c++: excessive satisfaction in check_methods [PR108579]
Checks
Commit Message
In check_methods we're unnecessarily checking satisfaction for all
constructors and assignment operators, even those that don't look like
copy/move special members. In the testcase below this manifests as an
unstable satisfaction error because the satisfaction result is first
determined to be false during check_methods (since A<int> is incomplete
at this point) and later true after completion of A<int>.
This patch fixes this simply by swapping the order of the
constraint_satisfied_p and copy_fn_p / move_fn_p tests.
Bootstrapped and regtested on x86_64-pc-linux-gnu, does this look
OK for trunk? This doesn't fix the regression completely, since
we get a similar unstable satisfaction error if one of the constrained
members is actually a copy/move special member. I suppose we need to
rearrange things in finish_struct_1 so that check_methods gets called in
a complete class context?
PR c++/108579
gcc/cp/ChangeLog:
* class.cc (check_methods): Test constraints_satisfied_p after
testing copy_fn_p / move_fn_p instead of beforehand.
gcc/testsuite/ChangeLog:
* g++.dg/cpp2a/concepts-pr108579.C: New test.
---
gcc/cp/class.cc | 16 ++++++++--------
gcc/testsuite/g++.dg/cpp2a/concepts-pr108579.C | 14 ++++++++++++++
2 files changed, 22 insertions(+), 8 deletions(-)
create mode 100644 gcc/testsuite/g++.dg/cpp2a/concepts-pr108579.C
Comments
On 1/30/23 14:10, Patrick Palka wrote:
> In check_methods we're unnecessarily checking satisfaction for all
> constructors and assignment operators, even those that don't look like
> copy/move special members. In the testcase below this manifests as an
> unstable satisfaction error because the satisfaction result is first
> determined to be false during check_methods (since A<int> is incomplete
> at this point) and later true after completion of A<int>.
>
> This patch fixes this simply by swapping the order of the
> constraint_satisfied_p and copy_fn_p / move_fn_p tests.
>
> Bootstrapped and regtested on x86_64-pc-linux-gnu, does this look
> OK for trunk? This doesn't fix the regression completely, since
> we get a similar unstable satisfaction error if one of the constrained
> members is actually a copy/move special member. I suppose we need to
> rearrange things in finish_struct_1 so that check_methods gets called in
> a complete class context?
I think the way to make that work, if indeed that's desirable, would be
to determine those properties lazily instead of at finish_struct time.
The patch is OK.
> PR c++/108579
>
> gcc/cp/ChangeLog:
>
> * class.cc (check_methods): Test constraints_satisfied_p after
> testing copy_fn_p / move_fn_p instead of beforehand.
>
> gcc/testsuite/ChangeLog:
>
> * g++.dg/cpp2a/concepts-pr108579.C: New test.
> ---
> gcc/cp/class.cc | 16 ++++++++--------
> gcc/testsuite/g++.dg/cpp2a/concepts-pr108579.C | 14 ++++++++++++++
> 2 files changed, 22 insertions(+), 8 deletions(-)
> create mode 100644 gcc/testsuite/g++.dg/cpp2a/concepts-pr108579.C
>
> diff --git a/gcc/cp/class.cc b/gcc/cp/class.cc
> index 351de6c5419..d3ce8532d56 100644
> --- a/gcc/cp/class.cc
> +++ b/gcc/cp/class.cc
> @@ -4822,11 +4822,11 @@ check_methods (tree t)
> /* Might be trivial. */;
> else if (TREE_CODE (fn) == TEMPLATE_DECL)
> /* Templates are never special members. */;
> - else if (!constraints_satisfied_p (fn))
> - /* Not eligible. */;
> - else if (copy_fn_p (fn))
> + else if (copy_fn_p (fn)
> + && constraints_satisfied_p (fn))
> TYPE_HAS_COMPLEX_COPY_CTOR (t) = true;
> - else if (move_fn_p (fn))
> + else if (move_fn_p (fn)
> + && constraints_satisfied_p (fn))
> TYPE_HAS_COMPLEX_MOVE_CTOR (t) = true;
> }
>
> @@ -4836,11 +4836,11 @@ check_methods (tree t)
> /* Might be trivial. */;
> else if (TREE_CODE (fn) == TEMPLATE_DECL)
> /* Templates are never special members. */;
> - else if (!constraints_satisfied_p (fn))
> - /* Not eligible. */;
> - else if (copy_fn_p (fn))
> + else if (copy_fn_p (fn)
> + && constraints_satisfied_p (fn))
> TYPE_HAS_COMPLEX_COPY_ASSIGN (t) = true;
> - else if (move_fn_p (fn))
> + else if (move_fn_p (fn)
> + && constraints_satisfied_p (fn))
> TYPE_HAS_COMPLEX_MOVE_ASSIGN (t) = true;
> }
> }
> diff --git a/gcc/testsuite/g++.dg/cpp2a/concepts-pr108579.C b/gcc/testsuite/g++.dg/cpp2a/concepts-pr108579.C
> new file mode 100644
> index 00000000000..bc7d709f889
> --- /dev/null
> +++ b/gcc/testsuite/g++.dg/cpp2a/concepts-pr108579.C
> @@ -0,0 +1,14 @@
> +// PR c++/108579
> +// { dg-do compile { target c++20 } }
> +
> +template<class T>
> +struct A {
> + A(double, char);
> + A(int) requires requires { A(0.0, 'c'); };
> + A& operator=(int) requires requires { A(1.0, 'd'); };
> +};
> +
> +int main() {
> + A<int> x(3);
> + x = 5;
> +}
@@ -4822,11 +4822,11 @@ check_methods (tree t)
/* Might be trivial. */;
else if (TREE_CODE (fn) == TEMPLATE_DECL)
/* Templates are never special members. */;
- else if (!constraints_satisfied_p (fn))
- /* Not eligible. */;
- else if (copy_fn_p (fn))
+ else if (copy_fn_p (fn)
+ && constraints_satisfied_p (fn))
TYPE_HAS_COMPLEX_COPY_CTOR (t) = true;
- else if (move_fn_p (fn))
+ else if (move_fn_p (fn)
+ && constraints_satisfied_p (fn))
TYPE_HAS_COMPLEX_MOVE_CTOR (t) = true;
}
@@ -4836,11 +4836,11 @@ check_methods (tree t)
/* Might be trivial. */;
else if (TREE_CODE (fn) == TEMPLATE_DECL)
/* Templates are never special members. */;
- else if (!constraints_satisfied_p (fn))
- /* Not eligible. */;
- else if (copy_fn_p (fn))
+ else if (copy_fn_p (fn)
+ && constraints_satisfied_p (fn))
TYPE_HAS_COMPLEX_COPY_ASSIGN (t) = true;
- else if (move_fn_p (fn))
+ else if (move_fn_p (fn)
+ && constraints_satisfied_p (fn))
TYPE_HAS_COMPLEX_MOVE_ASSIGN (t) = true;
}
}
new file mode 100644
@@ -0,0 +1,14 @@
+// PR c++/108579
+// { dg-do compile { target c++20 } }
+
+template<class T>
+struct A {
+ A(double, char);
+ A(int) requires requires { A(0.0, 'c'); };
+ A& operator=(int) requires requires { A(1.0, 'd'); };
+};
+
+int main() {
+ A<int> x(3);
+ x = 5;
+}