be forbidden also for -std=gnu/legacy: implicit none dimension :: i1(5) integer :: i1 = transfer([1,2,3,4,5], i1) integer :: i2 = transfer([1,2,3,4,5], i2) dimension :: i2(5) print *, i1 ! prints: 1 2 3 4 5 print *, i2 ! prints: 1 1 1 1 1 end
In words of the standard (esp. last sentence): "If an initialization expression includes a specification inquiry that depends on a type parameter or an array bound of an entity specified in the same specification-part, the type parameter or array bound shall be specified in a prior specification of the specification-part. The prior specification may be to the left of the specification inquiry in the same statement, but shall not be within the same entity-decl."
Mine. Got a patch, but one more error message to fix ...
Subject: Bug 34495 Author: dfranke Date: Tue Dec 18 23:39:56 2007 New Revision: 131047 URL: http://gcc.gnu.org/viewcvs?root=gcc&view=rev&rev=131047 Log: gcc/fortran: 2007-12-19 Daniel Franke <franke.daniel@gmail.com> PR fortran/34495 * expr.c (check_init_expr): Check whether variables with flavor FL_PARAMETER do have a value assigned. Added error messages where appropriate. * simplify.c (gfc_simplify_transfer): Added check if the MOLD argument is a constant if working with initialization expressions. gcc/testsuite: 2007-12-19 Daniel Franke <franke.daniel@gmail.com> PR fortran/34495 * gfortran.dg/transfer_simplify_2.f90: Fixed invalid initialization expressions. * gfortran.dg/transfer_simplify_7.f90: New test. Added: trunk/gcc/testsuite/gfortran.dg/transfer_simplify_7.f90 Modified: trunk/gcc/fortran/ChangeLog trunk/gcc/fortran/expr.c trunk/gcc/fortran/simplify.c trunk/gcc/testsuite/ChangeLog trunk/gcc/testsuite/gfortran.dg/transfer_simplify_2.f90
Fixed in trunk. Closing.