diff mbox

libstdc++: Support std::is_aggregate on clang++ (was [cfe-dev] clang++: std::is_aggregate unusable with clang-5.0/libstdc++-7)

Message ID 20170727072715.43fqiep4nawwvzlc@netzach.ktns.kdns.info
State New
Headers show

Commit Message

Katsuhiko Nishimra July 27, 2017, 7:27 a.m. UTC
From 56c4a18d0d8c8ce7aa1239880138775e4db06645 Mon Sep 17 00:00:00 2001
From: Katsuhiko Nishimra <ktns.87@gmail.com>
Date: Thu, 27 Jul 2017 16:03:54 +0900
Subject: [PATCH] libstdc++: Support std::is_aggregate on clang++

Currently, libstdc++ tries to detect __is_aggregate built-in macro using
__has_builtin, but this fails on clang++ because __has_builtin on
clang++ detects only built-in functions, not built-in macros. This patch
adds a test using __is_identifier. Tested on clang++
5.0.0-svn308422-1~exp1 and g++ 7.1.0-10 from Debian unstable.
---
 libstdc++-v3/include/std/type_traits | 5 +++++
 1 file changed, 5 insertions(+)

Comments

Jonathan Wakely July 31, 2017, 2:53 p.m. UTC | #1
On 27/07/17 16:27 +0900, Katsuhiko Nishimra wrote:
>From 56c4a18d0d8c8ce7aa1239880138775e4db06645 Mon Sep 17 00:00:00 2001
>From: Katsuhiko Nishimra <ktns.87@gmail.com>
>Date: Thu, 27 Jul 2017 16:03:54 +0900
>Subject: [PATCH] libstdc++: Support std::is_aggregate on clang++
>
>Currently, libstdc++ tries to detect __is_aggregate built-in macro using
>__has_builtin, but this fails on clang++ because __has_builtin on
>clang++ detects only built-in functions, not built-in macros. This patch
>adds a test using __is_identifier. Tested on clang++
>5.0.0-svn308422-1~exp1 and g++ 7.1.0-10 from Debian unstable.
>---
> libstdc++-v3/include/std/type_traits | 5 +++++
> 1 file changed, 5 insertions(+)
>
>diff --git a/libstdc++-v3/include/std/type_traits b/libstdc++-v3/include/std/type_traits
>index 390b6f40a..e7ec402fb 100644
>--- a/libstdc++-v3/include/std/type_traits
>+++ b/libstdc++-v3/include/std/type_traits
>@@ -2894,6 +2894,11 @@ template <typename _From, typename _To>
>
> #if __GNUC__ >= 7
> # define _GLIBCXX_HAVE_BUILTIN_IS_AGGREGATE 1
>+#elif defined(__is_identifier)
>+// For clang
>+# if ! __is_identifier(__is_aggregate)
>+#  define _GLIBCXX_HAVE_BUILTIN_IS_AGGREGATE 1
>+# endif
> #elif defined __has_builtin
> // For non-GNU compilers:
> # if __has_builtin(__is_aggregate)

This __has_bultin check only exists for Clang, so should be replaced
by the correct __is_identifier check, not left there in addition to
it.
Tim Song July 31, 2017, 3:13 p.m. UTC | #2
On Mon, Jul 31, 2017 at 10:53 AM, Jonathan Wakely <jwakely@redhat.com> wrote:
> On 27/07/17 16:27 +0900, Katsuhiko Nishimra wrote:
>>
>> From 56c4a18d0d8c8ce7aa1239880138775e4db06645 Mon Sep 17 00:00:00 2001
>> From: Katsuhiko Nishimra <ktns.87@gmail.com>
>> Date: Thu, 27 Jul 2017 16:03:54 +0900
>> Subject: [PATCH] libstdc++: Support std::is_aggregate on clang++
>>
>> Currently, libstdc++ tries to detect __is_aggregate built-in macro using
>> __has_builtin, but this fails on clang++ because __has_builtin on
>> clang++ detects only built-in functions, not built-in macros. This patch
>> adds a test using __is_identifier. Tested on clang++
>> 5.0.0-svn308422-1~exp1 and g++ 7.1.0-10 from Debian unstable.
>> ---
>> libstdc++-v3/include/std/type_traits | 5 +++++
>> 1 file changed, 5 insertions(+)
>>
>> diff --git a/libstdc++-v3/include/std/type_traits
>> b/libstdc++-v3/include/std/type_traits
>> index 390b6f40a..e7ec402fb 100644
>> --- a/libstdc++-v3/include/std/type_traits
>> +++ b/libstdc++-v3/include/std/type_traits
>> @@ -2894,6 +2894,11 @@ template <typename _From, typename _To>
>>
>> #if __GNUC__ >= 7
>> # define _GLIBCXX_HAVE_BUILTIN_IS_AGGREGATE 1
>> +#elif defined(__is_identifier)
>> +// For clang
>> +# if ! __is_identifier(__is_aggregate)
>> +#  define _GLIBCXX_HAVE_BUILTIN_IS_AGGREGATE 1
>> +# endif
>> #elif defined __has_builtin
>> // For non-GNU compilers:
>> # if __has_builtin(__is_aggregate)
>
>
> This __has_bultin check only exists for Clang, so should be replaced
> by the correct __is_identifier check, not left there in addition to
> it.
>
>

https://clang.llvm.org/docs/LanguageExtensions.html#checks-for-type-trait-primitives
seems to suggest using __has_extension instead.
Tim Song July 31, 2017, 5:23 p.m. UTC | #3
On Mon, Jul 31, 2017 at 11:13 AM, Tim Song <t.canens.cpp@gmail.com> wrote:
>
> https://clang.llvm.org/docs/LanguageExtensions.html#checks-for-type-trait-primitives
> seems to suggest using __has_extension instead.

Hmm, that doesn't work. Oh well.
diff mbox

Patch

diff --git a/libstdc++-v3/include/std/type_traits b/libstdc++-v3/include/std/type_traits
index 390b6f40a..e7ec402fb 100644
--- a/libstdc++-v3/include/std/type_traits
+++ b/libstdc++-v3/include/std/type_traits
@@ -2894,6 +2894,11 @@  template <typename _From, typename _To>
 
 #if __GNUC__ >= 7
 # define _GLIBCXX_HAVE_BUILTIN_IS_AGGREGATE 1
+#elif defined(__is_identifier)
+// For clang
+# if ! __is_identifier(__is_aggregate)
+#  define _GLIBCXX_HAVE_BUILTIN_IS_AGGREGATE 1
+# endif
 #elif defined __has_builtin
 // For non-GNU compilers:
 # if __has_builtin(__is_aggregate)