The following (valid) test case is currently rejected: implicit none type :: t1 integer :: i end type type, extends(t1) :: t2 end type type(t1),target :: x1 type(t2),target :: x2 select type ( y => fun(1) ) type is (t1) print *,"t1" type is (t2) print *,"t2" class default print *,"default" end select select type ( y => fun(-1) ) type is (t1) print *,"t1" type is (t2) print *,"t2" class default print *,"default" end select contains function fun(i) class(t1),pointer :: fun integer :: i if (i>0) then fun => x1 else if (i<0) then fun => x2 else fun => NULL() end if end function end ... with the bogus message: select type ( y => fun(1) ) 1 Error: Selector shall be polymorphic in SELECT TYPE statement at (1)
Side note: Invalid code like function fun() class(t1) :: fun end function is not rejected, although the polymorphic 'fun' is neither a pointer, allocatable nor a dummy.
Bonus feature #1: Adding this to comment #0 ... select type ( y => fun(0) ) type is (t1) print *,"t1" type is (t2) print *,"t2" class default print *,"default" end select ... should give a runtime error, since fun(0) evaluates to NULL. F08 section 16.5.1.6 demands that, if the selector is a pointer, it should be associated. In the same way, having an allocatable selector which is not allocated, should also give a runtime error.
Bonus feature #2: select type ( y => fun(1) ) type is (t1) y%i = 1 type is (t2) y%i = 2 end select ... should be rejected, due to (F08): C836 (R847) If selector is not a variable or is a variable that has a vector subscript, associate-name shall not appear in a variable definition context (16.6.7).
(In reply to comment #2) > ... should give a runtime error, since fun(0) evaluates to NULL. > > F08 section 16.5.1.6 demands that, if the selector is a pointer, it should be > associated. > In the same way, having an allocatable selector which is not allocated, should > also give a runtime error. "If the selector has the POINTER attribute, it shall be associated" "If the selector is allocatable, it shall be allocated" Note: Those are not constraints, which the compiler has to diagnose but constraints to the programmer. The compiler might diagnose those, but I think it should not do so by default. Similarly to ptr = 5 where gfortran also does not do any checking. (Some checks are done for -fcheck=pointer, though not for this one.) You can add a check using -fcheck=* but doing a general check I do not like for performance / code-size reasons.
Subject: Bug 44044 Author: janus Date: Mon May 10 12:54:25 2010 New Revision: 159217 URL: http://gcc.gnu.org/viewcvs?root=gcc&view=rev&rev=159217 Log: 2010-05-10 Janus Weil <janus@gcc.gnu.org> PR fortran/44044 * match.c (gfc_match_select_type): Move error message to resolve_select_type. * resolve.c (resolve_select_type): Error message moved here from gfc_match_select_type. Correctly set type of temporary. 2010-05-10 Janus Weil <janus@gcc.gnu.org> PR fortran/44044 * gfortran.dg/class_7.f03: Modified. * gfortran.dg/select_type_1.f03: Modified. * gfortran.dg/select_type_12.f03: New. Added: trunk/gcc/testsuite/gfortran.dg/select_type_12.f03 Modified: trunk/gcc/fortran/ChangeLog trunk/gcc/fortran/match.c trunk/gcc/fortran/resolve.c trunk/gcc/testsuite/ChangeLog trunk/gcc/testsuite/gfortran.dg/class_7.f03 trunk/gcc/testsuite/gfortran.dg/select_type_1.f03
Subject: Bug 44044 Author: janus Date: Mon May 17 08:25:06 2010 New Revision: 159476 URL: http://gcc.gnu.org/viewcvs?root=gcc&view=rev&rev=159476 Log: 2010-05-17 Janus Weil <janus@gcc.gnu.org> PR fortran/44044 * resolve.c (resolve_fl_var_and_proc): Move error messages here from ... (resolve_fl_variable_derived): ... this place. (resolve_symbol): Make sure function symbols (and their result variables) are not resolved twice. 2010-05-17 Janus Weil <janus@gcc.gnu.org> PR fortran/44044 * gfortran.dg/class_20.f03: New. Added: trunk/gcc/testsuite/gfortran.dg/class_20.f03 Modified: trunk/gcc/fortran/ChangeLog trunk/gcc/fortran/resolve.c trunk/gcc/testsuite/ChangeLog
Comment #0 and #1 are fixed at this point. I think the runtime checking in comment #2 one can ignore for the moment, since most usage cases of ordinary pointers are also not checked for. So the only ToDo item left here is comment #3.
(In reply to comment #7) > So the only ToDo item left here is comment #3. Which is C836, i.e. checking invalid use of the selector in a variable definition context. This will probably be fixes as part of PR 38936 (ASSOCIATE construct; SELECT TYPE uses it internally) - thus mark this PR as depending on PR 38936.
Subject: Bug 44044 Author: domob Date: Thu Sep 23 08:37:54 2010 New Revision: 164550 URL: http://gcc.gnu.org/viewcvs?root=gcc&view=rev&rev=164550 Log: 2010-09-23 Daniel Kraft <d@domob.eu> PR fortran/38936 PR fortran/44044 PR fortran/45474 * gfortran.h (gfc_check_vardef_context): New method. (struct symbol_attribute): New flag `select_type_temporary'. * primary.c (gfc_variable_attr): Clarify initialization of ref. (match_variable): Remove PROTECTED check and assignment check for PARAMETERs (this is now done later). * match.c (gfc_match_iterator): Remove INTENT(IN) check. (gfc_match_associate): Defer initialization of newAssoc->variable. (gfc_match_nullify): Remove PURE definability check. (select_type_set_tmp): Set new `select_type_temporary' flag. * expr.c (gfc_check_assign): Remove INTENT(IN) check here. (gfc_check_pointer_assign): Ditto (and other checks removed). (gfc_check_vardef_context): New method. * interface.c (compare_parameter_protected): Removed. (compare_actual_formal): Use `gfc_check_vardef_context' for checks related to INTENT([IN]OUT) arguments. * intrinsic.c (check_arglist): Check INTENT for intrinsics. * resolve.c (gfc_resolve_iterator): Use `gfc_check_vardef_context'. (remove_last_array_ref): New method. (resolve_deallocate_expr), (resolve_allocate_expr): Ditto. (resolve_allocate_deallocate): Ditto (for STAT and ERRMSG). (resolve_assoc_var): Remove checks for definability here. (resolve_select_type): Handle resolving of code->block here. (resolve_ordinary_assign): Remove PURE check. (resolve_code): Do not resolve code->blocks for SELECT TYPE here. Use `gfc_check_vardef_context' for assignments and pointer-assignments. 2010-09-23 Daniel Kraft <d@domob.eu> PR fortran/38936 PR fortran/44044 PR fortran/45474 * gfortran.dg/intrinsic_intent_1.f03: New test. * gfortran.dg/select_type_17.f03: New test. * gfortran.dg/associate_5.f03: More definability tests. * gfortran.dg/enum_2.f90: Check definability. * gfortran.dg/allocatable_dummy_2.f90: Change expected error message. * gfortran.dg/allocate_alloc_opt_2.f90: Ditto. * gfortran.dg/char_expr_2.f90: Ditto. * gfortran.dg/deallocate_alloc_opt_2.f90: Ditto. * gfortran.dg/enum_5.f90: Ditto. * gfortran.dg/equiv_constraint_8.f90: Ditto. * gfortran.dg/impure_assignment_2.f90: Ditto. * gfortran.dg/impure_assignment_3.f90: Ditto. * gfortran.dg/intent_out_1.f90: Ditto. * gfortran.dg/intent_out_3.f90: Ditto. * gfortran.dg/pointer_assign_7.f90: Ditto. * gfortran.dg/pointer_intent_3.f90: Ditto. * gfortran.dg/pr19936_1.f90: Ditto. * gfortran.dg/proc_ptr_comp_3.f90: Ditto. * gfortran.dg/simpleif_2.f90: Ditto. * gfortran.dg/protected_5.f90: Ditto. * gfortran.dg/protected_4.f90: Ditto and remove invalid error check. * gfortran.dg/protected_6.f90: Ditto. * gfortran.dg/protected_7.f90: Ditto. Added: trunk/gcc/testsuite/gfortran.dg/intrinsic_intent_1.f03 trunk/gcc/testsuite/gfortran.dg/select_type_17.f03 Modified: trunk/gcc/fortran/ChangeLog trunk/gcc/fortran/expr.c trunk/gcc/fortran/gfortran.h trunk/gcc/fortran/interface.c trunk/gcc/fortran/intrinsic.c trunk/gcc/fortran/match.c trunk/gcc/fortran/primary.c trunk/gcc/fortran/resolve.c trunk/gcc/testsuite/ChangeLog trunk/gcc/testsuite/gfortran.dg/allocatable_dummy_2.f90 trunk/gcc/testsuite/gfortran.dg/allocate_alloc_opt_2.f90 trunk/gcc/testsuite/gfortran.dg/associate_5.f03 trunk/gcc/testsuite/gfortran.dg/char_expr_2.f90 trunk/gcc/testsuite/gfortran.dg/deallocate_alloc_opt_2.f90 trunk/gcc/testsuite/gfortran.dg/enum_2.f90 trunk/gcc/testsuite/gfortran.dg/enum_5.f90 trunk/gcc/testsuite/gfortran.dg/equiv_constraint_8.f90 trunk/gcc/testsuite/gfortran.dg/impure_assignment_2.f90 trunk/gcc/testsuite/gfortran.dg/impure_assignment_3.f90 trunk/gcc/testsuite/gfortran.dg/intent_out_1.f90 trunk/gcc/testsuite/gfortran.dg/intent_out_3.f90 trunk/gcc/testsuite/gfortran.dg/pointer_assign_7.f90 trunk/gcc/testsuite/gfortran.dg/pointer_intent_3.f90 trunk/gcc/testsuite/gfortran.dg/pr19936_1.f90 trunk/gcc/testsuite/gfortran.dg/proc_ptr_comp_3.f90 trunk/gcc/testsuite/gfortran.dg/protected_4.f90 trunk/gcc/testsuite/gfortran.dg/protected_5.f90 trunk/gcc/testsuite/gfortran.dg/protected_6.f90 trunk/gcc/testsuite/gfortran.dg/protected_7.f90 trunk/gcc/testsuite/gfortran.dg/simpleif_2.f90
This implemented bonus feature #2 from comment #3. Closing now.