I recently added an assert to cp-gimplify to catch any
TARGET_EXPR_DIRECT_INIT_P being expanded without a target object, and this
testcase found one. We started out with a TARGET_EXPR around the
CONSTRUCTOR, which would normally mean that the member initializer would be
used to directly initialize the appropriate member of whatever object the
TARGET_EXPR ends up initializing. But then gimplify_modify_expr_rhs
stripped the TARGET_EXPR in order to assign directly from the elements of
the CONSTRUCTOR, leaving no object for the TARGET_EXPR_DIRECT_INIT_P to
initialize. I considered setting CONSTRUCTOR_PLACEHOLDER_BOUNDARY in that
case, which implies TARGET_EXPR_NO_ELIDE, but decided that there's no
particular reason the A initializer needs to initialize a member of a B
rather than a distinct A object, so let's only set TARGET_EXPR_DIRECT_INIT_P
when we're using the DMI in a constructor.
* init.c (get_nsdmi): Set TARGET_EXPR_DIRECT_INIT_P here.
* typeck2.c (digest_nsdmi_init): Not here.
+2020-01-16 Jason Merrill <jason@redhat.com>
+
+ PR c++/93280 - ICE with aggregate assignment and DMI.
+ * init.c (get_nsdmi): Set TARGET_EXPR_DIRECT_INIT_P here.
+ * typeck2.c (digest_nsdmi_init): Not here.
+
2020-01-15 Paolo Carlini <paolo.carlini@oracle.com>
PR c++/91073
if (simple_target)
init = TARGET_EXPR_INITIAL (init);
init = break_out_target_exprs (init, /*loc*/true);
+ if (in_ctor && init && TREE_CODE (init) == TARGET_EXPR)
+ /* This expresses the full initialization, prevent perform_member_init from
+ calling another constructor (58162). */
+ TARGET_EXPR_DIRECT_INIT_P (init) = true;
if (simple_target && TREE_CODE (init) != CONSTRUCTOR)
/* Now put it back so C++17 copy elision works. */
init = get_target_expr (init);
&& CP_AGGREGATE_TYPE_P (type))
init = reshape_init (type, init, complain);
init = digest_init_flags (type, init, flags, complain);
- if (TREE_CODE (init) == TARGET_EXPR)
- /* This represents the whole initialization. */
- TARGET_EXPR_DIRECT_INIT_P (init) = true;
return init;
}
\f