C++ PATCH to fix rejects-valid with constexpr ctor in C++17 (PR c++/83692)

C++ PATCH to fix rejects-valid with constexpr ctor in C++17 (PR c++/83692)

[Marek Polacek]

This is a similar problem to 83116: we'd cached a constexpr call, but after a
store the result had become invalid, yet we used the wrong result again when
encountering the same call later.  This resulted in evaluating a THROW_EXPR
which doesn't work.  Details in
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=83692#c5

The fix for 83116 didn't work here, because when evaluating the body of the
ctor via store_init_value -> cxx_constant_value we are in STRICT, so we do
cache.

It seems that we may no longer rely on the constexpr call table when we
do cxx_eval_store_expression, because that just rewrites *valp, i.e. the
value of an object.  Might be too big a hammer again, but I couldn't think
of how I could guard the caching of a constexpr call.

This doesn't manifest in C++14 because build_special_member_call in C++17 is
more aggressive with copy elisions (as required by P0135 which changed how we
view prvalues).  In C++14 build_special_member_call produces a CALL_EXPR, so
expand_default_init calls maybe_constant_init, for which STRICT is false, so
we avoid caching as per 83116.

Bootstrapped/regtested on x86_64-linux, ok for trunk?

2018-01-25  Marek Polacek  <polacek@redhat.com>

	PR c++/83692
	* constexpr.c (cxx_eval_store_expression): Clear constexpr_call_table.

	* g++.dg/cpp1y/constexpr-83692.C: New test.

diff --git gcc/cp/constexpr.c gcc/cp/constexpr.c
index 4d2ee4a28fc..0202d22f320 100644
--- gcc/cp/constexpr.c
+++ gcc/cp/constexpr.c
@@ -3663,6 +3663,10 @@ cxx_eval_store_expression (const constexpr_ctx *ctx, tree t,
   else
     *valp = init;
 
+  /* We've rewritten a value of a temporary in this constexpr
+     context which might invalide a cached call.  */
+  constexpr_call_table = NULL;
+
   /* Update TREE_CONSTANT and TREE_SIDE_EFFECTS on enclosing
      CONSTRUCTORs, if any.  */
   tree elt;
diff --git gcc/testsuite/g++.dg/cpp1y/constexpr-83692.C gcc/testsuite/g++.dg/cpp1y/constexpr-83692.C
index e69de29bb2d..292ba7c22e9 100644
--- gcc/testsuite/g++.dg/cpp1y/constexpr-83692.C
+++ gcc/testsuite/g++.dg/cpp1y/constexpr-83692.C
@@ -0,0 +1,21 @@
+// PR c++/83692
+// { dg-do compile { target c++14 } }
+
+struct integer {
+  constexpr int value() const { return m_value;	}
+  int m_value;
+};
+
+struct outer {
+  integer m_x{0};
+  constexpr outer()
+    {
+      if (m_x.value() != 0)
+	throw 0;
+      m_x.m_value = integer{1}.value();
+      if (m_x.value() != 1)
+	throw 0;
+    }
+};
+
+constexpr outer o{};

	Marek

Re: C++ PATCH to fix rejects-valid with constexpr ctor in C++17 (PR c++/83692)

[Marek Polacek]

On Thu, Jan 25, 2018 at 10:16:39PM +0100, Marek Polacek wrote:
> This is a similar problem to 83116: we'd cached a constexpr call, but after a
> store the result had become invalid, yet we used the wrong result again when
> encountering the same call later.  This resulted in evaluating a THROW_EXPR
> which doesn't work.  Details in
> https://gcc.gnu.org/bugzilla/show_bug.cgi?id=83692#c5
> 
> The fix for 83116 didn't work here, because when evaluating the body of the
> ctor via store_init_value -> cxx_constant_value we are in STRICT, so we do
> cache.
> 
> It seems that we may no longer rely on the constexpr call table when we
> do cxx_eval_store_expression, because that just rewrites *valp, i.e. the
> value of an object.  Might be too big a hammer again, but I couldn't think
> of how I could guard the caching of a constexpr call.
> 
> This doesn't manifest in C++14 because build_special_member_call in C++17 is
> more aggressive with copy elisions (as required by P0135 which changed how we
> view prvalues).  In C++14 build_special_member_call produces a CALL_EXPR, so
> expand_default_init calls maybe_constant_init, for which STRICT is false, so
> we avoid caching as per 83116.

...so the testcase should actually test c++17.  Consider that fixed.

	Marek

Re: C++ PATCH to fix rejects-valid with constexpr ctor in C++17 (PR c++/83692)

[Jason Merrill]

On Thu, Jan 25, 2018 at 4:16 PM, Marek Polacek <polacek@redhat.com> wrote:
> This is a similar problem to 83116: we'd cached a constexpr call, but after a
> store the result had become invalid, yet we used the wrong result again when
> encountering the same call later.  This resulted in evaluating a THROW_EXPR
> which doesn't work.  Details in
> https://gcc.gnu.org/bugzilla/show_bug.cgi?id=83692#c5
>
> The fix for 83116 didn't work here, because when evaluating the body of the
> ctor via store_init_value -> cxx_constant_value we are in STRICT, so we do
> cache.

> It seems that we may no longer rely on the constexpr call table when we
> do cxx_eval_store_expression, because that just rewrites *valp, i.e. the
> value of an object.  Might be too big a hammer again, but I couldn't think
> of how I could guard the caching of a constexpr call.

> This doesn't manifest in C++14 because build_special_member_call in C++17 is
> more aggressive with copy elisions (as required by P0135 which changed how we
> view prvalues).  In C++14 build_special_member_call produces a CALL_EXPR, so
> expand_default_init calls maybe_constant_init, for which STRICT is false, so
> we avoid caching as per 83116.

So it sounds like the problem is using cxx_constant_value for the
diagnostic when it has different semantics from the
maybe_constant_init that follows right after.  I guess we want a
cxx_constant_init function that is a hybrid of the two.

Jason

Re: C++ PATCH to fix rejects-valid with constexpr ctor in C++17 (PR c++/83692)

[Marek Polacek]

On Fri, Feb 02, 2018 at 02:11:27PM -0500, Jason Merrill wrote:
> On Thu, Jan 25, 2018 at 4:16 PM, Marek Polacek <polacek@redhat.com> wrote:
> > This is a similar problem to 83116: we'd cached a constexpr call, but after a
> > store the result had become invalid, yet we used the wrong result again when
> > encountering the same call later.  This resulted in evaluating a THROW_EXPR
> > which doesn't work.  Details in
> > https://gcc.gnu.org/bugzilla/show_bug.cgi?id=83692#c5
> >
> > The fix for 83116 didn't work here, because when evaluating the body of the
> > ctor via store_init_value -> cxx_constant_value we are in STRICT, so we do
> > cache.
> 
> > It seems that we may no longer rely on the constexpr call table when we
> > do cxx_eval_store_expression, because that just rewrites *valp, i.e. the
> > value of an object.  Might be too big a hammer again, but I couldn't think
> > of how I could guard the caching of a constexpr call.
> 
> > This doesn't manifest in C++14 because build_special_member_call in C++17 is
> > more aggressive with copy elisions (as required by P0135 which changed how we
> > view prvalues).  In C++14 build_special_member_call produces a CALL_EXPR, so
> > expand_default_init calls maybe_constant_init, for which STRICT is false, so
> > we avoid caching as per 83116.
> 
> So it sounds like the problem is using cxx_constant_value for the
> diagnostic when it has different semantics from the
> maybe_constant_init that follows right after.  I guess we want a
> cxx_constant_init function that is a hybrid of the two.

So like the following?  Thanks,

Bootstrapped/regtested on x86_64-linux, ok for trunk?

2018-02-04  Marek Polacek  <polacek@redhat.com>

	PR c++/83692
	* constexpr.c (cxx_constant_init): New function.
	* cp-tree.h (cxx_constant_init): Declare.
	* typeck2.c (store_init_value): Call cxx_constant_init instead of
	cxx_constant_value.

	* g++.dg/cpp1z/constexpr-83692.C: New test.

diff --git gcc/cp/constexpr.c gcc/cp/constexpr.c
index 93dd8ae049c..f95aacf2580 100644
--- gcc/cp/constexpr.c
+++ gcc/cp/constexpr.c
@@ -4902,6 +4902,14 @@ cxx_constant_value (tree t, tree decl)
   return cxx_eval_outermost_constant_expr (t, false, true, decl);
 }
 
+/* Like cxx_constant_value, but non-strict mode.  */
+
+tree
+cxx_constant_init (tree t, tree decl)
+{
+  return cxx_eval_outermost_constant_expr (t, false, false, decl);
+}
+
 /* Helper routine for fold_simple function.  Either return simplified
    expression T, otherwise NULL_TREE.
    In contrast to cp_fully_fold, and to maybe_constant_value, we try to fold
diff --git gcc/cp/cp-tree.h gcc/cp/cp-tree.h
index a53f4fd9c03..9f973305fbb 100644
--- gcc/cp/cp-tree.h
+++ gcc/cp/cp-tree.h
@@ -7417,6 +7417,7 @@ extern bool require_potential_constant_expression (tree);
 extern bool require_constant_expression (tree);
 extern bool require_potential_rvalue_constant_expression (tree);
 extern tree cxx_constant_value			(tree, tree = NULL_TREE);
+extern tree cxx_constant_init			(tree, tree = NULL_TREE);
 extern tree maybe_constant_value		(tree, tree = NULL_TREE);
 extern tree maybe_constant_init			(tree, tree = NULL_TREE);
 extern tree fold_non_dependent_expr		(tree);
diff --git gcc/cp/typeck2.c gcc/cp/typeck2.c
index 899d60e8535..b4abc54f537 100644
--- gcc/cp/typeck2.c
+++ gcc/cp/typeck2.c
@@ -830,7 +830,7 @@ store_init_value (tree decl, tree init, vec<tree, va_gc>** cleanups, int flags)
 	  if (!require_constant_expression (value))
 	    value = error_mark_node;
 	  else
-	    value = cxx_constant_value (value, decl);
+	    value = cxx_constant_init (value, decl);
 	}
       value = maybe_constant_init (value, decl);
       if (TREE_CODE (value) == CONSTRUCTOR && cp_has_mutable_p (type))
diff --git gcc/testsuite/g++.dg/cpp1z/constexpr-83692.C gcc/testsuite/g++.dg/cpp1z/constexpr-83692.C
index e69de29bb2d..f6b61eeab85 100644
--- gcc/testsuite/g++.dg/cpp1z/constexpr-83692.C
+++ gcc/testsuite/g++.dg/cpp1z/constexpr-83692.C
@@ -0,0 +1,21 @@
+// PR c++/83692
+// { dg-options -std=c++17 }
+
+struct integer {
+  constexpr int value() const { return m_value;	}
+  int m_value;
+};
+
+struct outer {
+  integer m_x{0};
+  constexpr outer()
+    {
+      if (m_x.value() != 0)
+	throw 0;
+      m_x.m_value = integer{1}.value();
+      if (m_x.value() != 1)
+	throw 0;
+    }
+};
+
+constexpr outer o{};

	Marek

Re: C++ PATCH to fix rejects-valid with constexpr ctor in C++17 (PR c++/83692)

[Jason Merrill]

On Mon, Feb 5, 2018 at 8:37 AM, Marek Polacek <polacek@redhat.com> wrote:
> On Fri, Feb 02, 2018 at 02:11:27PM -0500, Jason Merrill wrote:
>> On Thu, Jan 25, 2018 at 4:16 PM, Marek Polacek <polacek@redhat.com> wrote:
>> > This is a similar problem to 83116: we'd cached a constexpr call, but after a
>> > store the result had become invalid, yet we used the wrong result again when
>> > encountering the same call later.  This resulted in evaluating a THROW_EXPR
>> > which doesn't work.  Details in
>> > https://gcc.gnu.org/bugzilla/show_bug.cgi?id=83692#c5
>> >
>> > The fix for 83116 didn't work here, because when evaluating the body of the
>> > ctor via store_init_value -> cxx_constant_value we are in STRICT, so we do
>> > cache.
>>
>> > It seems that we may no longer rely on the constexpr call table when we
>> > do cxx_eval_store_expression, because that just rewrites *valp, i.e. the
>> > value of an object.  Might be too big a hammer again, but I couldn't think
>> > of how I could guard the caching of a constexpr call.
>>
>> > This doesn't manifest in C++14 because build_special_member_call in C++17 is
>> > more aggressive with copy elisions (as required by P0135 which changed how we
>> > view prvalues).  In C++14 build_special_member_call produces a CALL_EXPR, so
>> > expand_default_init calls maybe_constant_init, for which STRICT is false, so
>> > we avoid caching as per 83116.
>>
>> So it sounds like the problem is using cxx_constant_value for the
>> diagnostic when it has different semantics from the
>> maybe_constant_init that follows right after.  I guess we want a
>> cxx_constant_init function that is a hybrid of the two.
>
> So like the following?  Thanks,
>
> Bootstrapped/regtested on x86_64-linux, ok for trunk?
>
> 2018-02-04  Marek Polacek  <polacek@redhat.com>
>
>         PR c++/83692
>         * constexpr.c (cxx_constant_init): New function.
>         * cp-tree.h (cxx_constant_init): Declare.
>         * typeck2.c (store_init_value): Call cxx_constant_init instead of
>         cxx_constant_value.
>
> +/* Like cxx_constant_value, but non-strict mode.  */
> +
> +tree
> +cxx_constant_init (tree t, tree decl)
> +{
> +  return cxx_eval_outermost_constant_expr (t, false, false, decl);
> +}

Hmm, that doesn't do the TARGET_EXPR stripping that
maybe_constant_init does.  I was thinking of a version of
maybe_constant_init that passes false to allow_non_constant.  Probably
by making "maybe_constant_init" and cxx_constant_init both call the
current function with an additional parameter.  And then the existing
call to maybe_constant_init can move under an 'else' to avoid
redundant constexpr evaluation.

Jason

Re: C++ PATCH to fix rejects-valid with constexpr ctor in C++17 (PR c++/83692)

[Jason Merrill]

On Mon, Feb 5, 2018 at 1:45 PM, Jason Merrill <jason@redhat.com> wrote:
> On Mon, Feb 5, 2018 at 8:37 AM, Marek Polacek <polacek@redhat.com> wrote:
>> On Fri, Feb 02, 2018 at 02:11:27PM -0500, Jason Merrill wrote:
>>> On Thu, Jan 25, 2018 at 4:16 PM, Marek Polacek <polacek@redhat.com> wrote:
>>> > This is a similar problem to 83116: we'd cached a constexpr call, but after a
>>> > store the result had become invalid, yet we used the wrong result again when
>>> > encountering the same call later.  This resulted in evaluating a THROW_EXPR
>>> > which doesn't work.  Details in
>>> > https://gcc.gnu.org/bugzilla/show_bug.cgi?id=83692#c5
>>> >
>>> > The fix for 83116 didn't work here, because when evaluating the body of the
>>> > ctor via store_init_value -> cxx_constant_value we are in STRICT, so we do
>>> > cache.
>>>
>>> > It seems that we may no longer rely on the constexpr call table when we
>>> > do cxx_eval_store_expression, because that just rewrites *valp, i.e. the
>>> > value of an object.  Might be too big a hammer again, but I couldn't think
>>> > of how I could guard the caching of a constexpr call.
>>>
>>> > This doesn't manifest in C++14 because build_special_member_call in C++17 is
>>> > more aggressive with copy elisions (as required by P0135 which changed how we
>>> > view prvalues).  In C++14 build_special_member_call produces a CALL_EXPR, so
>>> > expand_default_init calls maybe_constant_init, for which STRICT is false, so
>>> > we avoid caching as per 83116.
>>>
>>> So it sounds like the problem is using cxx_constant_value for the
>>> diagnostic when it has different semantics from the
>>> maybe_constant_init that follows right after.  I guess we want a
>>> cxx_constant_init function that is a hybrid of the two.
>>
>> So like the following?  Thanks,
>>
>> Bootstrapped/regtested on x86_64-linux, ok for trunk?
>>
>> 2018-02-04  Marek Polacek  <polacek@redhat.com>
>>
>>         PR c++/83692
>>         * constexpr.c (cxx_constant_init): New function.
>>         * cp-tree.h (cxx_constant_init): Declare.
>>         * typeck2.c (store_init_value): Call cxx_constant_init instead of
>>         cxx_constant_value.
>>
>> +/* Like cxx_constant_value, but non-strict mode.  */
>> +
>> +tree
>> +cxx_constant_init (tree t, tree decl)
>> +{
>> +  return cxx_eval_outermost_constant_expr (t, false, false, decl);
>> +}
>
> Hmm, that doesn't do the TARGET_EXPR stripping that
> maybe_constant_init does.  I was thinking of a version of
> maybe_constant_init that passes false to allow_non_constant.  Probably
> by making "maybe_constant_init" and cxx_constant_init both call the
> current function with an additional parameter.  And then the existing
> call to maybe_constant_init can move under an 'else' to avoid
> redundant constexpr evaluation.

Want me to take this over?

Jason

Re: C++ PATCH to fix rejects-valid with constexpr ctor in C++17 (PR c++/83692)

[Marek Polacek]

On Fri, Feb 16, 2018 at 04:10:20PM -0500, Jason Merrill wrote:
> On Mon, Feb 5, 2018 at 1:45 PM, Jason Merrill <jason@redhat.com> wrote:
> > On Mon, Feb 5, 2018 at 8:37 AM, Marek Polacek <polacek@redhat.com> wrote:
> >> On Fri, Feb 02, 2018 at 02:11:27PM -0500, Jason Merrill wrote:
> >>> On Thu, Jan 25, 2018 at 4:16 PM, Marek Polacek <polacek@redhat.com> wrote:
> >>> > This is a similar problem to 83116: we'd cached a constexpr call, but after a
> >>> > store the result had become invalid, yet we used the wrong result again when
> >>> > encountering the same call later.  This resulted in evaluating a THROW_EXPR
> >>> > which doesn't work.  Details in
> >>> > https://gcc.gnu.org/bugzilla/show_bug.cgi?id=83692#c5
> >>> >
> >>> > The fix for 83116 didn't work here, because when evaluating the body of the
> >>> > ctor via store_init_value -> cxx_constant_value we are in STRICT, so we do
> >>> > cache.
> >>>
> >>> > It seems that we may no longer rely on the constexpr call table when we
> >>> > do cxx_eval_store_expression, because that just rewrites *valp, i.e. the
> >>> > value of an object.  Might be too big a hammer again, but I couldn't think
> >>> > of how I could guard the caching of a constexpr call.
> >>>
> >>> > This doesn't manifest in C++14 because build_special_member_call in C++17 is
> >>> > more aggressive with copy elisions (as required by P0135 which changed how we
> >>> > view prvalues).  In C++14 build_special_member_call produces a CALL_EXPR, so
> >>> > expand_default_init calls maybe_constant_init, for which STRICT is false, so
> >>> > we avoid caching as per 83116.
> >>>
> >>> So it sounds like the problem is using cxx_constant_value for the
> >>> diagnostic when it has different semantics from the
> >>> maybe_constant_init that follows right after.  I guess we want a
> >>> cxx_constant_init function that is a hybrid of the two.
> >>
> >> So like the following?  Thanks,
> >>
> >> Bootstrapped/regtested on x86_64-linux, ok for trunk?
> >>
> >> 2018-02-04  Marek Polacek  <polacek@redhat.com>
> >>
> >>         PR c++/83692
> >>         * constexpr.c (cxx_constant_init): New function.
> >>         * cp-tree.h (cxx_constant_init): Declare.
> >>         * typeck2.c (store_init_value): Call cxx_constant_init instead of
> >>         cxx_constant_value.
> >>
> >> +/* Like cxx_constant_value, but non-strict mode.  */
> >> +
> >> +tree
> >> +cxx_constant_init (tree t, tree decl)
> >> +{
> >> +  return cxx_eval_outermost_constant_expr (t, false, false, decl);
> >> +}
> >
> > Hmm, that doesn't do the TARGET_EXPR stripping that
> > maybe_constant_init does.  I was thinking of a version of
> > maybe_constant_init that passes false to allow_non_constant.  Probably
> > by making "maybe_constant_init" and cxx_constant_init both call the
> > current function with an additional parameter.  And then the existing
> > call to maybe_constant_init can move under an 'else' to avoid
> > redundant constexpr evaluation.
> 
> Want me to take this over?

Sorry again for the delay.

I tried to do what you suggested.  There was one twist: it regressed
constexpr-nullptr-2.C, in particular we lost diagnostics for

constexpr int* pj0 = &((S*)0)->j;       // { dg-error "not a constant expression" }
constexpr int* pj1 = &((S*)nullptr)->j;  // { dg-error "not a constant expression" }

because when maybe_constant_init_1 saw a constant:

5142   else if (CONSTANT_CLASS_P (t))
5143     /* No evaluation needed.  */;

so it didn't call cxx_eval_outermost_constant_expr which is supposed to give
the error.  I fixed it by adding "&& allow_non_constant" so now it gives the
proper diagnostics.

Bootstrapped/regtested on x86_64-linux, ok for trunk?

2018-02-23  Marek Polacek  <polacek@redhat.com>

	PR c++/83692
	* constexpr.c (maybe_constant_init_1): New function.
	(maybe_constant_init): Make it a wrapper around maybe_constant_init_1.
	(cxx_constant_init): New function.
	* cp-tree.h (cxx_constant_init): Declare.
	* typeck2.c (store_init_value): Call cxx_constant_init instead of
	cxx_constant_value.  Move the maybe_constant_init call under an 'else'.

	* g++.dg/cpp1z/constexpr-83692.C: New test.

diff --git gcc/cp/constexpr.c gcc/cp/constexpr.c
index 47ff90cb055..26d0d099a05 100644
--- gcc/cp/constexpr.c
+++ gcc/cp/constexpr.c
@@ -5123,8 +5123,8 @@ fold_non_dependent_expr (tree t)
 /* Like maybe_constant_value, but returns a CONSTRUCTOR directly, rather
    than wrapped in a TARGET_EXPR.  */
 
-tree
-maybe_constant_init (tree t, tree decl)
+static tree
+maybe_constant_init_1 (tree t, tree decl, bool allow_non_constant)
 {
   if (!t)
     return t;
@@ -5139,10 +5139,10 @@ maybe_constant_init (tree t, tree decl)
     t = TARGET_EXPR_INITIAL (t);
   if (!is_nondependent_static_init_expression (t))
     /* Don't try to evaluate it.  */;
-  else if (CONSTANT_CLASS_P (t))
+  else if (CONSTANT_CLASS_P (t) && allow_non_constant)
     /* No evaluation needed.  */;
   else
-    t = cxx_eval_outermost_constant_expr (t, true, false, decl);
+    t = cxx_eval_outermost_constant_expr (t, allow_non_constant, false, decl);
   if (TREE_CODE (t) == TARGET_EXPR)
     {
       tree init = TARGET_EXPR_INITIAL (t);
@@ -5152,6 +5152,22 @@ maybe_constant_init (tree t, tree decl)
   return t;
 }
 
+/* Wrapper for maybe_constant_init_1 which permits non constants.  */
+
+tree
+maybe_constant_init (tree t, tree decl)
+{
+  return maybe_constant_init_1 (t, decl, true);
+}
+
+/* Wrapper for maybe_constant_init_1 which does not permit non constants.  */
+
+tree
+cxx_constant_init (tree t, tree decl)
+{
+  return maybe_constant_init_1 (t, decl, false);
+}
+
 #if 0
 /* FIXME see ADDR_EXPR section in potential_constant_expression_1.  */
 /* Return true if the object referred to by REF has automatic or thread
diff --git gcc/cp/cp-tree.h gcc/cp/cp-tree.h
index 9038d677b2d..04c7b7ce3a9 100644
--- gcc/cp/cp-tree.h
+++ gcc/cp/cp-tree.h
@@ -7411,6 +7411,7 @@ extern bool require_potential_constant_expression (tree);
 extern bool require_constant_expression (tree);
 extern bool require_potential_rvalue_constant_expression (tree);
 extern tree cxx_constant_value			(tree, tree = NULL_TREE);
+extern tree cxx_constant_init			(tree, tree = NULL_TREE);
 extern tree maybe_constant_value		(tree, tree = NULL_TREE);
 extern tree maybe_constant_init			(tree, tree = NULL_TREE);
 extern tree fold_non_dependent_expr		(tree);
diff --git gcc/cp/typeck2.c gcc/cp/typeck2.c
index 899d60e8535..153b46cca77 100644
--- gcc/cp/typeck2.c
+++ gcc/cp/typeck2.c
@@ -830,9 +830,10 @@ store_init_value (tree decl, tree init, vec<tree, va_gc>** cleanups, int flags)
 	  if (!require_constant_expression (value))
 	    value = error_mark_node;
 	  else
-	    value = cxx_constant_value (value, decl);
+	    value = cxx_constant_init (value, decl);
 	}
-      value = maybe_constant_init (value, decl);
+      else
+	value = maybe_constant_init (value, decl);
       if (TREE_CODE (value) == CONSTRUCTOR && cp_has_mutable_p (type))
 	/* Poison this CONSTRUCTOR so it can't be copied to another
 	   constexpr variable.  */
diff --git gcc/testsuite/g++.dg/cpp1z/constexpr-83692.C gcc/testsuite/g++.dg/cpp1z/constexpr-83692.C
index e69de29bb2d..f6b61eeab85 100644
--- gcc/testsuite/g++.dg/cpp1z/constexpr-83692.C
+++ gcc/testsuite/g++.dg/cpp1z/constexpr-83692.C
@@ -0,0 +1,21 @@
+// PR c++/83692
+// { dg-options -std=c++17 }
+
+struct integer {
+  constexpr int value() const { return m_value;	}
+  int m_value;
+};
+
+struct outer {
+  integer m_x{0};
+  constexpr outer()
+    {
+      if (m_x.value() != 0)
+	throw 0;
+      m_x.m_value = integer{1}.value();
+      if (m_x.value() != 1)
+	throw 0;
+    }
+};
+
+constexpr outer o{};

	Marek

Re: C++ PATCH to fix rejects-valid with constexpr ctor in C++17 (PR c++/83692)

[Jason Merrill]

OK, thanks.

On Fri, Feb 23, 2018 at 9:29 AM, Marek Polacek <polacek@redhat.com> wrote:
> On Fri, Feb 16, 2018 at 04:10:20PM -0500, Jason Merrill wrote:
>> On Mon, Feb 5, 2018 at 1:45 PM, Jason Merrill <jason@redhat.com> wrote:
>> > On Mon, Feb 5, 2018 at 8:37 AM, Marek Polacek <polacek@redhat.com> wrote:
>> >> On Fri, Feb 02, 2018 at 02:11:27PM -0500, Jason Merrill wrote:
>> >>> On Thu, Jan 25, 2018 at 4:16 PM, Marek Polacek <polacek@redhat.com> wrote:
>> >>> > This is a similar problem to 83116: we'd cached a constexpr call, but after a
>> >>> > store the result had become invalid, yet we used the wrong result again when
>> >>> > encountering the same call later.  This resulted in evaluating a THROW_EXPR
>> >>> > which doesn't work.  Details in
>> >>> > https://gcc.gnu.org/bugzilla/show_bug.cgi?id=83692#c5
>> >>> >
>> >>> > The fix for 83116 didn't work here, because when evaluating the body of the
>> >>> > ctor via store_init_value -> cxx_constant_value we are in STRICT, so we do
>> >>> > cache.
>> >>>
>> >>> > It seems that we may no longer rely on the constexpr call table when we
>> >>> > do cxx_eval_store_expression, because that just rewrites *valp, i.e. the
>> >>> > value of an object.  Might be too big a hammer again, but I couldn't think
>> >>> > of how I could guard the caching of a constexpr call.
>> >>>
>> >>> > This doesn't manifest in C++14 because build_special_member_call in C++17 is
>> >>> > more aggressive with copy elisions (as required by P0135 which changed how we
>> >>> > view prvalues).  In C++14 build_special_member_call produces a CALL_EXPR, so
>> >>> > expand_default_init calls maybe_constant_init, for which STRICT is false, so
>> >>> > we avoid caching as per 83116.
>> >>>
>> >>> So it sounds like the problem is using cxx_constant_value for the
>> >>> diagnostic when it has different semantics from the
>> >>> maybe_constant_init that follows right after.  I guess we want a
>> >>> cxx_constant_init function that is a hybrid of the two.
>> >>
>> >> So like the following?  Thanks,
>> >>
>> >> Bootstrapped/regtested on x86_64-linux, ok for trunk?
>> >>
>> >> 2018-02-04  Marek Polacek  <polacek@redhat.com>
>> >>
>> >>         PR c++/83692
>> >>         * constexpr.c (cxx_constant_init): New function.
>> >>         * cp-tree.h (cxx_constant_init): Declare.
>> >>         * typeck2.c (store_init_value): Call cxx_constant_init instead of
>> >>         cxx_constant_value.
>> >>
>> >> +/* Like cxx_constant_value, but non-strict mode.  */
>> >> +
>> >> +tree
>> >> +cxx_constant_init (tree t, tree decl)
>> >> +{
>> >> +  return cxx_eval_outermost_constant_expr (t, false, false, decl);
>> >> +}
>> >
>> > Hmm, that doesn't do the TARGET_EXPR stripping that
>> > maybe_constant_init does.  I was thinking of a version of
>> > maybe_constant_init that passes false to allow_non_constant.  Probably
>> > by making "maybe_constant_init" and cxx_constant_init both call the
>> > current function with an additional parameter.  And then the existing
>> > call to maybe_constant_init can move under an 'else' to avoid
>> > redundant constexpr evaluation.
>>
>> Want me to take this over?
>
> Sorry again for the delay.
>
> I tried to do what you suggested.  There was one twist: it regressed
> constexpr-nullptr-2.C, in particular we lost diagnostics for
>
> constexpr int* pj0 = &((S*)0)->j;       // { dg-error "not a constant expression" }
> constexpr int* pj1 = &((S*)nullptr)->j;  // { dg-error "not a constant expression" }
>
> because when maybe_constant_init_1 saw a constant:
>
> 5142   else if (CONSTANT_CLASS_P (t))
> 5143     /* No evaluation needed.  */;
>
> so it didn't call cxx_eval_outermost_constant_expr which is supposed to give
> the error.  I fixed it by adding "&& allow_non_constant" so now it gives the
> proper diagnostics.
>
> Bootstrapped/regtested on x86_64-linux, ok for trunk?
>
> 2018-02-23  Marek Polacek  <polacek@redhat.com>
>
>         PR c++/83692
>         * constexpr.c (maybe_constant_init_1): New function.
>         (maybe_constant_init): Make it a wrapper around maybe_constant_init_1.
>         (cxx_constant_init): New function.
>         * cp-tree.h (cxx_constant_init): Declare.
>         * typeck2.c (store_init_value): Call cxx_constant_init instead of
>         cxx_constant_value.  Move the maybe_constant_init call under an 'else'.
>
>         * g++.dg/cpp1z/constexpr-83692.C: New test.
>
> diff --git gcc/cp/constexpr.c gcc/cp/constexpr.c
> index 47ff90cb055..26d0d099a05 100644
> --- gcc/cp/constexpr.c
> +++ gcc/cp/constexpr.c
> @@ -5123,8 +5123,8 @@ fold_non_dependent_expr (tree t)
>  /* Like maybe_constant_value, but returns a CONSTRUCTOR directly, rather
>     than wrapped in a TARGET_EXPR.  */
>
> -tree
> -maybe_constant_init (tree t, tree decl)
> +static tree
> +maybe_constant_init_1 (tree t, tree decl, bool allow_non_constant)
>  {
>    if (!t)
>      return t;
> @@ -5139,10 +5139,10 @@ maybe_constant_init (tree t, tree decl)
>      t = TARGET_EXPR_INITIAL (t);
>    if (!is_nondependent_static_init_expression (t))
>      /* Don't try to evaluate it.  */;
> -  else if (CONSTANT_CLASS_P (t))
> +  else if (CONSTANT_CLASS_P (t) && allow_non_constant)
>      /* No evaluation needed.  */;
>    else
> -    t = cxx_eval_outermost_constant_expr (t, true, false, decl);
> +    t = cxx_eval_outermost_constant_expr (t, allow_non_constant, false, decl);
>    if (TREE_CODE (t) == TARGET_EXPR)
>      {
>        tree init = TARGET_EXPR_INITIAL (t);
> @@ -5152,6 +5152,22 @@ maybe_constant_init (tree t, tree decl)
>    return t;
>  }
>
> +/* Wrapper for maybe_constant_init_1 which permits non constants.  */
> +
> +tree
> +maybe_constant_init (tree t, tree decl)
> +{
> +  return maybe_constant_init_1 (t, decl, true);
> +}
> +
> +/* Wrapper for maybe_constant_init_1 which does not permit non constants.  */
> +
> +tree
> +cxx_constant_init (tree t, tree decl)
> +{
> +  return maybe_constant_init_1 (t, decl, false);
> +}
> +
>  #if 0
>  /* FIXME see ADDR_EXPR section in potential_constant_expression_1.  */
>  /* Return true if the object referred to by REF has automatic or thread
> diff --git gcc/cp/cp-tree.h gcc/cp/cp-tree.h
> index 9038d677b2d..04c7b7ce3a9 100644
> --- gcc/cp/cp-tree.h
> +++ gcc/cp/cp-tree.h
> @@ -7411,6 +7411,7 @@ extern bool require_potential_constant_expression (tree);
>  extern bool require_constant_expression (tree);
>  extern bool require_potential_rvalue_constant_expression (tree);
>  extern tree cxx_constant_value                 (tree, tree = NULL_TREE);
> +extern tree cxx_constant_init                  (tree, tree = NULL_TREE);
>  extern tree maybe_constant_value               (tree, tree = NULL_TREE);
>  extern tree maybe_constant_init                        (tree, tree = NULL_TREE);
>  extern tree fold_non_dependent_expr            (tree);
> diff --git gcc/cp/typeck2.c gcc/cp/typeck2.c
> index 899d60e8535..153b46cca77 100644
> --- gcc/cp/typeck2.c
> +++ gcc/cp/typeck2.c
> @@ -830,9 +830,10 @@ store_init_value (tree decl, tree init, vec<tree, va_gc>** cleanups, int flags)
>           if (!require_constant_expression (value))
>             value = error_mark_node;
>           else
> -           value = cxx_constant_value (value, decl);
> +           value = cxx_constant_init (value, decl);
>         }
> -      value = maybe_constant_init (value, decl);
> +      else
> +       value = maybe_constant_init (value, decl);
>        if (TREE_CODE (value) == CONSTRUCTOR && cp_has_mutable_p (type))
>         /* Poison this CONSTRUCTOR so it can't be copied to another
>            constexpr variable.  */
> diff --git gcc/testsuite/g++.dg/cpp1z/constexpr-83692.C gcc/testsuite/g++.dg/cpp1z/constexpr-83692.C
> index e69de29bb2d..f6b61eeab85 100644
> --- gcc/testsuite/g++.dg/cpp1z/constexpr-83692.C
> +++ gcc/testsuite/g++.dg/cpp1z/constexpr-83692.C
> @@ -0,0 +1,21 @@
> +// PR c++/83692
> +// { dg-options -std=c++17 }
> +
> +struct integer {
> +  constexpr int value() const { return m_value;        }
> +  int m_value;
> +};
> +
> +struct outer {
> +  integer m_x{0};
> +  constexpr outer()
> +    {
> +      if (m_x.value() != 0)
> +       throw 0;
> +      m_x.m_value = integer{1}.value();
> +      if (m_x.value() != 1)
> +       throw 0;
> +    }
> +};
> +
> +constexpr outer o{};
>
>         Marek