Re: [PATCH] c++: noexcept and copy elision [PR109030]

[Patrick Palka <ppalka@redhat.com>]

On Thu, 9 Mar 2023, Jason Merrill wrote:

> On 3/9/23 14:32, Patrick Palka wrote:
> > On Mon, 6 Mar 2023, Marek Polacek via Gcc-patches wrote:
> > 
> > > When processing a noexcept, constructors aren't elided: build_over_call
> > > has
> > > 	 /* It's unsafe to elide the constructor when handling
> > > 	    a noexcept-expression, it may evaluate to the wrong
> > > 	    value (c++/53025).  */
> > > 	 && (force_elide || cp_noexcept_operand == 0))
> > > so the assert I added recently needs to be relaxed a little bit.
> > > 
> > > Bootstrapped/regtested on x86_64-pc-linux-gnu, ok for trunk?
> > > 
> > > 	PR c++/109030
> > > 
> > > gcc/cp/ChangeLog:
> > > 
> > > 	* constexpr.cc (cxx_eval_call_expression): Relax assert.
> > > 
> > > gcc/testsuite/ChangeLog:
> > > 
> > > 	* g++.dg/cpp0x/noexcept77.C: New test.
> > > ---
> > >   gcc/cp/constexpr.cc                     | 6 +++++-
> > >   gcc/testsuite/g++.dg/cpp0x/noexcept77.C | 9 +++++++++
> > >   2 files changed, 14 insertions(+), 1 deletion(-)
> > >   create mode 100644 gcc/testsuite/g++.dg/cpp0x/noexcept77.C
> > > 
> > > diff --git a/gcc/cp/constexpr.cc b/gcc/cp/constexpr.cc
> > > index 364695b762c..5384d0e8e46 100644
> > > --- a/gcc/cp/constexpr.cc
> > > +++ b/gcc/cp/constexpr.cc
> > > @@ -2869,7 +2869,11 @@ cxx_eval_call_expression (const constexpr_ctx *ctx,
> > > tree t,
> > >       /* We used to shortcut trivial constructor/op= here, but nowadays
> > >        we can only get a trivial function here with
> > > -fno-elide-constructors.  */
> > > -  gcc_checking_assert (!trivial_fn_p (fun) || !flag_elide_constructors);
> > > +  gcc_checking_assert (!trivial_fn_p (fun)
> > > +		       || !flag_elide_constructors
> > > +		       /* We don't elide constructors when processing
> > > +			  a noexcept-expression.  */
> > > +		       || cp_noexcept_operand);
> > 
> > It seems weird that we're performing constant evaluation within an
> > unevaluated operand.  Would it make sense to also fix this a second way
> > by avoiding constant evaluation from maybe_constant_init when
> > cp_unevaluated_operand && !manifestly_const_eval, like in
> > maybe_constant_value?
> 
> Sounds good.

Hmm, while working on this I noticed we currently don't reject a version of
g++.dg/cpp2a/constexpr-inst1.C that list initializes an aggregate instead of
int (ever since r12-4425-g1595fe44e11a96):

  struct A { int m; };
  template<typename T> constexpr int f() { return T::value; }
  template<bool B, typename T> void h(decltype(A{B ? f<T>() : 0})); // was int{...}
  template<bool B, typename T> void h(...);
  void x() {
    h<false, int>(0); // OK?
  }

ISTM we should instantiate f<int> here for the same reason we do in the
original version of the testcase, and for that to happen we need to
pass manifestly_const_eval=true in massage_init_elt.  Does that seem
reasonable?