https://gcc.gnu.org/bugzilla/show_bug.cgi?id=87893
Jakub Jelinek <jakub at gcc dot gnu.org> changed:
What |Removed |Added
----------------------------------------------------------------------------
CC| |jakub at gcc dot gnu.org
--- Comment #2 from Jakub Jelinek <jakub at gcc dot gnu.org> ---
When trying to cp_fold pair::pair(&p) ctor call, guess related to arm ABI
feature of returning pointer to the object from the ctor,
2542 if (callee && DECL_DECLARED_CONSTEXPR_P (callee)
2543 && !flag_no_inline)
2544 r = maybe_constant_value (x);
returns for some reason a VOID_CST and then
2549 if (DECL_CONSTRUCTOR_P (callee))
2550 {
2551 loc = EXPR_LOCATION (x);
2552 tree s = build_fold_indirect_ref_loc (loc,
2553 CALL_EXPR_ARG (x,
0));
2554 r = build2_loc (loc, INIT_EXPR, TREE_TYPE (s), s, r);
2555 }
tries to initialize the class with that. VOID_CST is a result of the ctor call
constexpr evaluation doing:
if (DECL_CONSTRUCTOR_P (fun))
/* This can be null for a subobject constructor call, in
which case what we care about is the initialization
side-effects rather than the value. We could get at the
value by evaluating *this, but we don't bother; there's
no need to put such a call in the hash table. */
result = lval ? ctx->object : ctx->ctor;
where for some reason both ctx->object and ctx->ctor are NULL and then
else if (!result)
result = void_node;
both in cxx_eval_call_expression. No idea why that happens though and why
pair::pair(&p) call isn't being folded at all on the same testcase on
x86_64-linux.
