The (bool) cast wrt callable fiasco aside for a moment. This version has
the fences exactly where they need to be. Using C++ CAS with its membar
A for true and membar B for fail is RIPE for bugs anyway, well, imvvho:


#pragma once
#include <concepts>
#include <atomic>

struct xonce_flag
{
xonce_flag() noexcept = default;
private:
friend bool xcall_once( xonce_flag &, std::invocable auto );
using flag_t = std::atomic<signed char>;
flag_t m_flag = 0;
};

bool xcall_once( xonce_flag &xflag, std::invocable auto callable )
{
using namespace std;
xonce_flag::flag_t &flag = xflag.m_flag;
for( signed char ref = flag.load( memory_order_relaxed ); ; )
if( ref > 0 ) [[likely]]
{

std::atomic_thread_fence(std::memory_order_acquire);

return true;
}
else if( ref < 0 ) [[unlikely]]
{
flag.wait( ref, memory_order_relaxed );
ref = flag.load( memory_order_relaxed );
}
else if( flag.compare_exchange_strong( ref, -1,
memory_order_relaxed, memory_order_relaxed ) ) [[likely]]
break;
bool succ = true;

std::atomic_thread_fence(std::memory_order_acquire);

try
{
if constexpr( requires { (bool)callable(); } )
succ = (bool)callable();
else
callable();
}
catch( ... )
{
std::atomic_thread_fence(std::memory_order_release);

flag.store( 0, memory_order_relaxed );
flag.notify_one();
throw;
}

std::atomic_thread_fence(std::memory_order_release);

flag.store( (char)succ, memory_order_relaxed );
if( succ )
flag.notify_all();
else
flag.notify_one();
return succ;
}


--- PyGate Linux v1.5.2
* Origin: Dragon's Lair, PyGate NNTP<>Fido Gate (3:633/10)

You're making my code more complicated for nothing.

Am 03.01.2026 um 04:59 schrieb Chris M. Thomasson:

The (bool) cast wrt callable fiasco aside for a moment. This version
has the fences exactly where they need to be. Using C++ CAS with its
membar A for true and membar B for fail is RIPE for bugs anyway, well, imvvho:

#pragma once
#include <concepts>
#include <atomic>

struct xonce_flag
{
ÿÿÿ xonce_flag() noexcept = default;
private:
ÿÿÿ friend bool xcall_once( xonce_flag &, std::invocable auto );
ÿÿÿ using flag_t = std::atomic<signed char>;
ÿÿÿ flag_t m_flag = 0;
};

bool xcall_once( xonce_flag &xflag, std::invocable auto callable )
{
ÿÿÿ using namespace std;
ÿÿÿ xonce_flag::flag_t &flag = xflag.m_flag;
ÿÿÿ for( signed char ref = flag.load( memory_order_relaxed ); ; )
ÿÿÿÿÿÿÿ if( ref > 0 ) [[likely]]
ÿÿÿÿÿÿÿ {

ÿÿÿÿÿÿÿÿÿÿÿ std::atomic_thread_fence(std::memory_order_acquire);

ÿÿÿÿÿÿÿÿÿÿÿ return true;
ÿÿÿÿÿÿÿ }
ÿÿÿÿÿÿÿ else if( ref < 0 ) [[unlikely]]
ÿÿÿÿÿÿÿ {
ÿÿÿÿÿÿÿÿÿÿÿ flag.wait( ref, memory_order_relaxed );
ÿÿÿÿÿÿÿÿÿÿÿ ref = flag.load( memory_order_relaxed );
ÿÿÿÿÿÿÿ }
ÿÿÿÿÿÿÿ else if( flag.compare_exchange_strong( ref, -1, memory_order_relaxed, memory_order_relaxed ) ) [[likely]]
ÿÿÿÿÿÿÿÿÿÿÿ break;
ÿÿÿ bool succ = true;

ÿÿÿ std::atomic_thread_fence(std::memory_order_acquire);

ÿÿÿ try
ÿÿÿ {
ÿÿÿÿÿÿÿ if constexpr( requires { (bool)callable(); } )
ÿÿÿÿÿÿÿÿÿÿÿ succ = (bool)callable();
ÿÿÿÿÿÿÿ else
ÿÿÿÿÿÿÿÿÿÿÿ callable();
ÿÿÿ }
ÿÿÿ catch( ... )
ÿÿÿ {
ÿÿÿÿÿÿÿ std::atomic_thread_fence(std::memory_order_release);

ÿÿÿÿÿÿÿ flag.store( 0, memory_order_relaxed );
ÿÿÿÿÿÿÿ flag.notify_one();
ÿÿÿÿÿÿÿ throw;
ÿÿÿ }

ÿÿÿ std::atomic_thread_fence(std::memory_order_release);

ÿÿÿ flag.store( (char)succ, memory_order_relaxed );
ÿÿÿ if( succ )
ÿÿÿÿÿÿÿ flag.notify_all();
ÿÿÿ else
ÿÿÿÿÿÿÿ flag.notify_one();
ÿÿÿ return succ;
}

--- PyGate Linux v1.5.2
* Origin: Dragon's Lair, PyGate NNTP<>Fido Gate (3:633/10)

On 1/2/2026 8:02 PM, Bonita Montero wrote:

You're making my code more complicated for nothing.

I think it makes it MUCH easier to read, and performs just as well. The membars are exactly right where they need to be.

Am 03.01.2026 um 04:59 schrieb Chris M. Thomasson:

The (bool) cast wrt callable fiasco aside for a moment. This version
has the fences exactly where they need to be. Using C++ CAS with its
membar A for true and membar B for fail is RIPE for bugs anyway, well,
imvvho:


#pragma once
#include <concepts>
#include <atomic>

struct xonce_flag
{
ÿÿÿ xonce_flag() noexcept = default;
private:
ÿÿÿ friend bool xcall_once( xonce_flag &, std::invocable auto );
ÿÿÿ using flag_t = std::atomic<signed char>;
ÿÿÿ flag_t m_flag = 0;
};

bool xcall_once( xonce_flag &xflag, std::invocable auto callable )
{
ÿÿÿ using namespace std;
ÿÿÿ xonce_flag::flag_t &flag = xflag.m_flag;
ÿÿÿ for( signed char ref = flag.load( memory_order_relaxed ); ; )
ÿÿÿÿÿÿÿ if( ref > 0 ) [[likely]]
ÿÿÿÿÿÿÿ {

ÿÿÿÿÿÿÿÿÿÿÿ std::atomic_thread_fence(std::memory_order_acquire);

ÿÿÿÿÿÿÿÿÿÿÿ return true;
ÿÿÿÿÿÿÿ }
ÿÿÿÿÿÿÿ else if( ref < 0 ) [[unlikely]]
ÿÿÿÿÿÿÿ {
ÿÿÿÿÿÿÿÿÿÿÿ flag.wait( ref, memory_order_relaxed );
ÿÿÿÿÿÿÿÿÿÿÿ ref = flag.load( memory_order_relaxed );
ÿÿÿÿÿÿÿ }
ÿÿÿÿÿÿÿ else if( flag.compare_exchange_strong( ref, -1,
memory_order_relaxed, memory_order_relaxed ) ) [[likely]]
ÿÿÿÿÿÿÿÿÿÿÿ break;
ÿÿÿ bool succ = true;

ÿÿÿ std::atomic_thread_fence(std::memory_order_acquire);

ÿÿÿ try
ÿÿÿ {
ÿÿÿÿÿÿÿ if constexpr( requires { (bool)callable(); } )
ÿÿÿÿÿÿÿÿÿÿÿ succ = (bool)callable();
ÿÿÿÿÿÿÿ else
ÿÿÿÿÿÿÿÿÿÿÿ callable();
ÿÿÿ }
ÿÿÿ catch( ... )
ÿÿÿ {
ÿÿÿÿÿÿÿ std::atomic_thread_fence(std::memory_order_release);

ÿÿÿÿÿÿÿ flag.store( 0, memory_order_relaxed );
ÿÿÿÿÿÿÿ flag.notify_one();
ÿÿÿÿÿÿÿ throw;
ÿÿÿ }

ÿÿÿ std::atomic_thread_fence(std::memory_order_release);

ÿÿÿ flag.store( (char)succ, memory_order_relaxed );
ÿÿÿ if( succ )
ÿÿÿÿÿÿÿ flag.notify_all();
ÿÿÿ else
ÿÿÿÿÿÿÿ flag.notify_one();
ÿÿÿ return succ;
}

--- PyGate Linux v1.5.2
* Origin: Dragon's Lair, PyGate NNTP<>Fido Gate (3:633/10)

Am 03.01.2026 um 05:09 schrieb Chris M. Thomasson:

On 1/2/2026 8:02 PM, Bonita Montero wrote:

You're making my code more complicated for nothing.

I think it makes it MUCH easier to read, and performs just as well.
The membars are exactly right where they need to be.

You're crazy.


--- PyGate Linux v1.5.2
* Origin: Dragon's Lair, PyGate NNTP<>Fido Gate (3:633/10)

On 1/2/2026 8:18 PM, Bonita Montero wrote:

Am 03.01.2026 um 05:09 schrieb Chris M. Thomasson:

On 1/2/2026 8:02 PM, Bonita Montero wrote:

You're making my code more complicated for nothing.

I think it makes it MUCH easier to read, and performs just as well.
The membars are exactly right where they need to be.

You're crazy.

Na. (bool)callable, well, on the other hand... Shit happens.

--- PyGate Linux v1.5.2
* Origin: Dragon's Lair, PyGate NNTP<>Fido Gate (3:633/10)

Am 03.01.2026 um 20:43 schrieb Chris M. Thomasson:

On 1/2/2026 8:18 PM, Bonita Montero wrote:

Am 03.01.2026 um 05:09 schrieb Chris M. Thomasson:

On 1/2/2026 8:02 PM, Bonita Montero wrote:

You're making my code more complicated for nothing.

I think it makes it MUCH easier to read, and performs just as well.
The membars are exactly right where they need to be.

You're crazy.

Na. (bool)callable, well, on the other hand... Shit happens.

That's while I intented this new kind of "once_flag"; to make
*that* possible. I can't understand any objections against
that; the interface is rather simple.
The callback is concept'ed with std::invocable auto, i.e it
must be callable without parameters; that makes the inter-
face even easier to understand.


--- PyGate Linux v1.5.2
* Origin: Dragon's Lair, PyGate NNTP<>Fido Gate (3:633/10)

On 1/3/2026 11:14 PM, Bonita Montero wrote:

Am 03.01.2026 um 20:43 schrieb Chris M. Thomasson:

On 1/2/2026 8:18 PM, Bonita Montero wrote:

Am 03.01.2026 um 05:09 schrieb Chris M. Thomasson:

On 1/2/2026 8:02 PM, Bonita Montero wrote:

You're making my code more complicated for nothing.

I think it makes it MUCH easier to read, and performs just as well.
The membars are exactly right where they need to be.

You're crazy.

Na. (bool)callable, well, on the other hand... Shit happens.

That's while I intented this new kind of "once_flag"; to make
*that* possible. I can't understand any objections against
that; the interface is rather simple.
The callback is concept'ed with std::invocable auto, i.e it
must be callable without parameters; that makes the inter-
face even easier to understand.

You need to document it, before anybody needs to read the code to find out.

--- PyGate Linux v1.5.2
* Origin: Dragon's Lair, PyGate NNTP<>Fido Gate (3:633/10)

Am 04.01.2026 um 23:21 schrieb Chris M. Thomasson:

That's while I intented this new kind of "once_flag"; to make
*that* possible. I can't understand any objections against
that; the interface is rather simple.
The callback is concept'ed with std::invocable auto, i.e it
must be callable without parameters; that makes the inter-
face even easier to understand.

You need to document it, before anybody needs to read the code to find
out.

It's only one line of code to find that out (if constexpr( ... )).



--- PyGate Linux v1.5.2
* Origin: Dragon's Lair, PyGate NNTP<>Fido Gate (3:633/10)

On 1/4/2026 3:59 PM, Bonita Montero wrote:

Am 04.01.2026 um 23:21 schrieb Chris M. Thomasson:

That's while I intented this new kind of "once_flag"; to make
*that* possible. I can't understand any objections against
that; the interface is rather simple.
The callback is concept'ed with std::invocable auto, i.e it
must be callable without parameters; that makes the inter-
face even easier to understand.

You need to document it, before anybody needs to read the code to find
out.

It's only one line of code to find that out (if constexpr( ... )).

But, I mean for somebody who might want to use it. Why not document it
up front?

--- PyGate Linux v1.5.2
* Origin: Dragon's Lair, PyGate NNTP<>Fido Gate (3:633/10)

On 1/4/2026 3:59 PM, Bonita Montero wrote:

Am 04.01.2026 um 23:21 schrieb Chris M. Thomasson:

That's while I intented this new kind of "once_flag"; to make
*that* possible. I can't understand any objections against
that; the interface is rather simple.
The callback is concept'ed with std::invocable auto, i.e it
must be callable without parameters; that makes the inter-
face even easier to understand.

You need to document it, before anybody needs to read the code to find
out.

It's only one line of code to find that out (if constexpr( ... )).

Oh no. If somebody wants to use it... I can see it now. The user asks so
what are the rules of callable? You say read the code. That's like the opengroup saying read the code for a mutex impl to find out how it
works. We don't need to document anything, just read the impl code.
Common man!

--- PyGate Linux v1.5.2
* Origin: Dragon's Lair, PyGate NNTP<>Fido Gate (3:633/10)

Am 05.01.2026 um 01:02 schrieb Chris M. Thomasson:

Oh no. If somebody wants to use it... I can see it now. The user asks
so what are the rules of callable? You say read the code. That's like
the opengroup saying read the code for a mutex impl to find out how it works. We don't need to document anything, just read the impl code.
Common man!

The code is easy to read.


--- PyGate Linux v1.5.2
* Origin: Dragon's Lair, PyGate NNTP<>Fido Gate (3:633/10)

On 1/4/2026 4:04 PM, Bonita Montero wrote:

Am 05.01.2026 um 01:02 schrieb Chris M. Thomasson:

Oh no. If somebody wants to use it... I can see it now. The user asks
so what are the rules of callable? You say read the code. That's like
the opengroup saying read the code for a mutex impl to find out how it
works. We don't need to document anything, just read the impl code.
Common man!

The code is easy to read.

I can read it for sure. Yes I can understand it for sure. But, that (bool)callable() simply needs a clearly documented point. Fair enough?

--- PyGate Linux v1.5.2
* Origin: Dragon's Lair, PyGate NNTP<>Fido Gate (3:633/10)

Am 05.01.2026 um 03:10 schrieb Chris M. Thomasson:

On 1/4/2026 4:04 PM, Bonita Montero wrote:

Am 05.01.2026 um 01:02 schrieb Chris M. Thomasson:

Oh no. If somebody wants to use it... I can see it now. The user
asks so what are the rules of callable? You say read the code.
That's like the opengroup saying read the code for a mutex impl to
find out how it works. We don't need to document anything, just read
the impl code. Common man!

The code is easy to read.

I can read it for sure. Yes I can understand it for sure. But, that (bool)callable() simply needs a clearly documented point. Fair enough?

Sorry, if you understand the rest this part is rather easy to read
if you can read concept'ed code.



--- PyGate Linux v1.5.2
* Origin: Dragon's Lair, PyGate NNTP<>Fido Gate (3:633/10)

On 1/5/2026 10:39 AM, Bonita Montero wrote:

Am 05.01.2026 um 03:10 schrieb Chris M. Thomasson:

On 1/4/2026 4:04 PM, Bonita Montero wrote:

Am 05.01.2026 um 01:02 schrieb Chris M. Thomasson:

Oh no. If somebody wants to use it... I can see it now. The user
asks so what are the rules of callable? You say read the code.
That's like the opengroup saying read the code for a mutex impl to
find out how it works. We don't need to document anything, just read
the impl code. Common man!

The code is easy to read.

I can read it for sure. Yes I can understand it for sure. But, that
(bool)callable() simply needs a clearly documented point. Fair enough?

Sorry, if you understand the rest this part is rather easy to read
if you can read concept'ed code.

Have you ever created documentation about your code before? I can
imagine you creating a doc on, say, opengroup for a mutex:

https://pubs.opengroup.org/onlinepubs/9699919799/functions/pthread_mutex_lock.html

You say read the code, instead of providing a detailed explanation.

--- PyGate Linux v1.5.2
* Origin: Dragon's Lair, PyGate NNTP<>Fido Gate (3:633/10)

Am 05.01.2026 um 21:34 schrieb Chris M. Thomasson:

Have you ever created documentation about your code before? I can
imagine you creating a doc on, say, opengroup for a mutex: https://pubs.opengroup.org/onlinepubs/9699919799/functions/pthread_mutex_lock.html

You say read the code, instead of providing a detailed explanation.

I use Doxygen sometimes.
But you think you're an expert in parallel programming.
Then you should understand the code easily.
It's only 34 lines of code.


--- PyGate Linux v1.5.2
* Origin: Dragon's Lair, PyGate NNTP<>Fido Gate (3:633/10)

On 1/5/2026 9:07 PM, Bonita Montero wrote:

Am 05.01.2026 um 21:34 schrieb Chris M. Thomasson:

Have you ever created documentation about your code before? I can
imagine you creating a doc on, say, opengroup for a mutex:
https://pubs.opengroup.org/onlinepubs/9699919799/functions/
pthread_mutex_lock.html
You say read the code, instead of providing a detailed explanation.

I use Doxygen sometimes.
But you think you're an expert in parallel programming.

I have had to correct you in the past. Remember?

Then you should understand the code easily.
It's only 34 lines of code.

I do understand it! I even know how to make it use standalone fences.
The (bool)callable raised some red flags. If you doc that clearly,
well... Then, why not use it?

--- PyGate Linux v1.5.2
* Origin: Dragon's Lair, PyGate NNTP<>Fido Gate (3:633/10)

On 1/6/2026 4:56 PM, Chris M. Thomasson wrote:

On 1/5/2026 9:07 PM, Bonita Montero wrote:

Am 05.01.2026 um 21:34 schrieb Chris M. Thomasson:

Have you ever created documentation about your code before? I can
imagine you creating a doc on, say, opengroup for a mutex:
https://pubs.opengroup.org/onlinepubs/9699919799/functions/
pthread_mutex_lock.html
You say read the code, instead of providing a detailed explanation.

I use Doxygen sometimes.
But you think you're an expert in parallel programming.

I have had to correct you in the past. Remember?

It was fun when Sun gave me a T2000 sunfire before it was available for purchase by the general public.

Then you should understand the code easily.
It's only 34 lines of code.

I do understand it! I even know how to make it use standalone fences.
The (bool)callable raised some red flags. If you doc that clearly,
well... Then, why not use it?

--- PyGate Linux v1.5.2
* Origin: Dragon's Lair, PyGate NNTP<>Fido Gate (3:633/10)

Am 07.01.2026 um 01:56 schrieb Chris M. Thomasson:

On 1/5/2026 9:07 PM, Bonita Montero wrote:

Am 05.01.2026 um 21:34 schrieb Chris M. Thomasson:

Have you ever created documentation about your code before? I can
imagine you creating a doc on, say, opengroup for a mutex:
https://pubs.opengroup.org/onlinepubs/9699919799/functions/
pthread_mutex_lock.html
You say read the code, instead of providing a detailed explanation.

I use Doxygen sometimes.
But you think you're an expert in parallel programming.

I have had to correct you in the past. Remember?

Then you should understand the code easily.
It's only 34 lines of code.

I do understand it! I even know how to make it use standalone fences.
The (bool)callable raised some red flags. If you doc that clearly,
well... Then, why not use it?

Clearly documented ...; dude that code is 34 only !


--- PyGate Linux v1.5.2
* Origin: Dragon's Lair, PyGate NNTP<>Fido Gate (3:633/10)

Am 07.01.2026 um 06:58 schrieb Bonita Montero:

Clearly documented ...; dude that code is 34 only

This is the unit-test:

#include <iostream>
#include <thread>
#include <vector>
#include <optional>
#include <atomic>
#include <barrier>
#include "xonce.h"

using namespace std;

int main()
{
ÿ ÿ constexpr int N_THREADS = 10;
ÿ ÿ atomic_int nTh = N_THREADS;
ÿ ÿ optional<xonce_flag> flag;
ÿ ÿ barrier bar( N_THREADS );
ÿ ÿ vector<jthread> threads;
ÿ ÿ atomic_int ctr = 0;
ÿ ÿ for( int t = N_THREADS; t--; )
ÿ ÿ ÿ ÿ threads.emplace_back( [&]
ÿ ÿ ÿ ÿ ÿ ÿ {
ÿ ÿ ÿ ÿ ÿ ÿ ÿ ÿ for( size_t r = 1'000'000; r; --r )
ÿ ÿ ÿ ÿ ÿ ÿ ÿ ÿ {
ÿ ÿ ÿ ÿ ÿ ÿ ÿ ÿ ÿ ÿ if( !--nTh )
ÿ ÿ ÿ ÿ ÿ ÿ ÿ ÿ ÿ ÿ {
ÿ ÿ ÿ ÿ ÿ ÿ ÿ ÿ ÿ ÿ ÿ ÿ flag.emplace();
ÿ ÿ ÿ ÿ ÿ ÿ ÿ ÿ ÿ ÿ ÿ ÿ nTh = N_THREADS;
ÿ ÿ ÿ ÿ ÿ ÿ ÿ ÿ ÿ ÿ }
ÿ ÿ ÿ ÿ ÿ ÿ ÿ ÿ ÿ ÿ bar.arrive_and_wait();
ÿ ÿ ÿ ÿ ÿ ÿ ÿ ÿ ÿ ÿ xcall_once( *flag, [&ctr] { ++ctr; } );
ÿ ÿ ÿ ÿ ÿ ÿ ÿ ÿ }
ÿ ÿ ÿ ÿ ÿ ÿ } );
ÿ ÿ threads.resize( 0 );
ÿ ÿ cout << ctr << endl;
}


--- PyGate Linux v1.5.2
* Origin: Dragon's Lair, PyGate NNTP<>Fido Gate (3:633/10)

On 1/6/2026 9:58 PM, Bonita Montero wrote:

Am 07.01.2026 um 01:56 schrieb Chris M. Thomasson:

On 1/5/2026 9:07 PM, Bonita Montero wrote:

Am 05.01.2026 um 21:34 schrieb Chris M. Thomasson:

Have you ever created documentation about your code before? I can
imagine you creating a doc on, say, opengroup for a mutex:
https://pubs.opengroup.org/onlinepubs/9699919799/functions/
pthread_mutex_lock.html
You say read the code, instead of providing a detailed explanation.

I use Doxygen sometimes.
But you think you're an expert in parallel programming.

I have had to correct you in the past. Remember?

Then you should understand the code easily.
It's only 34 lines of code.

I do understand it! I even know how to make it use standalone fences.
The (bool)callable raised some red flags. If you doc that clearly,
well... Then, why not use it?

Clearly documented ...; dude that code is 34 only !

I know the code is as is. But, well, what if somebody wants to use it as
a primitive, and wants to read a spec, instead of the code?

--- PyGate Linux v1.5.2
* Origin: Dragon's Lair, PyGate NNTP<>Fido Gate (3:633/10)

Am 07.01.2026 um 07:15 schrieb Chris M. Thomasson:

I know the code is as is. But, well, what if somebody wants to use it
as a primitive, and wants to read a spec, instead of the code?

The code doesn't need to be documented but the interface.


--- PyGate Linux v1.5.2
* Origin: Dragon's Lair, PyGate NNTP<>Fido Gate (3:633/10)

On 1/6/2026 10:45 PM, Bonita Montero wrote:

Am 07.01.2026 um 07:15 schrieb Chris M. Thomasson:

I know the code is as is. But, well, what if somebody wants to use it
as a primitive, and wants to read a spec, instead of the code?

The code doesn't need to be documented but the interface.

Right, and (bool)callable rules/spec needs to be in the docs.

--- PyGate Linux v1.5.2
* Origin: Dragon's Lair, PyGate NNTP<>Fido Gate (3:633/10)