Mirror of the gdb-patches mailing list
 help / color / mirror / Atom feed
From: Andrew Burgess <andrew.burgess@embecosm.com>
To: gdb-patches@sourceware.org
Subject: [PATCH 1/8] gdbsupport: Provide global operators |=, &=, and ^= for enum bit flags
Date: Thu, 13 Aug 2020 13:58:38 +0100	[thread overview]
Message-ID: <8eb7f409ca572c526de08ab23878be905f5fef42.1597319264.git.andrew.burgess@embecosm.com> (raw)
In-Reply-To: <cover.1597319264.git.andrew.burgess@embecosm.com>

The current enum bit flags mechanism provides global operators &, |,
^, ~, but does not provide &=, |=, ^=.

The implementation for one of these, |=, would look like this:

  template <typename enum_type>
  typename enum_flags_type<enum_type>::type &
  operator|= (enum_type &e1, enum_type e2)
  {
    e1 = enum_flags<enum_type> (e1) | e2;
    return e1;
  }

Then if we create a test within GDB like this:

  enum some_flag
  {
    flag_val1 = 1 << 1,
    flag_val2 = 1 << 2,
    flag_val3 = 1 << 3,
    flag_val4 = 1 << 4,
  };
  DEF_ENUM_FLAGS_TYPE(enum some_flag, some_flags);

  ...

  enum some_flag f = flag_val1 | flag_val2;
  f |= flag_val3;

Initially this didn't compile, with an error like:

  ..../gdbsupport/enum-flags.h: In instantiation of ‘typename enum_flags_type<T>::type& operator|=(enum_type&, enum_type) [with enum_type = some_flag; typename enum_flags_type<T>::type = enum_flags<some_flag>]’:
  ..../gdb/main.c:168:8:   required from here
  ..../gdbsupport/enum-flags.h:217:10: error: cannot bind non-const lvalue reference of type ‘enum_flags_type<some_flag>::type&’ {aka ‘enum_flags<some_flag>&’} to an rvalue of type ‘enum_flags_type<some_flag>::type’ {aka ‘enum_flags<some_flag>’}
    217 |   return e1;
        |          ^~
  ..../gdbsupport/enum-flags.h:125:3: note:   after user-defined conversion: ‘enum_flags<E>::enum_flags(enum_flags<E>::enum_type) [with E = some_flag; enum_flags<E>::enum_type = some_flag]’
    125 |   enum_flags (enum_type e)
        |   ^~~~~~~~~~

If we look at enum_flags_type, which is used in the return type of our
operator, it looks like this:

  template<>
  struct enum_flags_type<enum_type>
  {
    typedef enum_flags<enum_type> type;
  }

So, with `typename' and `typedef' resolved, our new operator actually
compiled like this:

  template <typename enum_type>
  enum_flags<enum_type> &
  operator|= (enum_type &e1, enum_type e2)
  {
    e1 = enum_flags<enum_type> (e1) | e2;
    return e1;
  }

And I think this makes the problem clearer (maybe), the return type
reference is different to the argument types, so the compiler is
forced to try and create a temporary (the result of converting `e1' to
type `enum_flags<enum_type>') and return a reference to that temporary.

What I'd like is for the return type to match the arguments, and this
can be achieved I think by changing the enum_flags_type definition to
this:

  template<>
  struct enum_flags_type<enum_type>
  {
    typedef enum_type type;
  }

Now, once the `typename' and `typedef' are resolved, our new operator
compiles as this:

  template <typename enum_type>
  enum_type &
  operator|= (enum_type &e1, enum_type e2)
  {
    e1 = enum_flags<enum_type> (e1) | e2;
    return e1;
  }

And indeed our original example now compiles fine.

The final thing we must take care to check is that the global
operators are not now picking up _all_ enums, after all, this is part
of what the enum-flags.h file does for us, bitwise operators where we
ask for it, and not otherwise.

A manual test on my demo above (with some_flags) confirms that
removing the DEF_ENUM_FLAGS_TYPE line still prevents the code from
compiling as we'd hope, this is because, just like before
`enum_flags_type<enum_type>::type' is only defined for the enums that
have DEF_ENUM_FLAGS_TYPE applied to them.

gdbsupport/ChangeLog:

	* enum-flags.h (DEF_ENUM_FLAGS_TYPE): Update header comment, and
	define enum_flags_type<enum_type>::type differently.
	(operator|=): New.
	(operator&=): New.
	(operator^=): New.
---
 gdbsupport/ChangeLog    |  8 ++++++++
 gdbsupport/enum-flags.h | 30 +++++++++++++++++++++++++++---
 2 files changed, 35 insertions(+), 3 deletions(-)

diff --git a/gdbsupport/enum-flags.h b/gdbsupport/enum-flags.h
index 825ff4faf2c..78db3c7d88e 100644
--- a/gdbsupport/enum-flags.h
+++ b/gdbsupport/enum-flags.h
@@ -55,15 +55,15 @@
    instantiating for non-flag enums.  */
 template<typename T> struct enum_flags_type {};
 
-/* Use this to mark an enum as flags enum.  It defines FLAGS as
+/* Use this to mark an enum as flags enum.  It defines FLAGS_TYPE as
    enum_flags wrapper class for ENUM, and enables the global operator
-   overloads for ENUM.  */
+   overloads for ENUM_TYPE.  */
 #define DEF_ENUM_FLAGS_TYPE(enum_type, flags_type)	\
   typedef enum_flags<enum_type> flags_type;		\
   template<>						\
   struct enum_flags_type<enum_type>			\
   {							\
-    typedef enum_flags<enum_type> type;			\
+    typedef enum_type type;				\
   }
 
 /* Until we can rely on std::underlying type being universally
@@ -209,6 +209,30 @@ operator~ (enum_type e)
   return ~enum_flags<enum_type> (e);
 }
 
+template <typename enum_type>
+typename enum_flags_type<enum_type>::type
+operator|= (enum_type &e1, const enum_type &e2)
+{
+  e1 = enum_flags<enum_type> (e1) | e2;
+  return e1;
+}
+
+template <typename enum_type>
+typename enum_flags_type<enum_type>::type
+operator&= (enum_type &e1, const enum_type &e2)
+{
+  e1 = enum_flags<enum_type> (e1) & e2;
+  return e1;
+}
+
+template <typename enum_type>
+typename enum_flags_type<enum_type>::type
+operator^= (enum_type &e1, const enum_type &e2)
+{
+  e1 = enum_flags<enum_type> (e1) ^ e2;
+  return e1;
+}
+
 #else /* __cplusplus */
 
 /* In C, the flags type is just a typedef for the enum type.  */
-- 
2.25.4



  reply	other threads:[~2020-08-13 12:58 UTC|newest]

Thread overview: 62+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2020-08-13 12:58 [PATCH 0/8] Fortran Array Slicing and Striding Support Andrew Burgess
2020-08-13 12:58 ` Andrew Burgess [this message]
2020-08-15 17:16   ` [PATCH 1/8] gdbsupport: Provide global operators |=, &=, and ^= for enum bit flags Tom Tromey
2020-08-16  9:13     ` Andrew Burgess
2020-08-17 10:40     ` Andrew Burgess
2020-08-20 16:00       ` Pedro Alves
2020-08-21 14:49       ` Pedro Alves
2020-08-21 15:57         ` Andrew Burgess
2020-08-21 18:10           ` Pedro Alves
2020-08-13 12:58 ` [PATCH 2/8] gdbsupport: Make function arguments constant in enum-flags.h Andrew Burgess
2020-08-15 19:45   ` Tom Tromey
2020-08-16  9:08     ` Andrew Burgess
2020-08-13 12:58 ` [PATCH 3/8] gdb/fortran: Clean up array/string expression evaluation Andrew Burgess
2020-08-13 12:58 ` [PATCH 4/8] gdb/fortran: Move Fortran expression handling into f-lang.c Andrew Burgess
2020-08-13 12:58 ` [PATCH 5/8] gdb/fortran: Change whitespace when printing arrays Andrew Burgess
2020-08-13 12:58 ` [PATCH 6/8] gdb: Convert enum range_type to a bit field enum Andrew Burgess
2020-08-13 12:58 ` [PATCH 7/8] gdb/testsuite: Add missing expected results Andrew Burgess
2020-08-13 12:58 ` [PATCH 8/8] gdb/fortran: Add support for Fortran array slices at the GDB prompt Andrew Burgess
2020-08-13 13:31   ` Eli Zaretskii
2020-08-26 14:49 ` [PATCHv2 00/10] Fortran Array Slicing and Striding Support Andrew Burgess
2020-08-26 14:49   ` [PATCHv2 01/10] Rewrite valid-expr.h's internals in terms of the detection idiom (C++17/N4502) Andrew Burgess
2020-08-26 14:49   ` [PATCHv2 02/10] Use type_instance_flags more throughout Andrew Burgess
2020-08-26 14:49   ` [PATCHv2 03/10] Rewrite enum_flags, add unit tests, fix problems Andrew Burgess
2020-08-26 14:49   ` [PATCHv2 04/10] gdb: additional changes to make use of type_instance_flags more Andrew Burgess
2020-08-26 14:49   ` [PATCHv2 05/10] gdb/fortran: Clean up array/string expression evaluation Andrew Burgess
2020-09-19  8:53     ` Andrew Burgess
2020-08-26 14:49   ` [PATCHv2 06/10] gdb/fortran: Move Fortran expression handling into f-lang.c Andrew Burgess
2020-09-19  8:53     ` Andrew Burgess
2020-08-26 14:49   ` [PATCHv2 07/10] gdb/fortran: Change whitespace when printing arrays Andrew Burgess
2020-09-19  8:54     ` Andrew Burgess
2020-08-26 14:49   ` [PATCHv2 08/10] gdb: Convert enum range_type to a bit field enum Andrew Burgess
2020-08-26 14:49   ` [PATCHv2 09/10] gdb/testsuite: Add missing expected results Andrew Burgess
2020-09-18  9:53     ` Andrew Burgess
2020-08-26 14:49   ` [PATCHv2 10/10] gdb/fortran: Add support for Fortran array slices at the GDB prompt Andrew Burgess
2020-08-26 17:02     ` Eli Zaretskii
2020-09-19  9:47   ` [PATCHv3 0/2] Fortran Array Slicing and Striding Support Andrew Burgess
2020-09-19  9:48     ` [PATCHv3 1/2] gdb: Convert enum range_type to a bit field enum Andrew Burgess
2020-09-19 13:50       ` Simon Marchi
2020-09-19  9:48     ` [PATCHv3 2/2] gdb/fortran: Add support for Fortran array slices at the GDB prompt Andrew Burgess
2020-09-19 10:03       ` Eli Zaretskii
2020-09-28  9:40     ` [PATCHv4 0/3] Fortran Array Slicing and Striding Support Andrew Burgess
2020-09-28  9:40       ` [PATCHv4 1/3] gdb: Convert enum range_type to a bit field enum Andrew Burgess
2020-09-28  9:40       ` [PATCHv4 2/3] gdb: rename 'enum range_type' to 'enum range_flag' Andrew Burgess
2020-09-28  9:40       ` [PATCHv4 3/3] gdb/fortran: Add support for Fortran array slices at the GDB prompt Andrew Burgess
2020-09-28  9:52         ` Eli Zaretskii via Gdb-patches
2020-10-11 18:12       ` [PATCHv5 0/4] Fortran Array Slicing and Striding Support Andrew Burgess
2020-10-11 18:12         ` [PATCHv5 1/4] gdb: Convert enum range_type to a bit field enum Andrew Burgess
2020-10-20 20:16           ` Tom Tromey
2020-10-11 18:12         ` [PATCHv5 2/4] gdb: rename 'enum range_type' to 'enum range_flag' Andrew Burgess
2020-10-20 20:16           ` Tom Tromey
2020-10-11 18:12         ` [PATCHv5 3/4] gdb/fortran: add support for parsing array strides in expressions Andrew Burgess
2020-10-12 13:21           ` Simon Marchi
2020-10-20 20:17           ` Tom Tromey
2020-10-22 10:42           ` Andrew Burgess
2020-10-11 18:12         ` [PATCHv5 4/4] gdb/fortran: Add support for Fortran array slices at the GDB prompt Andrew Burgess
2020-10-12 14:10           ` Simon Marchi
2020-10-20 20:45           ` Tom Tromey
2020-10-29 11:08             ` Andrew Burgess
2020-10-31 22:16           ` [PATCHv6] " Andrew Burgess
2020-11-12 12:09             ` Andrew Burgess
2020-11-12 18:58             ` Tom Tromey
2020-11-19 11:56             ` Andrew Burgess

Reply instructions:

You may reply publicly to this message via plain-text email
using any one of the following methods:

* Save the following mbox file, import it into your mail client,
  and reply-to-all from there: mbox

  Avoid top-posting and favor interleaved quoting:
  https://en.wikipedia.org/wiki/Posting_style#Interleaved_style

* Reply using the --to, --cc, and --in-reply-to
  switches of git-send-email(1):

  git send-email \
    --in-reply-to=8eb7f409ca572c526de08ab23878be905f5fef42.1597319264.git.andrew.burgess@embecosm.com \
    --to=andrew.burgess@embecosm.com \
    --cc=gdb-patches@sourceware.org \
    /path/to/YOUR_REPLY

  https://kernel.org/pub/software/scm/git/docs/git-send-email.html

* If your mail client supports setting the In-Reply-To header
  via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox