[Bug c++/11548] New: [3.4 regression] Forwarding default template args ICEs

[Bug c++/11548] New: [3.4 regression] Forwarding default template args ICEs

[bangerth at dealii dot org]

PLEASE REPLY TO gcc-bugzilla@gcc.gnu.org ONLY, *NOT* gcc-bugs@gcc.gnu.org.

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=11548

           Summary: [3.4 regression] Forwarding default template args ICEs
           Product: gcc
           Version: 3.4
            Status: UNCONFIRMED
          Severity: normal
          Priority: P2
         Component: c++
        AssignedTo: unassigned at gcc dot gnu dot org
        ReportedBy: bangerth at dealii dot org
                CC: gcc-bugs at gcc dot gnu dot org

This was recently broken:
---------------------------------
template <typename> struct length {
    static const unsigned int value = 0;
};

template <typename T, int N = length<T>::value> class C;
template <typename T>                           class C<T, 0> {};

template <typename> C<int> foo ();
----------------------------------
Note that we need to look at the default argument of the general
template to realize that the value of N in C<int> equals zero, and
then take the partial specialization of C<T,0>. It's a funny idiom (used
in boost, for example), but it shouldn't ICE:

tmp/gg> ../build-gcc/gcc-install/bin/c++ -c x.cc
x.cc:8: internal compiler error: Segmentation fault
Please submit a full bug report,
with preprocessed source if appropriate.
See <URL:http://gcc.gnu.org/bugs.html> for instructions.

A backtrace in the compiler is this:
#0  convert_nontype_argument (type=0x4001e57c, expr=0x40023348)
    at ../../gcc/gcc/cp/pt.c:2999
#1  0x080734bb in convert_template_argument (parm=0x403d9ca8, arg=0x40023348, 
    args=0x40023330, complain=tf_error, i=1, in_decl=0x403d9f30)
    at ../../gcc/gcc/cp/pt.c:3574
#2  0x08073619 in coerce_template_parms (parms=0x400232e8, args=0x2, 
    in_decl=0x403d9f30, complain=tf_error, require_all_arguments=1)
    at ../../gcc/gcc/cp/pt.c:3676
#3  0x08074ad8 in lookup_template_class (d1=0x403dc118, arglist=0x403dc410, 
    in_decl=0x0, context=0x0, entering_scope=0, complain=tf_error)
    at ../../gcc/gcc/cp/pt.c:4175
#4  0x080c3867 in finish_template_type (name=0x4001e57c, args=0x4001e57c, 
    entering_scope=1073866108) at ../../gcc/gcc/cp/semantics.c:2172

The crash is in the if()-clause of this snippet, since expr_type==0:
  /* An integral constant-expression can include const variables or
     enumerators.  Simplify things by folding them to their values,
     unless we're about to bind the declaration to a reference
     parameter.  */
  if (INTEGRAL_TYPE_P (expr_type)
      && TREE_CODE (type) != REFERENCE_TYPE)
    expr = decl_constant_value (expr);

None of the code in the lowest 3 frames seems to have been changed
recently, so I can't tell which the breaking patch is. Unfortunately,
I also can't see a work-around for the problem, short of actually
writing in the value of the default argument at the point of use :-(

W.

[Bug c++/11548] [3.4 regression] Forwarding default template args ICEs

[pinskia at physics dot uc dot edu]

PLEASE REPLY TO gcc-bugzilla@gcc.gnu.org ONLY, *NOT* gcc-bugs@gcc.gnu.org.

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=11548


pinskia at physics dot uc dot edu changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
             Status|UNCONFIRMED                 |NEW
     Ever Confirmed|                            |1
   Last reconfirmed|0000-00-00 00:00:00         |2003-07-16 16:01:04
               date|                            |


------- Additional Comments From pinskia at physics dot uc dot edu  2003-07-16 16:01 -------
I can confirm this on the mainline (20030715).
Looks like an other fall out from Mark's patch on 2003-07-08
: Search converges between 2003-07-08-trunk (#330) and 2003-07-09-trunk (#331). 
2003-07-08  Mark Mitchell

        * cp-tree.def (NON_DEPENDENT_EXPR): New node.
        * cp-tree.h (build_call_from_tree): Remove.
        (build_member_call): Likewise.
        (dependent_template_arg_p): Remove.
        (any_dependent_template_arguments_p): New function.
        (dependent_template_id_p): Likewise.
        (any_type_dependent_arguments_p): Likewise.
        (build_non_dependent_expr): Likewise.
        (build_non_dependent_args): Likewise.
        (build_x_compound_expr): Adjust prototype.
        * call.c (build_new_method_call): Handle non-dependent expressions
        correctly.
        * decl2.c (grok_array_decl): Likewise.
        (build_offset_ref_call_from_tree): Likewise.
        (build_call_from_tree): Remove.
        * error.c (dump_decl): Handle NON_DEPENDENT_EXPR.
        (dump_expr): Likewise.
        * init.c (build_member_call): Remove.
        * mangle.c (write_expression): Update handling for template-ids.
        * parser.c (cp_parser_primary_expression): Use
        any_dependent_template_arguments_p.  Update constant-expression
        handling.
        (cp_parser_postfix_expression): Use
        any_type_dependent_arguments_p.  Simplify call processing.
        (cp_parser_unary_expression): Simplify.
        (cp_parser_expression): Adjust for changes to
        build_x_compound_expr.
        (cp_parser_template_argument): Implement standard-conforming
        parsing of non-type template arguments.
        (cp_parser_direct_declarator): Use
        cp_parser_fold_non_dependent_expr.
        (cp_parser_fold_non_dependent_expr): New function.
        (cp_parser_next_token_ends_template_argument_p): Likewise.
        * pt.c (convert_template_argument): Do not call
        maybe_fold_nontype_arg.
        (tsubst_baselink): Likewise.
        (tsubst_copy_and_build): Share common code.  Make sizeof/alignof
        processing work correctly for non-dependent expressions.  Adjust
        handling of COMPOUND_EXPR.  Simplify call processing.
        (value_dependent_expression_p): Deal with functional casts and
        sizeof/alignof correctly.
        (type_dependent_expression_p): Handle overloaded functions.
        (any_type_dependent_arguments_p): New function.
        (any_dependent_template_arguments_p): Likewise.
        (dependent_template_p): Treat SCOPE_REFs as dependent.
        (dependent_template_id_p): Simplify.
        (build_non_dependent_expr): New function.
        (build_non_dependent_args): Likewise.
        * semantics.c (finish_stmt_expr): Don't make dependent
        statement-expresions have void type.
        (finish_call_expr): Handle non-dependent expressions
        correctly.
        * tree.c (lvalue_p_1): Treat NON_DEPENDENT_EXPRs as lvalues.
        * typeck.c (cxx_sizeof_or_alignof_type): Give the expression
        type size_t, even in templates.
        (expr_sizeof): Likewise.
        (finish_class_member_access_expr): Handle non-dependent expressions
        correctly.
        (build_x_indirect_ref): Likewise.
        (build_x_binary_op): Likewise.
        (build_x_unary_op): Likewise.
        (build_x_conditional_expr): Likewise.
        (build_x_compound_expr): Likewise.
        * typeck2.c (build_x_arrow): Likewise.

[Bug c++/11548] [3.4 regression] Forwarding default template args ICEs

[nathan at gcc dot gnu dot org]

PLEASE REPLY TO gcc-bugzilla@gcc.gnu.org ONLY, *NOT* gcc-bugs@gcc.gnu.org.

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=11548


nathan at gcc dot gnu dot org changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
         AssignedTo|unassigned at gcc dot gnu   |nathan at gcc dot gnu dot
                   |dot org                     |org
             Status|NEW                         |ASSIGNED

[Bug c++/11548] [3.4 regression] Forwarding default template args ICEs

[nathan at gcc dot gnu dot org]

PLEASE REPLY TO gcc-bugzilla@gcc.gnu.org ONLY, *NOT* gcc-bugs@gcc.gnu.org.

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=11548


nathan at gcc dot gnu dot org changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
             Status|ASSIGNED                    |RESOLVED
         Resolution|                            |DUPLICATE


------- Additional Comments From nathan at gcc dot gnu dot org  2003-07-25 13:05 -------


*** This bug has been marked as a duplicate of 11596 ***