[Bug c++/64100] New: A static assert using the the current class in a noexcept test leads to a segfault

[Bug c++/64100] New: A static assert using the the current class in a noexcept test leads to a segfault

[ville.voutilainen at gmail dot com]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=64100

            Bug ID: 64100
           Summary: A static assert using the the current class in a
                    noexcept test leads to a segfault
           Product: gcc
           Version: 5.0
            Status: UNCONFIRMED
          Severity: normal
          Priority: P3
         Component: c++
          Assignee: unassigned at gcc dot gnu.org
          Reporter: ville.voutilainen at gmail dot com

template<typename> struct foo 
{ 
    static_assert(noexcept(((foo *)1)->~foo()), ""); 
}; 

template class foo<int>;

Clang diagnoses the use of the constructor of foo as an attempt to
use an incomplete type, gcc segfaults. See
http://melpon.org/wandbox/permlink/Xz0rchej4eij4tOA
for the actual trace.

This looks like a regression, since my build from 20141121 just silently
eats the code. 20141127 produces the segfault.

[Bug c++/64100] A static assert using the the current class in a noexcept test leads to a segfault

[mpolacek at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=64100

Marek Polacek <mpolacek at gcc dot gnu.org> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
             Status|UNCONFIRMED                 |NEW
   Last reconfirmed|                            |2014-12-03
                 CC|                            |mpolacek at gcc dot gnu.org
   Target Milestone|---                         |4.8.4
     Ever confirmed|0                           |1

--- Comment #1 from Marek Polacek <mpolacek at gcc dot gnu.org> ---
I see ICE even with 4.6.

[Bug c++/64100] A static assert using the the current class in a noexcept test leads to a segfault

[ktietz at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=64100

Kai Tietz <ktietz at gcc dot gnu.org> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
                 CC|                            |ktietz at gcc dot gnu.org

--- Comment #2 from Kai Tietz <ktietz at gcc dot gnu.org> ---
Issue is that lookup_destructor calls adjust_result_of_qualified_name_lookup
with an NULL_TREE decl (returned by lookup_member).  Here the error-check seems
to be missing.

Following patch fixes this issue for me:
Index: typeck.c
===================================================================
--- typeck.c    (Revision 218142)
+++ typeck.c    (Arbeitskopie)
@@ -2536,6 +2536,11 @@ lookup_destructor (tree object, tree scope, tree d
   expr = lookup_member (dtor_type, complete_dtor_identifier,
                        /*protect=*/1, /*want_type=*/false,
                        tf_warning_or_error);
+  if (!expr)
+    {
+      cxx_incomplete_type_error (dtor_name, dtor_type);
+      return error_mark_node;
+    }
   expr = (adjust_result_of_qualified_name_lookup
          (expr, dtor_type, object_type));
   if (scope == NULL_TREE)
Index: search.c
===================================================================
--- search.c    (Revision 218142)
+++ search.c    (Arbeitskopie)
@@ -1530,6 +1530,9 @@ adjust_result_of_qualified_name_lookup (tree decl,
                                        tree qualifying_scope,
                                        tree context_class)
 {
+  if (!decl)
+    return NULL_TREE;
+
   if (context_class && context_class != error_mark_node
       && CLASS_TYPE_P (context_class)
       && CLASS_TYPE_P (qualifying_scope)

The change to search.c isn't really required, but avoids to show the ice in
adjust_result_of_qualifying.

[Bug c++/64100] A static assert using the the current class in a noexcept test leads to a segfault

[jakub at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=64100

Jakub Jelinek <jakub at gcc dot gnu.org> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
                 CC|                            |jakub at gcc dot gnu.org

--- Comment #3 from Jakub Jelinek <jakub at gcc dot gnu.org> ---
Yeah, r160250 still errored out on it, r160311 already ICEs.

[Bug c++/64100] A static assert using the the current class in a noexcept test leads to a segfault

[jason at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=64100

Jason Merrill <jason at gcc dot gnu.org> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
                 CC|                            |jason at gcc dot gnu.org

--- Comment #4 from Jason Merrill <jason at gcc dot gnu.org> ---
(In reply to Kai Tietz from comment #2)
> Following patch fixes this issue for me:

Were you going to send the patch to the list?

> +  if (!expr)
> +    {
> +      cxx_incomplete_type_error (dtor_name, dtor_type);
> +      return error_mark_node;
> +    }

The error needs to be conditional on (complain & tf_error).

Does the caller handle error_mark_node properly?

[Bug c++/64100] A static assert using the the current class in a noexcept test leads to a segfault

[ktietz at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=64100

Kai Tietz <ktietz at gcc dot gnu.org> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
             Status|NEW                         |ASSIGNED

--- Comment #5 from Kai Tietz <ktietz at gcc dot gnu.org> ---
(In reply to Jason Merrill from comment #4)
> (In reply to Kai Tietz from comment #2)
> > Following patch fixes this issue for me:
> 
> Were you going to send the patch to the list?

Yes, I will send it today to ML.  I will put you CC.

> > +  if (!expr)
> > +    {
> > +      cxx_incomplete_type_error (dtor_name, dtor_type);
> > +      return error_mark_node;
> > +    }
> 
> The error needs to be conditional on (complain & tf_error).

Ok, I'll add that.

> Does the caller handle error_mark_node properly?

Yes, I tested it before.  By just returning error_mark_node the testcase will
be processed completely silently by g++.

[Bug c++/64100] A static assert using the the current class in a noexcept test leads to a segfault

[ktietz at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=64100

--- Comment #6 from Kai Tietz <ktietz at gcc dot gnu.org> ---
Author: ktietz
Date: Wed Dec 10 11:22:34 2014
New Revision: 218571

URL: https://gcc.gnu.org/viewcvs?rev=218571&root=gcc&view=rev
Log:
    PR c++/64100
    * typeck.c (lookup_destructor): Handle incomplete type.


Modified:
    trunk/gcc/cp/ChangeLog
    trunk/gcc/cp/typeck.c

[Bug c++/64100] A static assert using the the current class in a noexcept test leads to a segfault

[ktietz at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=64100

--- Comment #7 from Kai Tietz <ktietz at gcc dot gnu.org> ---
Author: ktietz
Date: Wed Dec 10 11:26:47 2014
New Revision: 218572

URL: https://gcc.gnu.org/viewcvs?rev=218572&root=gcc&view=rev
Log:
    PR c++/64100
    * g++.dg/template/pr64100.C: New file.


Added:
    trunk/gcc/testsuite/g++.dg/template/pr64100.C
Modified:
    trunk/gcc/testsuite/ChangeLog

[Bug c++/64100] A static assert using the the current class in a noexcept test leads to a segfault

[ktietz at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=64100

Kai Tietz <ktietz at gcc dot gnu.org> changed:

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

--- Comment #8 from Kai Tietz <ktietz at gcc dot gnu.org> ---
Fixed.  As it is an error on invalid-code, it won't get backported
automatically.  Close it.