[1/3] Rewrite valid-expr.h's internals in terms of the detection idiom (C++17/N4502)

Message ID 20200821144523.19451-2-pedro@palves.net
State Superseded
Headers show
Series
  • Rewrite enum_flags, add unit tests, fix problems
Related show

Commit Message

Pedro Alves Aug. 21, 2020, 2:45 p.m.
An earlier attempt at doing this had failed (wouldn't work in GCCs
around 4.8, IIRC), but now that I try again, it works.  I suspect that
my previous attempt did not use the pre C++14-safe void_t (in
traits.h).

I want to switch to this model because:

 - It's the standard detection idiom that folks will learn starting
   with C++17.

 - In the enum_flags unit tests, I have a static_assert that triggers
   a warning (resulting in build error), which GCC does not suppress
   because the warning is not being triggered in the SFINAE context.
   Switching to the detection idiom fixes that.  Alternatively,
   switching to the C++03-style expression-validity checking with a
   varargs overload would allow addressing that, but I think that
   would be going backwards idiomatically speaking.

 - While this patch shows a net increase of lines of code, the magic
   being added to traits.h can be removed in a few years when we start
   requiring C++17.

gdbsupport/ChangeLog:

	* traits.h (struct nonesuch, struct detector, detected_or)
	(detected_or_t, is_detected, detected_t, detected_or)
	(detected_or_t, is_detected_exact, is_detected_convertible): New.
	* valid-expr.h (CHECK_VALID_EXPR_INT): Use gdb::is_detected_exact.
---
 gdbsupport/traits.h     | 67 +++++++++++++++++++++++++++++++++++++++++++++++++
 gdbsupport/valid-expr.h | 20 +++------------
 2 files changed, 70 insertions(+), 17 deletions(-)

-- 
2.14.5

Comments

Tom Tromey Sept. 28, 2020, 6:59 p.m. | #1
>>>>> "Pedro" == Pedro Alves <pedro@palves.net> writes:


Pedro> +  static_assert (gdb::is_detected_exact<EXPR_TYPE,			\
Pedro> +		 archetype, TYPES>::value == VALID,			\
Pedro>  		 "");							\


I'm not sure if it was exactly this patch, but after I merged this code
to our internal tree, we started having build problems on a machine that
uses GCC 6.4.  I've appended the error message.

I'm not sure it is worth doing anything about this, but I thought it
would be worth noting in case anybody else winds up in this situation.

Tom

../../src/gdb/unittests/offset-type-selftests.c: In substitution of 'template<class Expected, template<class ...> class Op, class ... Args> using is_detected_exact = std::is_same<Expected, typename gdb::detection_detail::detector<gdb::nonesuch, void, Op, Args ...>::type> [with Expected = selftests::offset_type::off_A&; Op = selftests::offset_type::check_valid_expr75::archetype; Args = {selftests::offset_type::off_A, selftests::offset_type::off_B}]':
../../src/gdb/unittests/offset-type-selftests.c:75:1:   required from here
../../src/gdb/../gdbsupport/valid-expr.h:65:20: error: type/value mismatch at argument 2 in template parameter list for 'template<class Expected, template<class ...> class Op, class ... Args> using is_detected_exact = std::is_same<Expected, typename gdb::detection_detail::detector<gdb::nonesuch, void, Op, Args ...>::type>'
    archetype, TYPES>::value == VALID,   \
                    ^
../../src/gdb/../gdbsupport/valid-expr.h:79:3: note: in expansion of macro 'CHECK_VALID_EXPR_INT'
   CHECK_VALID_EXPR_INT (ESC_PARENS(typename T1, typename T2),  \
   ^~~~~~~~~~~~~~~~~~~~
../../src/gdb/unittests/offset-type-selftests.c:42:3: note: in expansion of macro 'CHECK_VALID_EXPR_2'
   CHECK_VALID_EXPR_2 (off_A, off_B, VALID, EXPR_TYPE, EXPR)
   ^~~~~~~~~~~~~~~~~~
../../src/gdb/unittests/offset-type-selftests.c:75:1: note: in expansion of macro 'CHECK_VALID'
 CHECK_VALID (true,  off_A&, lval_a += undrl {});
 ^~~~~~~~~~~
../../src/gdb/../gdbsupport/valid-expr.h:65:20: note:   expected a class template, got 'selftests::offset_type::check_valid_expr75::archetype'
    archetype, TYPES>::value == VALID,   \
                    ^
../../src/gdb/../gdbsupport/valid-expr.h:79:3: note: in expansion of macro 'CHECK_VALID_EXPR_INT'
   CHECK_VALID_EXPR_INT (ESC_PARENS(typename T1, typename T2),  \
   ^~~~~~~~~~~~~~~~~~~~
../../src/gdb/unittests/offset-type-selftests.c:42:3: note: in expansion of macro 'CHECK_VALID_EXPR_2'
   CHECK_VALID_EXPR_2 (off_A, off_B, VALID, EXPR_TYPE, EXPR)
   ^~~~~~~~~~~~~~~~~~
../../src/gdb/unittests/offset-type-selftests.c:75:1: note: in expansion of macro 'CHECK_VALID'
 CHECK_VALID (true,  off_A&, lval_a += undrl {});
 ^~~~~~~~~~~
Ulrich Weigand via Gdb-patches Sept. 28, 2020, 7:58 p.m. | #2
On 9/28/20 3:59 PM, Tom Tromey wrote:
>>>>>> "Pedro" == Pedro Alves <pedro@palves.net> writes:

> 

> Pedro> +  static_assert (gdb::is_detected_exact<EXPR_TYPE,			\

> Pedro> +		 archetype, TYPES>::value == VALID,			\

> Pedro>  		 "");							\

> 

> 

> I'm not sure if it was exactly this patch, but after I merged this code

> to our internal tree, we started having build problems on a machine that

> uses GCC 6.4.  I've appended the error message.

> 

> I'm not sure it is worth doing anything about this, but I thought it

> would be worth noting in case anybody else winds up in this situation.

> 

> Tom


It was reported before on this thread a week or so ago. Unless GCC 6.4 
isn't supported any longer, I think it should be fixed.

If GCC 6.4 isn't supported, then we should make that explicit on the 
documentation to spare folks the trouble of having to figure it out.
Tom Tromey Sept. 28, 2020, 8 p.m. | #3
>> I'm not sure if it was exactly this patch, but after I merged this code

>> to our internal tree, we started having build problems on a machine that

>> uses GCC 6.4.  I've appended the error message.

>> I'm not sure it is worth doing anything about this, but I thought it

>> would be worth noting in case anybody else winds up in this situation.

>> Tom


Luis> It was reported before on this thread a week or so ago. Unless GCC 6.4
Luis> isn't supported any longer, I think it should be fixed.

Ok, thanks.  I will find out tonight.

Tom
Tom Tromey Sept. 29, 2020, 1:04 p.m. | #4
>>> I'm not sure if it was exactly this patch, but after I merged this code

>>> to our internal tree, we started having build problems on a machine that

>>> uses GCC 6.4.  I've appended the error message.

>>> I'm not sure it is worth doing anything about this, but I thought it

>>> would be worth noting in case anybody else winds up in this situation.

>>> Tom


Luis> It was reported before on this thread a week or so ago. Unless GCC 6.4
Luis> isn't supported any longer, I think it should be fixed.

Tom> Ok, thanks.  I will find out tonight.

Oops, I really misunderstood your note, and re-reading it now I can't
really understand why.  I thought you were saying it was fixed, but
really you were saying it is still there and should be fixed if we want
to support GCC 6.4.

I am not sure if we should try to fix it or not.  Locally I plan to
disable unit tests on this particular machine.  It is the only one using
6.4.

Tom
Pedro Alves Sept. 29, 2020, 7:24 p.m. | #5
On 9/29/20 2:04 PM, Tom Tromey wrote:
>>>> I'm not sure if it was exactly this patch, but after I merged this code

>>>> to our internal tree, we started having build problems on a machine that

>>>> uses GCC 6.4.  I've appended the error message.

>>>> I'm not sure it is worth doing anything about this, but I thought it

>>>> would be worth noting in case anybody else winds up in this situation.

>>>> Tom

> 

> Luis> It was reported before on this thread a week or so ago. Unless GCC 6.4

> Luis> isn't supported any longer, I think it should be fixed.

> 

> Tom> Ok, thanks.  I will find out tonight.

> 

> Oops, I really misunderstood your note, and re-reading it now I can't

> really understand why.  I thought you were saying it was fixed, but

> really you were saying it is still there and should be fixed if we want

> to support GCC 6.4.

> 

> I am not sure if we should try to fix it or not.  Locally I plan to

> disable unit tests on this particular machine.  It is the only one using

> 6.4.


I built a GCC 6.4 here, and could reproduce it.  I'm testing a fix.

Patch

diff --git a/gdbsupport/traits.h b/gdbsupport/traits.h
index 2a6f00654c..93b609ac10 100644
--- a/gdbsupport/traits.h
+++ b/gdbsupport/traits.h
@@ -52,6 +52,73 @@  struct make_void { typedef void type; };
 template<typename... Ts>
 using void_t = typename make_void<Ts...>::type;
 
+/* Implementation of the detection idiom:
+
+   - http://www.open-std.org/jtc1/sc22/wg21/docs/papers/2015/n4502.pdf
+   - http://en.cppreference.com/w/cpp/experimental/is_detected
+
+*/
+
+struct nonesuch
+{
+  nonesuch () = delete;
+  ~nonesuch () = delete;
+  nonesuch (const nonesuch &) = delete;
+  void operator= (const nonesuch &) = delete;
+};
+
+namespace detection_detail {
+/* Implementation of the detection idiom (negative case).  */
+template<typename Default, typename AlwaysVoid,
+	 template<typename...> class Op, typename... Args>
+struct detector
+{
+  using value_t = std::false_type;
+  using type = Default;
+};
+
+/* Implementation of the detection idiom (positive case).  */
+template<typename Default, template<typename...> class Op, typename... Args>
+struct detector<Default, void_t<Op<Args...>>, Op, Args...>
+{
+  using value_t = std::true_type;
+  using type = Op<Args...>;
+};
+
+/* Detect whether Op<Args...> is a valid type, use Default if not.  */
+template<typename Default, template<typename...> class Op,
+	 typename... Args>
+using detected_or = detector<Default, void, Op, Args...>;
+
+/* Op<Args...> if that is a valid type, otherwise Default.  */
+template<typename Default, template<typename...> class Op,
+	 typename... Args>
+using detected_or_t
+  = typename detected_or<Default, Op, Args...>::type;
+
+} /* detection_detail */
+
+template<template<typename...> class Op, typename... Args>
+using is_detected
+  = typename detection_detail::detector<nonesuch, void, Op, Args...>::value_t;
+
+template<template<typename...> class Op, typename... Args>
+using detected_t
+  = typename detection_detail::detector<nonesuch, void, Op, Args...>::type;
+
+template<typename Default, template<typename...> class Op, typename... Args>
+using detected_or = detection_detail::detected_or<Default, Op, Args...>;
+
+template<typename Default, template<typename...> class Op, typename... Args>
+using detected_or_t = typename detected_or<Default, Op, Args...>::type;
+
+template<typename Expected, template<typename...> class Op, typename... Args>
+using is_detected_exact = std::is_same<Expected, detected_t<Op, Args...>>;
+
+template<typename To, template<typename...> class Op, typename... Args>
+using is_detected_convertible
+  = std::is_convertible<detected_t<Op, Args...>, To>;
+
 /* A few trait helpers, mainly stolen from libstdc++.  Uppercase
    because "and/or", etc. are reserved keywords.  */
 
diff --git a/gdbsupport/valid-expr.h b/gdbsupport/valid-expr.h
index b1c8446814..a22fa61134 100644
--- a/gdbsupport/valid-expr.h
+++ b/gdbsupport/valid-expr.h
@@ -58,26 +58,12 @@ 
 #define CHECK_VALID_EXPR_INT(TYPENAMES, TYPES, VALID, EXPR_TYPE, EXPR)	\
   namespace CONCAT (check_valid_expr, __LINE__) {			\
 									\
-  template<typename, typename, typename = void>				\
-  struct is_valid_expression						\
-    : std::false_type {};						\
-									\
   template <TYPENAMES>							\
-    struct is_valid_expression<TYPES, gdb::void_t<decltype (EXPR)>>	\
-    : std::true_type {};						\
+    using archetype = decltype (EXPR);					\
 									\
-  static_assert (is_valid_expression<TYPES>::value == VALID,		\
+  static_assert (gdb::is_detected_exact<EXPR_TYPE,			\
+		 archetype, TYPES>::value == VALID,			\
 		 "");							\
-									\
-  template<TYPENAMES, typename = void>					\
-  struct is_same_type							\
-    : std::is_same<EXPR_TYPE, void> {};					\
-									\
-  template <TYPENAMES>							\
-    struct is_same_type<TYPES, gdb::void_t<decltype (EXPR)>>		\
-    : std::is_same<EXPR_TYPE, decltype (EXPR)> {};			\
-									\
-  static_assert (is_same_type<TYPES>::value, "");			\
   } /* namespace */
 
 /* A few convenience macros that support expressions involving a