Re: [PATCH v3] c++: fixes for derived-to-base reference binding [PR107085]

[Jason Merrill <jason@redhat.com>]

On 10/6/22 17:43, Marek Polacek wrote:
> On Thu, Oct 06, 2022 at 02:00:40PM -0400, Jason Merrill wrote:
>> On 10/6/22 13:51, Marek Polacek wrote:
>>> On Thu, Oct 06, 2022 at 10:58:44AM -0400, Jason Merrill wrote:
>>>> On 10/6/22 10:49, Marek Polacek wrote:
>>>>> On Wed, Oct 05, 2022 at 08:25:29PM -0400, Jason Merrill wrote:
>>>>>> On 10/5/22 17:27, Marek Polacek wrote:
>>>>>>> This PR reports that
>>>>>>>
>>>>>>>       struct Base {};
>>>>>>>       struct Derived : Base {};
>>>>>>>       static_assert(__reference_constructs_from_temporary(Base const&, Derived));
>>>>>>>
>>>>>>> doesn't pass, which it should: it's just like
>>>>>>>
>>>>>>>       const Base& b(Derived{});
>>>>>>>
>>>>>>> where we bind 'b' to the Base subobject of a temporary object of type
>>>>>>> Derived.  The ck_base conversion didn't have ->need_temporary_p set because
>>>>>>> we didn't need to create a temporary object just for the base, but the whole
>>>>>>> object is a temporary so we're still binding to a temporary.  Fixed by
>>>>>>> the conv_is_prvalue hunk.
>>>>>>>
>>>>>>> That broke a bunch of tests.  I've distilled the issue into a simple test
>>>>>>> in elision4.C.  Essentially, we have
>>>>>>>
>>>>>>>       struct B { /* ... */ };
>>>>>>>       struct D : B { };
>>>>>>>       B b = D();
>>>>>>>
>>>>>>> and we set force_elide in build_over_call, but we're unable to actually
>>>>>>> elide the B::B(B&&) call, and crash on gcc_assert (!force_elide);.
>>>>>>>
>>>>>>> <https://en.cppreference.com/w/cpp/language/copy_elision> says that copy
>>>>>>> elision "can only apply when the object being initialized is known not to be
>>>>>>> a potentially-overlapping subobject".  So I suppose we shouldn't force_elide
>>>>>>> the B::B(B&&) call.  I don't belive the CWG 2327 code was added to handle
>>>>>>> derived-to-base conversions, at that time conv_binds_ref_to_prvalue wasn't
>>>>>>> checking ck_base at all.
>>>>>>>
>>>>>>> Does that make sense?  If so...
>>>>>>>
>>>>>>> Bootstrapped/regtested on x86_64-pc-linux-gnu, ok for trunk?
>>>>>>>
>>>>>>> 	PR c++/107085
>>>>>>>
>>>>>>> gcc/cp/ChangeLog:
>>>>>>>
>>>>>>> 	* call.cc (conv_is_prvalue): Return true if the base subobject is part
>>>>>>> 	of a temporary object.
>>>>>>
>>>>>> No, the base subobject of a prvalue is an xvalue.
>>>>>
>>>>> Ah, so this is just like T().m where T() is a prvalue but the whole thing
>>>>> is an xvalue.  Duly noted.
>>>>
>>>> Exactly.
>>>>
>>>>>> I think the problem is that an expression being a prvalue is a subset of
>>>>>> binding a reference to a temporary, and we shouldn't try to express both of
>>>>>> those using the same function: you need a separate
>>>>>> conv_binds_ref_to_temporary.
>>>>>
>>>>> Ack, so how about this?  Thanks,
>>>>>
>>>>> Bootstrapped/regtested on x86_64-pc-linux-gnu, ok for trunk?
>>>>>
>>>>> -- >8 --
>>>>> This PR reports that
>>>>>
>>>>>      struct Base {};
>>>>>      struct Derived : Base {};
>>>>>      static_assert(__reference_constructs_from_temporary(Base const&, Derived));
>>>>>
>>>>> doesn't pass, which it should: it's just like
>>>>>
>>>>>      const Base& b(Derived{});
>>>>>
>>>>> where we bind 'b' to the Base subobject of a temporary object of type
>>>>> Derived.  The ck_base conversion didn't have ->need_temporary_p set because
>>>>> we didn't need to create a temporary object just for the base, but the whole
>>>>> object is a temporary so we're still binding to a temporary.  Since the
>>>>> Base subobject is an xvalue, a new function is introduced.
>>>>>
>>>>> 	PR c++/107085
>>>>>
>>>>> gcc/cp/ChangeLog:
>>>>>
>>>>> 	* call.cc (conv_binds_ref_to_temporary): New.
>>>>> 	(ref_conv_binds_directly): Use it.
>>>>>
>>>>> gcc/testsuite/ChangeLog:
>>>>>
>>>>> 	* g++.dg/ext/reference_constructs_from_temporary1.C: Adjust expected
>>>>> 	result.
>>>>> 	* g++.dg/ext/reference_converts_from_temporary1.C: Likewise.
>>>>> 	* g++.dg/cpp0x/elision4.C: New test.
>>>>> ---
>>>>>     gcc/cp/call.cc                                | 23 ++++++++++++++++++-
>>>>>     gcc/testsuite/g++.dg/cpp0x/elision4.C         | 15 ++++++++++++
>>>>>     .../reference_constructs_from_temporary1.C    |  2 +-
>>>>>     .../ext/reference_converts_from_temporary1.C  |  2 +-
>>>>>     4 files changed, 39 insertions(+), 3 deletions(-)
>>>>>     create mode 100644 gcc/testsuite/g++.dg/cpp0x/elision4.C
>>>>>
>>>>> diff --git a/gcc/cp/call.cc b/gcc/cp/call.cc
>>>>> index bd04a1d309a..715a83f5a69 100644
>>>>> --- a/gcc/cp/call.cc
>>>>> +++ b/gcc/cp/call.cc
>>>>> @@ -9210,6 +9210,27 @@ conv_binds_ref_to_prvalue (conversion *c)
>>>>>       return conv_is_prvalue (next_conversion (c));
>>>>>     }
>>>>> +/* True iff C is a conversion that binds a reference to a temporary.
>>>>> +   This is a superset of conv_binds_ref_to_prvalue: here we're also
>>>>> +   interested in xvalues.  */
>>>>> +
>>>>> +static bool
>>>>> +conv_binds_ref_to_temporary (conversion *c)
>>>>> +{
>>>>> +  if (conv_binds_ref_to_prvalue (c))
>>>>> +    return true;
>>>>> +  if (c->kind != ck_ref_bind)
>>>>> +    return false;
>>>>> +  c = next_conversion (c);
>>>>> +  /* This is the case for
>>>>> +       struct Base {};
>>>>> +       struct Derived : Base {};
>>>>> +       const Base& b(Derived{});
>>>>> +     where we bind 'b' to the Base subobject of a temporary object of type
>>>>> +     Derived.  The subobject is an xvalue; the whole object is a prvalue.  */
>>>>> +  return (c->kind == ck_base && conv_is_prvalue (next_conversion (c)));
>>>>
>>>> I think you also want to check for the case of c->u.expr being a
>>>> COMPONENT_REF/ARRAY_REF around a TARGET_EXPR, as you mentioned.
>>>
>>> I see.  So this would be achieved using e.g.
>>>
>>> struct B { };
>>> struct D : B { };
>>> struct C {
>>>     D d;
>>> };
>>>
>>> const B& b = C{}.d;
>>
>> Yes.
>>
>>> Except I'm not sure how to trigger this via the built-in, which takes two types.
>>> Am I missing something obvious?
>>
>> Indeed, it can't be triggered by the built-in.  But I see
>> ref_conv_binds_directly is also called from warn_for_range_copy, which ought
>> to be able to trigger it.
> 
> Even there, I think, it's not possible to trigger it with, say (d is an
> array of D):
> 
>       for (const B &b : C{}.d)
>         // ...
> 
> because warn_for_range_copy gets *__for_begin as the EXPR, which means that
> we won't get to the TARGET_EXPR.  It's like
> 
>    auto&& __for_begin = C{}.d;
>    const B &b = *__for_begin;
> 
> and the conversion warn_for_range_copy sees is D -> const B -> const B&, the
> original .u.expr is *__for_begin.

Ah, right, and so the C{} temporary gets lifetime-extended and there's 
no problem to warn about.

> I could add some checking assert to conv_binds_ref_to_temporary to see if
> we ever encounter a COMPONENT_REF/ARRAY_REF around a TARGET_EXPR...

Or try to handle it properly even if we can't exercise the code yet.

I'm thinking again about warning for e.g.

const B& b = frotz(C{}.d);

where frotz passes through its reference argument, which becomes a 
dangling reference.  Handling the C{}.d case seems relevant to such a 
warning.

I wonder how many false positives there would be for warning about this 
for any function that both takes and returns a reference?

Do you want to try that, or shall I?

>> Incidentally, ref_conv_binds_directly should also probably be reversed to
>> ref_conv_binds_to_temporary since you can "bind directly" to an xvalue that
>> refers to a temporary.
> 
> Good point, I at least did that.
> 
> -- >8 --
> This PR reports that
> 
>    struct Base {};
>    struct Derived : Base {};
>    static_assert(__reference_constructs_from_temporary(Base const&, Derived));
> 
> doesn't pass, which it should: it's just like
> 
>    const Base& b(Derived{});
> 
> where we bind 'b' to the Base subobject of a temporary object of type
> Derived.  The ck_base conversion didn't have ->need_temporary_p set because
> we didn't need to create a temporary object just for the base, but the whole
> object is a temporary so we're still binding to a temporary.  Since the
> Base subobject is an xvalue, a new function is introduced.
> 
> 	PR c++/107085
> 
> gcc/cp/ChangeLog:
> 
> 	* call.cc (conv_binds_ref_to_temporary): New.
> 	(ref_conv_binds_directly): Rename to...
> 	(ref_conv_binds_to_temporary): ...this.  Use
> 	conv_binds_ref_to_temporary.
> 	* cp-tree.h (ref_conv_binds_directly): Rename to...
> 	(ref_conv_binds_to_temporary): ...this.
> 	* method.cc (ref_xes_from_temporary): Use ref_conv_binds_to_temporary.
> 	* parser.cc (warn_for_range_copy): Likewise.
> 
> gcc/testsuite/ChangeLog:
> 
> 	* g++.dg/ext/reference_constructs_from_temporary1.C: Adjust expected
> 	result.
> 	* g++.dg/ext/reference_converts_from_temporary1.C: Likewise.
> 	* g++.dg/cpp0x/elision4.C: New test.
> ---
>   gcc/cp/call.cc                                | 33 +++++++++++++++----
>   gcc/cp/cp-tree.h                              |  2 +-
>   gcc/cp/method.cc                              |  2 +-
>   gcc/cp/parser.cc                              |  5 +--
>   gcc/testsuite/g++.dg/cpp0x/elision4.C         | 15 +++++++++
>   .../reference_constructs_from_temporary1.C    |  2 +-
>   .../ext/reference_converts_from_temporary1.C  |  2 +-
>   7 files changed, 49 insertions(+), 12 deletions(-)
>   create mode 100644 gcc/testsuite/g++.dg/cpp0x/elision4.C
> 
> diff --git a/gcc/cp/call.cc b/gcc/cp/call.cc
> index bd04a1d309a..5a3d509f6c8 100644
> --- a/gcc/cp/call.cc
> +++ b/gcc/cp/call.cc
> @@ -9210,15 +9210,36 @@ conv_binds_ref_to_prvalue (conversion *c)
>     return conv_is_prvalue (next_conversion (c));
>   }
>   
> -/* Return tristate::TS_TRUE if converting EXPR to a reference type TYPE does
> -   not involve creating a temporary.  Return tristate::TS_FALSE if converting
> -   EXPR to a reference type TYPE binds the reference to a temporary.  If the
> -   conversion is invalid or bad, return tristate::TS_UNKNOWN.  DIRECT_INIT_P
> +/* True iff C is a conversion that binds a reference to a temporary.
> +   This is a superset of conv_binds_ref_to_prvalue: here we're also
> +   interested in xvalues.  */
> +
> +static bool
> +conv_binds_ref_to_temporary (conversion *c)
> +{
> +  if (conv_binds_ref_to_prvalue (c))
> +    return true;
> +  if (c->kind != ck_ref_bind)
> +    return false;
> +  c = next_conversion (c);
> +  /* This is the case for
> +       struct Base {};
> +       struct Derived : Base {};
> +       const Base& b(Derived{});
> +     where we bind 'b' to the Base subobject of a temporary object of type
> +     Derived.  The subobject is an xvalue; the whole object is a prvalue.  */
> +  return (c->kind == ck_base && conv_is_prvalue (next_conversion (c)));
> +}
> +
> +/* Return tristate::TS_TRUE if converting EXPR to a reference type TYPE binds
> +   the reference to a temporary.  Return tristate::TS_FALSE if converting
> +   EXPR to a reference type TYPE doesn't bind the reference to a temporary.  If
> +   the conversion is invalid or bad, return tristate::TS_UNKNOWN.  DIRECT_INIT_P
>      says whether the conversion should be done in direct- or copy-initialization
>      context.  */
>   
>   tristate
> -ref_conv_binds_directly (tree type, tree expr, bool direct_init_p /*= false*/)
> +ref_conv_binds_to_temporary (tree type, tree expr, bool direct_init_p/*=false*/)
>   {
>     gcc_assert (TYPE_REF_P (type));
>   
> @@ -9230,7 +9251,7 @@ ref_conv_binds_directly (tree type, tree expr, bool direct_init_p /*= false*/)
>   					  /*c_cast_p=*/false, flags, tf_none);
>     tristate ret (tristate::TS_UNKNOWN);
>     if (conv && !conv->bad_p)
> -    ret = tristate (!conv_binds_ref_to_prvalue (conv));
> +    ret = tristate (conv_binds_ref_to_temporary (conv));
>   
>     /* Free all the conversions we allocated.  */
>     obstack_free (&conversion_obstack, p);
> diff --git a/gcc/cp/cp-tree.h b/gcc/cp/cp-tree.h
> index 8bc1c2dc7fd..469eb2fdb25 100644
> --- a/gcc/cp/cp-tree.h
> +++ b/gcc/cp/cp-tree.h
> @@ -6534,7 +6534,7 @@ extern bool sufficient_parms_p			(const_tree);
>   extern tree type_decays_to			(tree);
>   extern tree extract_call_expr			(tree);
>   extern tree build_trivial_dtor_call		(tree, bool = false);
> -extern tristate ref_conv_binds_directly		(tree, tree, bool = false);
> +extern tristate ref_conv_binds_to_temporary	(tree, tree, bool = false);
>   extern tree build_user_type_conversion		(tree, tree, int,
>   						 tsubst_flags_t);
>   extern tree build_new_function_call		(tree, vec<tree, va_gc> **,
> diff --git a/gcc/cp/method.cc b/gcc/cp/method.cc
> index 55af5c43c18..622e1b9802e 100644
> --- a/gcc/cp/method.cc
> +++ b/gcc/cp/method.cc
> @@ -2233,7 +2233,7 @@ ref_xes_from_temporary (tree to, tree from, bool direct_init_p)
>     tree val = build_stub_object (from);
>     if (!TYPE_REF_P (from) && TREE_CODE (from) != FUNCTION_TYPE)
>       val = CLASS_TYPE_P (from) ? force_rvalue (val, tf_none) : rvalue (val);
> -  return ref_conv_binds_directly (to, val, direct_init_p).is_false ();
> +  return ref_conv_binds_to_temporary (to, val, direct_init_p).is_true ();
>   }
>   
>   /* Worker for is_{,nothrow_}convertible.  Attempt to perform an implicit
> diff --git a/gcc/cp/parser.cc b/gcc/cp/parser.cc
> index 90e06f914fb..bceedab5cd6 100644
> --- a/gcc/cp/parser.cc
> +++ b/gcc/cp/parser.cc
> @@ -13731,7 +13731,8 @@ warn_for_range_copy (tree decl, tree expr)
>   
>     if (TYPE_REF_P (type))
>       {
> -      if (glvalue_p (expr) && ref_conv_binds_directly (type, expr).is_false ())
> +      if (glvalue_p (expr)
> +	  && ref_conv_binds_to_temporary (type, expr).is_true ())
>   	{
>   	  auto_diagnostic_group d;
>   	  if (warning_at (loc, OPT_Wrange_loop_construct,
> @@ -13762,7 +13763,7 @@ warn_for_range_copy (tree decl, tree expr)
>     /* If we can initialize a reference directly, suggest that to avoid the
>        copy.  */
>     tree rtype = cp_build_reference_type (type, /*rval*/false);
> -  if (ref_conv_binds_directly (rtype, expr).is_true ())
> +  if (ref_conv_binds_to_temporary (rtype, expr).is_false ())
>       {
>         auto_diagnostic_group d;
>         if (warning_at (loc, OPT_Wrange_loop_construct,
> diff --git a/gcc/testsuite/g++.dg/cpp0x/elision4.C b/gcc/testsuite/g++.dg/cpp0x/elision4.C
> new file mode 100644
> index 00000000000..3cc2e3afa5d
> --- /dev/null
> +++ b/gcc/testsuite/g++.dg/cpp0x/elision4.C
> @@ -0,0 +1,15 @@
> +// PR c++/107085
> +// { dg-do compile { target c++11 } }
> +
> +struct X {
> +  X();
> +  X(X&&);
> +};
> +struct Z : X {};
> +X x1 = Z();
> +X x2 = X(Z());
> +
> +struct B { };
> +struct D : B { };
> +B b1 = D();
> +B b2 = B(D());
> diff --git a/gcc/testsuite/g++.dg/ext/reference_constructs_from_temporary1.C b/gcc/testsuite/g++.dg/ext/reference_constructs_from_temporary1.C
> index 76de905a35d..5354b1dc4e6 100644
> --- a/gcc/testsuite/g++.dg/ext/reference_constructs_from_temporary1.C
> +++ b/gcc/testsuite/g++.dg/ext/reference_constructs_from_temporary1.C
> @@ -201,7 +201,7 @@ SA(!__reference_constructs_from_temporary(const int&, H));
>   SA(!__reference_constructs_from_temporary(int&&, G2));
>   SA(!__reference_constructs_from_temporary(const int&, H2));
>   
> -SA(!__reference_constructs_from_temporary(const Base&, Der));
> +SA(__reference_constructs_from_temporary(const Base&, Der));
>   
>   // This fails because std::is_constructible_v<int&&, id<int[3]>> is false.
>   SA(!__reference_constructs_from_temporary(int&&, id<int[3]>));
> diff --git a/gcc/testsuite/g++.dg/ext/reference_converts_from_temporary1.C b/gcc/testsuite/g++.dg/ext/reference_converts_from_temporary1.C
> index 90196c38742..e6c159e9b00 100644
> --- a/gcc/testsuite/g++.dg/ext/reference_converts_from_temporary1.C
> +++ b/gcc/testsuite/g++.dg/ext/reference_converts_from_temporary1.C
> @@ -201,7 +201,7 @@ SA( __reference_converts_from_temporary(const int&, H));
>   SA(!__reference_converts_from_temporary(int&&, G2));
>   SA(!__reference_converts_from_temporary(const int&, H2));
>   
> -SA(!__reference_converts_from_temporary(const Base&, Der));
> +SA(__reference_converts_from_temporary(const Base&, Der));
>   
>   // This fails because std::is_constructible_v<int&&, id<int[3]>> is false.
>   SA(!__reference_converts_from_temporary(int&&, id<int[3]>));
> 
> base-commit: 50c35c691517291dbb77b1661761bc59950ba101