From: Simon Marchi <simark@simark.ca>
To: Pedro Alves <pedro@palves.net>, gdb-patches@sourceware.org
Subject: Re: [PATCH v3] enum_flags to_string (was: Re: [PATCH v2 07/29] Thread options & clone events (core + remote))
Date: Fri, 28 Oct 2022 11:59:25 -0400 [thread overview]
Message-ID: <6aee3c00-b334-97e2-62f7-0434822bcf7a@simark.ca> (raw)
In-Reply-To: <6b91f8f6-d3b5-c44b-297c-ce1c3a1c80f6@palves.net>
On 2022-10-28 07 h 08, Pedro Alves wrote:
> On 2022-10-28 11:26 a.m., Pedro Alves wrote:
>> Some self review. :-)
>>
>> On 2022-10-27 8:55 p.m., Pedro Alves wrote:
>>
>>>
>>> static std::string
>>> to_string (some_flags flags)
>>> {
>>> #define CASE(X) { X, #X }
>>> static constexpr test_flags::string_map_element mapping[] = {
>>> CASE (SOME_FLAG1),
>>> CASE (SOME_FLAG2),
>>> CASE (SOME_FLAG3),
>>> };
>>> #undef CASE
>>>
>>> return flags.to_string (mapping);
>>> }
>>
>> I was looking at this again, and noticed I named the macro "CASE". That's just
>> because that style of macro is typically used in a switch/case, my first approach
>> used one. That isn't what's being done in the end, so better rename it. And while doing that,
>> I realized that you don't even need to name the array or its type. Like so:
>>
>> static std::string
>> to_string (some_flags flags)
>> {
>> #define MAP_ENUM_FLAG(X) { X, #X }
>> return flags.to_string ({
>> MAP_ENUM_FLAG (SOME_FLAG1),
>> MAP_ENUM_FLAG (SOME_FLAG3),
>> });
>> #undef MAP_ENUM_FLAG
>> }
>>
>> But then, every enum_flags to_string implementation will end up defining/undefining that
>> same macro, so might as well put it in gdbsupport/enum-flags.h, resulting in:
>>
>> static std::string
>> to_string (test_flags flags)
>> {
>> return flags.to_string ({
>> MAP_ENUM_FLAG (FLAG1),
>> MAP_ENUM_FLAG (FLAG3),
>> });
>> }
>
> Bah, I now looked at the generated code at -O2, and with the unnamed array, the compiler
> emits code to manually compose a temporary array on the stack at every call...
Maybe because of this?
https://tristanbrindle.com/posts/beware-copies-initializer-list
>
> 3rd time's a charm?
>
> -- >8 --
> From 0fe3edef69360c6795ab6811c4f3cba06159a50d Mon Sep 17 00:00:00 2001
> From: Pedro Alves <pedro@palves.net>
> Date: Tue, 25 Oct 2022 15:39:37 +0100
> Subject: [PATCH] enum_flags to_string
>
> This commit introduces shared infrastructure that can be used to
> implement enum_flags -> to_string functions. With this, if we want to
> support converting a given enum_flags specialization to string, we
> just need to implement a function that provides the enumerator->string
> mapping, like so:
>
> enum some_flag
> {
> SOME_FLAG1 = 1 << 0,
> SOME_FLAG2 = 1 << 1,
> SOME_FLAG3 = 1 << 2,
> };
>
> DEF_ENUM_FLAGS_TYPE (some_flag, some_flags);
>
> static std::string
> to_string (some_flags flags)
> {
> static constexpr some_flags::string_mapping mapping[] = {
> MAP_ENUM_FLAG (SOME_FLAG1),
> MAP_ENUM_FLAG (SOME_FLAG2),
> MAP_ENUM_FLAG (SOME_FLAG3),
> };
> return flags.to_string (mapping);
> }
>
> .. and then to_string(SOME_FLAG2 | SOME_FLAG3) produces a string like
> "0x6 [SOME_FLAG2 SOME_FLAG3]".
>
> If we happen to forget to update the mapping array when we introduce a
> new enumerator, then the string representation will pretty-print the
> flags it knows about, and then the leftover flags in hex (one single
> number). For example, if we had missed mapping SOME_FLAG2 above, we'd
> end up with:
>
> to_string(SOME_FLAG2 | SOME_FLAG3) => "0x6 [SOME_FLAG2 0x4]");
>
> Other than in the unit tests included, no actual usage of the
> functionality is added in this commit.
>
> Change-Id: I835de43c33d13bc0c95132f42c3f97318b875779
> ---
> gdb/unittests/enum-flags-selftests.c | 24 ++++++++++++
> gdbsupport/enum-flags.h | 57 ++++++++++++++++++++++++++++
> 2 files changed, 81 insertions(+)
>
> diff --git a/gdb/unittests/enum-flags-selftests.c b/gdb/unittests/enum-flags-selftests.c
> index f52fc7220a1..ef43ba23cdf 100644
> --- a/gdb/unittests/enum-flags-selftests.c
> +++ b/gdb/unittests/enum-flags-selftests.c
> @@ -374,6 +374,20 @@ enum test_uflag : unsigned
> DEF_ENUM_FLAGS_TYPE (test_flag, test_flags);
> DEF_ENUM_FLAGS_TYPE (test_uflag, test_uflags);
>
> +/* to_string enumerator->string mapping function used to test
> + enum_flags::to_string. This intentionally misses mapping one
> + enumerator (FLAG2). */
> +
> +static std::string
> +to_string (test_flags flags)
> +{
> + static constexpr test_flags::string_mapping mapping[] = {
> + MAP_ENUM_FLAG (FLAG1),
> + MAP_ENUM_FLAG (FLAG3),
> + };
> + return flags.to_string (mapping);
> +}
> +
> static void
> self_test ()
> {
> @@ -581,6 +595,16 @@ self_test ()
>
> SELF_CHECK (ok);
> }
> +
> + /* Check string conversion. */
> + {
> + SELF_CHECK (to_string (0) == "0x0 []");
> + SELF_CHECK (to_string (FLAG1) == "0x2 [FLAG1]");
> + SELF_CHECK (to_string (FLAG1 | FLAG3) == "0xa [FLAG1 FLAG3]");
> + SELF_CHECK (to_string (FLAG1 | FLAG2 | FLAG3) == "0xe [FLAG1 FLAG3 0x4]");
> + SELF_CHECK (to_string (FLAG2) == "0x4 [0x4]");
> + SELF_CHECK (to_string (test_flag (0xff)) == "0xff [FLAG1 FLAG3 0xf5]");
Ish, this crashes with:
/home/smarchi/src/binutils-gdb/gdb/../gdbsupport/enum-flags.h:191:12: runtime error: load of value 255, which is not a valid value for type 'test_flag'
with --enable-ubsan. We can disable UBSan per function using
__attribute__((no_sanitize_undefined)) or
__attribute__((no_sanitize("undefined"))). With clang, it appears
possible to push and pop function attributes, but it doesn't seem to be
possible with gcc.
I had a hard time identifying which functions needed it, so I slapped it
on all functions in enum-flags.h, and the test passed fine, so that's
good news. Here's my change, to get you started:
From a2cac0cd6b85de82ce96d5b565f83d152482d403 Mon Sep 17 00:00:00 2001
From: Simon Marchi <simon.marchi@efficios.com>
Date: Fri, 28 Oct 2022 11:56:01 -0400
Subject: [PATCH] fix
Change-Id: I229d00382d8c0e7af0987f6720bab1ef4094290f
---
gdbsupport/enum-flags.h | 25 ++++++++++++++++++++++++-
1 file changed, 24 insertions(+), 1 deletion(-)
diff --git a/gdbsupport/enum-flags.h b/gdbsupport/enum-flags.h
index 911f3fd7890..48b38dc88b5 100644
--- a/gdbsupport/enum-flags.h
+++ b/gdbsupport/enum-flags.h
@@ -158,16 +158,19 @@ class enum_flags
: m_enum_value ((enum_type) 0)
{}
+ __attribute__((no_sanitize_undefined))
enum_flags &operator&= (enum_flags e) &
{
m_enum_value = (enum_type) (m_enum_value & e.m_enum_value);
return *this;
}
+ __attribute__((no_sanitize_undefined))
enum_flags &operator|= (enum_flags e) &
{
m_enum_value = (enum_type) (m_enum_value | e.m_enum_value);
return *this;
}
+ __attribute__((no_sanitize_undefined))
enum_flags &operator^= (enum_flags e) &
{
m_enum_value = (enum_type) (m_enum_value ^ e.m_enum_value);
@@ -180,12 +183,14 @@ class enum_flags
void operator^= (enum_flags e) && = delete;
/* Like raw enums, allow conversion to the underlying type. */
+__attribute__((no_sanitize_undefined))
constexpr operator underlying_type () const
{
return m_enum_value;
}
/* Get the underlying value as a raw enum. */
+ __attribute__((no_sanitize_undefined))
constexpr enum_type raw () const
{
return m_enum_value;
@@ -201,7 +206,7 @@ class enum_flags
mapping array from callers. */
template<size_t N>
- std::string
+ __attribute__((no_sanitize_undefined)) std::string
to_string (const string_mapping (&mapping)[N]) const
{
enum_type flags = raw ();
@@ -259,6 +264,7 @@ using is_enum_flags_enum_type_t
/* Raw enum on both LHS/RHS. Returns raw enum type. */ \
template <typename enum_type, \
typename = is_enum_flags_enum_type_t<enum_type>> \
+ __attribute__((no_sanitize_undefined)) \
constexpr enum_type \
OPERATOR_OP (enum_type e1, enum_type e2) \
{ \
@@ -269,6 +275,7 @@ using is_enum_flags_enum_type_t
/* enum_flags on the LHS. */ \
template <typename enum_type, \
typename = is_enum_flags_enum_type_t<enum_type>> \
+ __attribute__((no_sanitize_undefined)) \
constexpr enum_flags<enum_type> \
OPERATOR_OP (enum_flags<enum_type> e1, enum_type e2) \
{ return e1.raw () OP e2; } \
@@ -276,6 +283,7 @@ using is_enum_flags_enum_type_t
/* enum_flags on the RHS. */ \
template <typename enum_type, \
typename = is_enum_flags_enum_type_t<enum_type>> \
+ __attribute__((no_sanitize_undefined)) \
constexpr enum_flags<enum_type> \
OPERATOR_OP (enum_type e1, enum_flags<enum_type> e2) \
{ return e1 OP e2.raw (); } \
@@ -283,6 +291,7 @@ using is_enum_flags_enum_type_t
/* enum_flags on both LHS/RHS. */ \
template <typename enum_type, \
typename = is_enum_flags_enum_type_t<enum_type>> \
+ __attribute__((no_sanitize_undefined)) \
constexpr enum_flags<enum_type> \
OPERATOR_OP (enum_flags<enum_type> e1, enum_flags<enum_type> e2) \
{ return e1.raw () OP e2.raw (); } \
@@ -291,21 +300,25 @@ using is_enum_flags_enum_type_t
\
template <typename enum_type, typename unrelated_type, \
typename = is_enum_flags_enum_type_t<enum_type>> \
+ __attribute__((no_sanitize_undefined)) \
constexpr enum_flags<enum_type> \
OPERATOR_OP (enum_type e1, unrelated_type e2) = delete; \
\
template <typename enum_type, typename unrelated_type, \
typename = is_enum_flags_enum_type_t<enum_type>> \
+ __attribute__((no_sanitize_undefined)) \
constexpr enum_flags<enum_type> \
OPERATOR_OP (unrelated_type e1, enum_type e2) = delete; \
\
template <typename enum_type, typename unrelated_type, \
typename = is_enum_flags_enum_type_t<enum_type>> \
+ __attribute__((no_sanitize_undefined)) \
constexpr enum_flags<enum_type> \
OPERATOR_OP (enum_flags<enum_type> e1, unrelated_type e2) = delete; \
\
template <typename enum_type, typename unrelated_type, \
typename = is_enum_flags_enum_type_t<enum_type>> \
+ __attribute__((no_sanitize_undefined)) \
constexpr enum_flags<enum_type> \
OPERATOR_OP (unrelated_type e1, enum_flags<enum_type> e2) = delete;
@@ -333,6 +346,7 @@ using is_enum_flags_enum_type_t
/* lval reference version. */ \
template <typename enum_type, \
typename = is_enum_flags_enum_type_t<enum_type>> \
+ __attribute__((no_sanitize_undefined)) \
constexpr enum_type & \
OPERATOR_OP (enum_type &e1, enum_type e2) \
{ return e1 = e1 OP e2; } \
@@ -377,6 +391,7 @@ ENUM_FLAGS_GEN_COMPOUND_ASSIGN (operator^=, ^)
/* enum_flags OP enum_flags */ \
\
template <typename enum_type> \
+ __attribute__((no_sanitize_undefined)) \
constexpr bool \
OPERATOR_OP (enum_flags<enum_type> lhs, enum_flags<enum_type> rhs) \
{ return lhs.raw () OP rhs.raw (); } \
@@ -384,32 +399,38 @@ ENUM_FLAGS_GEN_COMPOUND_ASSIGN (operator^=, ^)
/* enum_flags OP other */ \
\
template <typename enum_type> \
+ __attribute__((no_sanitize_undefined)) \
constexpr bool \
OPERATOR_OP (enum_flags<enum_type> lhs, enum_type rhs) \
{ return lhs.raw () OP rhs; } \
\
template <typename enum_type> \
+ __attribute__((no_sanitize_undefined)) \
constexpr bool \
OPERATOR_OP (enum_flags<enum_type> lhs, int rhs) \
{ return lhs.raw () OP rhs; } \
\
template <typename enum_type, typename U> \
+ __attribute__((no_sanitize_undefined)) \
constexpr bool \
OPERATOR_OP (enum_flags<enum_type> lhs, U rhs) = delete; \
\
/* other OP enum_flags */ \
\
template <typename enum_type> \
+ __attribute__((no_sanitize_undefined)) \
constexpr bool \
OPERATOR_OP (enum_type lhs, enum_flags<enum_type> rhs) \
{ return lhs OP rhs.raw (); } \
\
template <typename enum_type> \
+ __attribute__((no_sanitize_undefined)) \
constexpr bool \
OPERATOR_OP (int lhs, enum_flags<enum_type> rhs) \
{ return lhs OP rhs.raw (); } \
\
template <typename enum_type, typename U> \
+ __attribute__((no_sanitize_undefined)) \
constexpr bool \
OPERATOR_OP (U lhs, enum_flags<enum_type> rhs) = delete;
@@ -426,6 +447,7 @@ template <typename enum_type,
typename = is_enum_flags_enum_type_t<enum_type>,
typename
= gdb::Requires<enum_flags_detail::EnumIsUnsigned<enum_type>>>
+__attribute__((no_sanitize_undefined))
constexpr enum_type
operator~ (enum_type e)
{
@@ -442,6 +464,7 @@ template <typename enum_type,
typename = is_enum_flags_enum_type_t<enum_type>,
typename
= gdb::Requires<enum_flags_detail::EnumIsUnsigned<enum_type>>>
+__attribute__((no_sanitize_undefined))
constexpr enum_flags<enum_type>
operator~ (enum_flags<enum_type> e)
{
--
2.37.3
> + }
> }
>
> } /* namespace enum_flags_tests */
> diff --git a/gdbsupport/enum-flags.h b/gdbsupport/enum-flags.h
> index cd500f55ff3..911f3fd7890 100644
> --- a/gdbsupport/enum-flags.h
> +++ b/gdbsupport/enum-flags.h
> @@ -130,6 +130,17 @@ class enum_flags
> typedef E enum_type;
> typedef typename enum_underlying_type<enum_type>::type underlying_type;
>
> + /* For to_string. Maps one enumerator of E to a string. */
> + struct string_mapping
> + {
> + E flag;
> + const char *str;
> + };
> +
> + /* Convenience for to_string implementations, to build a
> + string_mapping array. */
> +#define MAP_ENUM_FLAG(ENUM_FLAG) { ENUM_FLAG, #ENUM_FLAG }
> +
> public:
> /* Allow default construction. */
> constexpr enum_flags ()
> @@ -183,6 +194,52 @@ class enum_flags
> /* Binary operations involving some unrelated type (which would be a
> bug) are implemented as non-members, and deleted. */
>
> + /* Convert this object to a std::string, using MAPPING as
> + enumerator-to-string mapping array. This is not meant to be
> + called directly. Instead, enum_flags specializations should have
> + their own to_string function wrapping this one, thus hidding the
> + mapping array from callers. */
> +
> + template<size_t N>
> + std::string
> + to_string (const string_mapping (&mapping)[N]) const
Just wondering, any reason why you chose a template rather than, say,
gdb::array_view<const string_mapping>?
Once the undefined behavior thing is fixed, this patch LGTM:
Approved-By: Simon Marchi <simon.marchi@efficios.com>
Simon
next prev parent reply other threads:[~2022-10-28 15:59 UTC|newest]
Thread overview: 72+ messages / expand[flat|nested] mbox.gz Atom feed top
2022-07-13 22:24 [PATCH v2 00/29] Step over thread clone and thread exit Pedro Alves
2022-07-13 22:24 ` [PATCH v2 01/29] displaced step: pass down target_waitstatus instead of gdb_signal Pedro Alves
2022-07-20 20:21 ` Simon Marchi
2022-07-13 22:24 ` [PATCH v2 02/29] linux-nat: introduce pending_status_str Pedro Alves
2022-07-20 20:38 ` Simon Marchi
2022-07-13 22:24 ` [PATCH v2 03/29] gdb/linux: Delete all other LWPs immediately on ptrace exec event Pedro Alves
2022-07-21 0:45 ` Simon Marchi
2022-07-13 22:24 ` [PATCH v2 04/29] Step over clone syscall w/ breakpoint, TARGET_WAITKIND_THREAD_CLONED Pedro Alves
2022-07-21 1:35 ` Simon Marchi
2022-10-17 18:54 ` Pedro Alves
2022-10-18 12:40 ` [PATCH] Don't explicitly set clone child ptrace options (was: Re: [PATCH v2 04/29] Step over clone syscall w/ breakpoint, TARGET_WAITKIND_THREAD_CLONED) Pedro Alves
2022-10-28 14:50 ` Simon Marchi
2022-12-12 20:13 ` [PATCH v2 04/29] Step over clone syscall w/ breakpoint, TARGET_WAITKIND_THREAD_CLONED Pedro Alves
2022-07-13 22:24 ` [PATCH v2 05/29] Support clone events in the remote protocol Pedro Alves
2022-07-21 2:25 ` Simon Marchi
2022-12-12 20:14 ` Pedro Alves
2022-07-13 22:24 ` [PATCH v2 06/29] Avoid duplicate QThreadEvents packets Pedro Alves
2022-07-21 2:30 ` Simon Marchi
2022-12-12 20:14 ` Pedro Alves
2022-07-13 22:24 ` [PATCH v2 07/29] Thread options & clone events (core + remote) Pedro Alves
2022-07-21 3:14 ` Simon Marchi
2022-10-27 19:55 ` [PATCH] enum_flags to_string (was: Re: [PATCH v2 07/29] Thread options & clone events (core + remote)) Pedro Alves
2022-10-28 10:26 ` [PATCH v2] " Pedro Alves
2022-10-28 11:08 ` [PATCH v3] " Pedro Alves
2022-10-28 15:59 ` Simon Marchi [this message]
2022-10-28 18:23 ` Pedro Alves
2022-10-31 12:47 ` Simon Marchi
2022-11-07 17:26 ` [PATCH v5] " Pedro Alves
2022-11-07 18:29 ` Simon Marchi
2022-11-08 14:56 ` Pedro Alves
2022-12-12 20:15 ` [PATCH v2 07/29] Thread options & clone events (core + remote) Pedro Alves
2022-07-13 22:24 ` [PATCH v2 08/29] Thread options & clone events (native Linux) Pedro Alves
2022-07-21 12:38 ` Simon Marchi
2022-12-12 20:16 ` Pedro Alves
2022-07-13 22:24 ` [PATCH v2 09/29] Thread options & clone events (Linux GDBserver) Pedro Alves
2022-07-21 13:11 ` Simon Marchi
2022-12-12 20:16 ` Pedro Alves
2022-07-13 22:24 ` [PATCH v2 10/29] gdbserver: Hide and don't detach pending clone children Pedro Alves
2022-07-13 22:24 ` [PATCH v2 11/29] Remove gdb/19675 kfails (displaced stepping + clone) Pedro Alves
2022-07-13 22:24 ` [PATCH v2 12/29] Add test for stepping over clone syscall Pedro Alves
2022-07-13 22:24 ` [PATCH v2 13/29] all-stop/synchronous RSP support thread-exit events Pedro Alves
2022-07-13 22:24 ` [PATCH v2 14/29] gdbserver/linux-low.cc: Ignore event_ptid if TARGET_WAITKIND_IGNORE Pedro Alves
2022-07-13 22:24 ` [PATCH v2 15/29] Introduce GDB_TO_EXIT thread option, fix step-over-thread-exit Pedro Alves
2022-07-13 22:24 ` [PATCH v2 16/29] Implement GDB_TO_EXIT support for Linux GDBserver Pedro Alves
2022-07-13 22:24 ` [PATCH v2 17/29] Implement GDB_TO_EXIT support for native Linux Pedro Alves
2022-07-21 15:26 ` Simon Marchi
2022-12-12 20:17 ` Pedro Alves
2022-07-13 22:24 ` [PATCH v2 18/29] gdb: clear step over information on thread exit (PR gdb/27338) Pedro Alves
2022-07-21 18:12 ` Simon Marchi
2022-12-12 20:18 ` Pedro Alves
2022-07-13 22:24 ` [PATCH v2 19/29] stop_all_threads: (re-)enable async before waiting for stops Pedro Alves
2022-07-21 18:21 ` Simon Marchi
2022-12-12 20:18 ` Pedro Alves
2022-07-13 22:24 ` [PATCH v2 20/29] gdbserver: Queue no-resumed event after thread exit Pedro Alves
2022-07-13 22:24 ` [PATCH v2 21/29] Don't resume new threads if scheduler-locking is in effect Pedro Alves
2022-07-14 5:28 ` Eli Zaretskii
2022-07-21 18:49 ` Simon Marchi
2022-12-12 20:19 ` Pedro Alves
2022-07-13 22:24 ` [PATCH v2 22/29] Report thread exit event for leader if reporting thread exit events Pedro Alves
2022-07-13 22:24 ` [PATCH v2 23/29] Ignore failure to read PC when resuming Pedro Alves
2022-07-13 22:24 ` [PATCH v2 24/29] gdb/testsuite/lib/my-syscalls.S: Refactor new SYSCALL macro Pedro Alves
2022-07-13 22:24 ` [PATCH v2 25/29] Testcases for stepping over thread exit syscall (PR gdb/27338) Pedro Alves
2022-07-13 22:24 ` [PATCH v2 26/29] Document remote clone events, and QThreadOptions packet Pedro Alves
2022-07-14 5:27 ` Eli Zaretskii
2022-07-13 22:24 ` [PATCH v2 27/29] inferior::clear_thread_list always silent Pedro Alves
2022-07-13 22:24 ` [PATCH v2 28/29] Centralize "[Thread ...exited]" notifications Pedro Alves
2022-07-13 22:24 ` [PATCH v2 29/29] Cancel execution command on thread exit, when stepping, nexting, etc Pedro Alves
2022-07-21 19:28 ` [PATCH v2 00/29] Step over thread clone and thread exit Simon Marchi
2022-10-03 13:46 ` Tom de Vries
2022-10-03 18:37 ` Tom de Vries
2022-12-12 20:20 ` Pedro Alves
2022-12-12 20:19 ` Pedro Alves
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=6aee3c00-b334-97e2-62f7-0434822bcf7a@simark.ca \
--to=simark@simark.ca \
--cc=gdb-patches@sourceware.org \
--cc=pedro@palves.net \
/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;
as well as URLs for read-only IMAP folder(s) and NNTP newsgroup(s).