On Fri, 29 Aug 2025 at 14:31, Tomasz Kaminski <tkami...@redhat.com> wrote:
>
>
>
> On Fri, Aug 29, 2025 at 2:06 AM Nathan Myers <n...@cantrip.org> wrote:
>>
>> Changes in v2:
>> * Make a global function __syncbuf_get_mutex, not an extern template
>> instantiation.
>>
>> This patch creates a global function __syncbuf_get_mutex, gated by
>> _GLIBCXX_HAS_GTHREADS, replacing a static instantiated member
>> _S_get_mutex used in syncbuf<> construction, and makes the global
>> symbol visible. A static local table of 16 mutexes is shared among
>> all specializations of syncbuf<>, chosen on construction by a hash
>> of the wrapped streambuf's address.
>>
>> It detaches the implementation of _S_get_mutex from the C++20 ABI.
>
> Wouldn't it be more extensible, if we always would define _S_get_mutex
Do you mean when ! HAS_GTHREADS? Is that useful?
>>
>>
>> libstdc++-v3/ChangeLog:
>> * include/std/syncstream: (syncbuf<>::__mutex) Remove _S_get_mutex,
>> use extern function instead.
>> * src/c++20/syncbuf.cc: Define global __syncbuf_get_mutex.
>> * src/c++20/Makefile.am: Mention syncbuf.cc.
>> * src/c++20/Makefile.in: Regenerate.
>> * config/abi/pre/gnu.ver: Mention mangled __syncbuf_get_mutex.
>> ---
>> libstdc++-v3/config/abi/pre/gnu.ver | 3 ++
>> libstdc++-v3/include/std/syncstream | 19 ++++--------
>> libstdc++-v3/src/c++20/Makefile.am | 2 +-
>> libstdc++-v3/src/c++20/Makefile.in | 2 +-
>> libstdc++-v3/src/c++20/syncbuf.cc | 45 +++++++++++++++++++++++++++++
>> 5 files changed, 56 insertions(+), 15 deletions(-)
>> create mode 100644 libstdc++-v3/src/c++20/syncbuf.cc
>>
>> diff --git a/libstdc++-v3/config/abi/pre/gnu.ver
>> b/libstdc++-v3/config/abi/pre/gnu.ver
>> index e1601dc39d2..2e48241d51f 100644
>> --- a/libstdc++-v3/config/abi/pre/gnu.ver
>> +++ b/libstdc++-v3/config/abi/pre/gnu.ver
>> @@ -2559,6 +2559,9 @@ GLIBCXX_3.4.35 {
>> _ZNSt6chrono9gps_clock3nowEv;
>> _ZNSt6chrono9tai_clock3nowEv;
>>
>> + # mutex& __syncbuf_get_mutex(void*)
>> + _ZSt19__syncbuf_get_mutexPv;
>> +
>> # __gnu_debug::_Safe_iterator_base and _Safe_sequence_base const
>>
>> _ZN11__gnu_debug19_Safe_iterator_base9_M_attachEPKNS_19_Safe_sequence_baseEb;
>>
>> _ZN11__gnu_debug19_Safe_iterator_base16_M_attach_singleEPKNS_19_Safe_sequence_baseEb;
>> diff --git a/libstdc++-v3/include/std/syncstream
>> b/libstdc++-v3/include/std/syncstream
>> index e2b5a199ec9..649756b2581 100644
>> --- a/libstdc++-v3/include/std/syncstream
>> +++ b/libstdc++-v3/include/std/syncstream
>> @@ -46,13 +46,17 @@
>> #include <bits/alloc_traits.h>
>> #include <bits/allocator.h>
>> #include <bits/functexcept.h>
>> -#include <bits/functional_hash.h>
>> #include <bits/std_mutex.h>
>>
>> namespace std _GLIBCXX_VISIBILITY(default)
>> {
>> _GLIBCXX_BEGIN_NAMESPACE_VERSION
>>
>> +#if _GLIBCXX_HAS_GTHREADS
>> + extern mutex&
>> + __syncbuf_get_mutex(void*); // in src/c++20/syncbuf.cc
>> +#endif
>> +
>> template<typename _CharT, typename _Traits, typename _Alloc>
>> class basic_syncbuf : public __syncbuf_base<_CharT, _Traits>
>> {
>> @@ -202,7 +206,7 @@ _GLIBCXX_BEGIN_NAMESPACE_VERSION
>> mutex* _M_mtx;
>>
>> __mutex(void* __t)
>> - : _M_mtx(__t ? &_S_get_mutex(__t) : nullptr)
>> + : _M_mtx(__t ? &__syncbuf_get_mutex(__t) : nullptr)
>> { }
>
> You could put the declaration of the __syncbuf_get_mutex function inside the
> constructor,
> to avoid having a separate _GLIBCXX_HAS_GTHREADS block.
> __mutex(void* __t)
> : _M_mtx(nullptr)
> {
> extern mutex& __syncbuf_get_mutex(void*); // in
> src/c++20/syncbuf.cc
> if (__t)
> _M_mtx = &_S_get_mutex(__t);
> }
> But let's see what other think.
Yes, we do that in a few other places, so that there's no
user-accessible declaration of the function.
