On 3/7/25 11:53 AM, Jakub Jelinek wrote:
Hi!
Since Marek's r14-4140 which moved immediate invocation evaluation
from build_over_call to cp_fold_r, the following testcase is miscompiled.
The a = foo (bar ()); case is actually handled right, that is handled
in cp_fold_r and the whole CALL_EXPR is at that point evaluated by
cp_fold_immediate_r with cxx_constant_value (stmt, tf_none);
and that uses mce_true for evaluation of the argument as well as the actual
call.
But in the bool b = foo (bar ()); case we actually try to evaluate this
as non-manifestly constant-evaluated. And while
/* Make sure we fold std::is_constant_evaluated to true in an
immediate function. */
if (DECL_IMMEDIATE_FUNCTION_P (fun))
call_ctx.manifestly_const_eval = mce_true;
ensures that if consteval and __builtin_is_constant_evaluated () is true
inside of that call, this happens after arguments to the function
have been already constant evaluated in cxx_bind_parameters_in_call.
The call_ctx in that case also includes new call_ctx.call, something that
shouldn't be used for the arguments, so the following patch just arranges
to call cxx_bind_parameters_in_call with manifestly_constant_evaluated =
mce_true.
Bootstrapped/regtested on x86_64-linux and i686-linux, ok for trunk?
2025-03-07 Jakub Jelinek <ja...@redhat.com>
PR c++/119150
* constexpr.cc (cxx_eval_call_expression): Call
cxx_bind_parameters_in_call for immediate invocations with
manifestly_const_eval set to mce_true.
* g++.dg/cpp2a/consteval41.C: New test.
--- gcc/cp/constexpr.cc.jj 2025-03-01 09:13:17.694075636 +0100
+++ gcc/cp/constexpr.cc 2025-03-07 13:45:35.909164389 +0100
@@ -3074,9 +3074,21 @@ cxx_eval_call_expression (const constexp
|| cp_noexcept_operand);
bool non_constant_args = false;
- new_call.bindings
- = cxx_bind_parameters_in_call (ctx, t, fun, non_constant_p,
- overflow_p, &non_constant_args);
+ /* An immediate invocation is manifestly constant evaluated including the
+ arguments of the call, so use mce_true for the argument evaluation. */
+ if (DECL_IMMEDIATE_FUNCTION_P (fun)
+ && ctx->manifestly_const_eval != mce_true)
+ {
+ constexpr_ctx call_ctx = *ctx;
Why not use new_ctx for this instead of creating another copy?
I would think once we see that we have an immediate function, we want to
go ahead and set ctx->manifestly_const_eval and
new_call->manifestly_const_eval to true. It looks like we currently
cache immediate invocations separately depending on whether they're in
an enclosing immediate function context, which seems redundant.
Incidentally, I don't remember why we need a separate call_ctx either.
Jason
