On 5/12/23 13:02, Patrick Palka wrote:
Hi Martin,
On Fri, May 12, 2023 at 12:13 PM Martin Jambor <mjam...@suse.cz> wrote:
Hello Patrick,
On Wed, May 03 2023, Patrick Palka via Gcc-patches wrote:
[...]
Subject: [PATCH] c++: potentiality of templated memfn call [PR109480]
Here we're incorrectly deeming the templated call a.g() inside b's
initializer as potentially constant, despite g being non-constexpr,
which leads to us wastefully instantiating the initializer ahead of time,
which incidentally tiggers a bug in access checking deferral (to be
fixed by the subsequent patch).
This patch fixes this by calling get_fns earlier during CALL_EXPR
potentiality checking so that we're able to extract a FUNCTION_DECL out
of a templated member function call (whose overall is typically a
COMPONENT_REF) and to the usual checking if the called function is
constexpr etc.
In passing, I noticed potential_constant_expression_1's special handling
of the object argument of a non-static member function call is effectively
the same as the generic argument handling a few lines later. So this
patch just gets rid of this special handling; otherwise we'd have to adapt
it to handle templated versions of such calls.
PR c++/109480
gcc/cp/ChangeLog:
* constexpr.cc (potential_constant_expression_1) <case CALL_EXPR>:
Reorganize to call get_fns sooner. Remove special handling of
the object argument of a non-static member function call. Remove
dead store to 'fun'.
This patch makes g++ no longer accept the following, complaining that
get_subsys is non-constexpr (with just -std=c++17 -S), which is of
course auto-reduced from a much larger source file from Ceph:
----------------------------------- 8< -----------------------------------
struct {
void get_subsys();
} PriorSet_dpp;
struct PriorSet {
template <typename> PriorSet();
};
template <typename> PriorSet::PriorSet() {
[](auto cctX) { cctX.template should_gather<PriorSet_dpp.get_subsys()>; };
}
----------------------------------- 8< -----------------------------------
I assume that is intentional and am actually somewhat surprised it was
accepted before, but can you please confirm?
Yes, this seems correct/intentional to me-- no instantiation of the
template would be valid because it's trying to use a non-constant
expression (which we now correctly identify as such) as a template
argument, so this snippet is IFNDR.
I don't think we have testsuite coverage for this QoI diagnostic, I'll add one.
Incidentally, I wonder about trying to make IFNDR diags in general
permerrors or default-error pedwarns, but that doesn't need to happen now.
Jason