https://gcc.gnu.org/bugzilla/show_bug.cgi?id=88294
David Malcolm <dmalcolm at gcc dot gnu.org> changed:
What |Removed |Added
----------------------------------------------------------------------------
CC| |dmalcolm at gcc dot gnu.org
--- Comment #3 from David Malcolm <dmalcolm at gcc dot gnu.org> ---
It's a failing assertion:
15401 gcc_assert (cp_unevaluated_operand != 0);
within tsubst_copy's PARM_DECL handling of a "this" PARM_DECL:
(gdb) p t
$2 = <parm_decl 0x7ffff18d7380 this>
15389 case PARM_DECL:
15390 r = retrieve_local_specialization (t);
15391
15392 if (r == NULL_TREE)
15393 {
15394 /* We get here for a use of 'this' in an NSDMI. */
15395 if (DECL_NAME (t) == this_identifier && current_class_ptr)
15396 return current_class_ptr;
15397
15398 /* This can happen for a parameter name used later in a
function
15399 declaration (such as in a late-specified return type).
Just
15400 make a dummy decl, since it's only used for its type. */
15401 gcc_assert (cp_unevaluated_operand != 0);
15402 r = tsubst_decl (t, args, complain);
where the local specialization is NULL.
The PARM_DECL in question is created by the hunk from r266224, which is calling
inject_this_parameter for the "noexcept" fn, in
cp_parser_noexcept_specification_opt
25032 if (current_class_type)
25033 inject_this_parameter (current_class_type,
TYPE_UNQUALIFIED);
This sets current_class_ptr to the "this" PARM_DECL.
It then resets current_class_ptr back to NULL here:
25055 current_class_ptr = save_ccp;
Later, in tsubst_copy on the PARM_DECL in the code quoted above, the PARM_DECL
matches only part of the test here:
15395 if (DECL_NAME (t) == this_identifier && current_class_ptr)
15396 return current_class_ptr;
in that it's DECL_NAME is indeed this_identifier, but current_class_ptr is
NULL, and so we carry on to the assertion:
15401 gcc_assert (cp_unevaluated_operand != 0);
which fails.
