[Thread] Can't compile package_task code with 1.54
Hi,
I just tried to compile my 1.52- based code with 1.54 beta and run into
package_task compilation issue.
In 1.52 I was using BOOST_THREAD_VERSION=3 define but with 1.54 I complied
with =4.
Test program:
#define BOOST_THREAD_VERSION 4
#include
Le 01/06/13 00:56, Szymon Gatner a écrit :
Hi,
I just tried to compile my 1.52- based code with 1.54 beta and run into package_task compilation issue.
In 1.52 I was using BOOST_THREAD_VERSION=3 define but with 1.54 I complied with =4.
Test program:
#define BOOST_THREAD_VERSION 4
#include
#include struct MyFunc { void operator()()const {} };
int main() { boost::function
f; MyFunc mf; boost::packaged_task
t1(f); // error 1 boost::packaged_task t2(mf); // error 2 } error 1 being:
d:\devel\boost_1_54_0_beta1\boost\thread\future.hpp(2848): error C2664: 'boost::detail::task_object
::task_object(boost::detail::task_object &)' : cannot convert parameter 1 from 'boost::function<Signature>' to 'boost::detail::task_object &' and error 2 similarly:
d:\devel\boost_1_54_0_beta1\boost\thread\future.hpp(2848): error C2664: 'boost::detail::task_object
::task_object(boost::detail::task_object &)' : cannot convert parameter 1 from 'MyFunc' to 'boost::detail::task_object &' (I tried with custom functor because I thought maybe error was related to boost function)
Compiling with Visual Studio 2012 Update 3
Hi, this is related to https://svn.boost.org/trac/boost/ticket/8596 which is fixed now, but maybe it doesn't works for you. Note that the trunk regression test since revision 84414 run with this compiler pass ( libs/thread/test/sync/futures/packaged_task/func_ctor_pass.cpp) https://svn.boost.org/trac/boost/browser/branches/release/libs/thread/test/s... http://www.boost.org/development/tests/trunk/developer/thread.html. Unfortunately not a single msvc tester has run after revision 84468 on release branch. I have no the possibility to reproduce it now. Please could you send the whole log? (privately if you prefer? Or via a ticket? Best, Vicente
Le 01/06/13 07:30, Vicente J. Botet Escriba a écrit :
Le 01/06/13 00:56, Szymon Gatner a écrit :
Hi,
I just tried to compile my 1.52- based code with 1.54 beta and run into package_task compilation issue.
In 1.52 I was using BOOST_THREAD_VERSION=3 define but with 1.54 I complied with =4.
Test program:
#define BOOST_THREAD_VERSION 4
#include
#include struct MyFunc { void operator()()const {} };
int main() { boost::function
f; MyFunc mf; boost::packaged_task
t1(f); // error 1 boost::packaged_task t2(mf); // error 2 } error 1 being:
d:\devel\boost_1_54_0_beta1\boost\thread\future.hpp(2848): error C2664: 'boost::detail::task_object
::task_object(boost::detail::task_object &)' : cannot convert parameter 1 from 'boost::function<Signature>' to 'boost::detail::task_object &' and error 2 similarly:
d:\devel\boost_1_54_0_beta1\boost\thread\future.hpp(2848): error C2664: 'boost::detail::task_object
::task_object(boost::detail::task_object &)' : cannot convert parameter 1 from 'MyFunc' to 'boost::detail::task_object &' (I tried with custom functor because I thought maybe error was related to boost function)
Compiling with Visual Studio 2012 Update 3
Hi,
this is related to https://svn.boost.org/trac/boost/ticket/8596 which is fixed now, but maybe it doesn't works for you. Note that the trunk regression test since revision 84414 run with this compiler pass ( libs/thread/test/sync/futures/packaged_task/func_ctor_pass.cpp) https://svn.boost.org/trac/boost/browser/branches/release/libs/thread/test/s... http://www.boost.org/development/tests/trunk/developer/thread.html.
Unfortunately not a single msvc tester has run after revision 84468 on release branch.
I have no the possibility to reproduce it now. Please could you send the whole log? (privately if you prefer? Or via a ticket?
I have found the error. My tests don't test for packaged_task
Hi Vincente,
I have found the error. My tests don't test for packaged_task
. The following patch should fix the issue (see attached file)
Please could you apply it a tell me if it fix with your compiler?
I am happy to confirm the fix, really appreciate so quick response from
you. To clarify: I was using boost_1_54_0_beta1_rc1.7z from beta
announcement.
Btw: I am using a workaround for packaged_task to be able store it in
std/boost::function:
typedef boost::packaged_task
Le 01/06/13 12:34, Szymon Gatner a écrit :
Hi Vincente,
I have found the error. My tests don't test for packaged_task
. The following patch should fix the issue (see attached file)
Please could you apply it a tell me if it fix with your compiler?
I am happy to confirm the fix, really appreciate so quick response from you. To clarify: I was using boost_1_54_0_beta1_rc1.7z from beta announcement.
Btw: I am using a workaround for packaged_task to be able store it in std/boost::function:
typedef boost::packaged_task
Task; typedef boost::shared_ptr<Task> TaskPtr; TaskP task(new Task(f)); addTask(boost::bind(&Task::operator (), task)); where:
void addTask(boost::function
task); not a problem of course but with 1.54 beta I was still not able to just move packaged_task to function directly:
boost::packaged_task
t1; boost::function
f2(boost::move(t1)); gives:
Error1error C2248: 'boost::packaged_task<<unnamed-symbol>>::packaged_task' : cannot access private member declared in class 'boost::packaged_task<<unnamed-symbol>>'d:\devel\boost_1_54_0_beta1\boost\function\function_template.hpp1070ConsoleApplication2
same error with std::function.
Is this an issue with packaged_task or a std/boost function implementations. I think boost::functions are not movable so they probably try to copy packaged_task but not sure why it also doesn't work with std::function.
IMO boost::function doesn't supports move semantics. I will take a look at std::function. Could you post the error with std::function? Best, Vicente
2013/6/1 Vicente J. Botet Escriba
IMO boost::function doesn't supports move semantics. I will take a look at std::function. Could you post the error with std::function?
code:
#define BOOST_THREAD_VERSION 4
#include
Le 01/06/13 13:53, Vicente J. Botet Escriba a écrit :
Le 01/06/13 12:34, Szymon Gatner a écrit :
Hi Vincente,
I have found the error. My tests don't test for packaged_task
. The following patch should fix the issue (see attached file)
Please could you apply it a tell me if it fix with your compiler?
I am happy to confirm the fix, really appreciate so quick response from you. To clarify: I was using boost_1_54_0_beta1_rc1.7z from beta announcement.
Btw: I am using a workaround for packaged_task to be able store it in std/boost::function:
typedef boost::packaged_task
Task; typedef boost::shared_ptr<Task> TaskPtr; TaskP task(new Task(f)); addTask(boost::bind(&Task::operator (), task)); where:
void addTask(boost::function
task); not a problem of course but with 1.54 beta I was still not able to just move packaged_task to function directly:
boost::packaged_task
t1; boost::function
f2(boost::move(t1)); gives:
Error1error C2248: 'boost::packaged_task<<unnamed-symbol>>::packaged_task' : cannot access private member declared in class 'boost::packaged_task<<unnamed-symbol>>'d:\devel\boost_1_54_0_beta1\boost\function\function_template.hpp1070ConsoleApplication2
same error with std::function.
Is this an issue with packaged_task or a std/boost function implementations. I think boost::functions are not movable so they probably try to copy packaged_task but not sure why it also doesn't work with std::function.
IMO boost::function doesn't supports move semantics. I will take a look at std::function. Could you post the error with std::function?
I don't find a constructor of function from a rvalue functor.
From the standard:
template<class F> function(F f);
template
2013/6/1 Vicente J. Botet Escriba
From the standard: template<class F> function(F f); template
function(allocator_arg_t, const A& a, F f); 7 Requires: F shall be *CopyConstructible*. f shall be Callable (20.10.11.2) for argument types ArgTypes and return type R. The copy constructor and destructor of A shall not throw exceptions. 8 Postconditions: !*this if any of the following hold: — f is a NULL function pointer. — f is a NULL pointer to member. — F is an instance of the function class template, and !f 9 Otherwise, *this targets a copy of f initialized with std::move(f). [Note: Implementations are encouraged to avoid the use of dynamically allocated memory for small callable objects, for example, where f’s target is an object holding only a pointer or reference to an object and a member function pointer. — end note ] 10 Throws: shall not throw exceptions when f is a function pointer or a reference_wrapper<T> for some T. Otherwise, may throw bad_alloc or any exception thrown by F’s copy or move constructor. But I find an assignment
template<class F> function& operator=(F&& f); 18 Effects: function(std::forward<F>(f)).swap(*this); 19 Returns: *this
This seems confusing. What am I missing?
I don't have a standard but this is in sync what you describe:
http://en.cppreference.com/w/cpp/utility/functional/function/function
http://en.cppreference.com/w/cpp/utility/functional/function/operator%3D
confusing indeed.
Anyway, trying to:
#define BOOST_THREAD_VERSION 4
#include
Le 01/06/13 15:15, Szymon Gatner a écrit :
2013/6/1 Vicente J. Botet Escriba
mailto:vicente.botet@wanadoo.fr> I don't find a constructor of function from a rvalue functor.
But I find an assignment
This seems confusing. What am I missing?
I don't have a standard but this is in sync what you describe:
http://en.cppreference.com/w/cpp/utility/functional/function/function
http://en.cppreference.com/w/cpp/utility/functional/function/operator%3D
confusing indeed.
Anyway, trying to:
#define BOOST_THREAD_VERSION 4
#include
#include <functional> int main() { boost::packaged_task
t; std::function
f2; f2 = std::move(t); } gives exactly the same compilation error:
I get the same kind of error for gcc-4.8.0 and clang-3.2. I suspect that either we are missing something basic or there is a issue on the standard specification and the standard library implementors have not identified it while implementing these operations. Best, Vicente
Le 01/06/13 14:38, Vicente J. Botet Escriba a écrit :
Le 01/06/13 13:53, Vicente J. Botet Escriba a écrit :
Le 01/06/13 12:34, Szymon Gatner a écrit :
Hi Vincente,
I have found the error. My tests don't test for packaged_task
. The following patch should fix the issue (see attached file)
Please could you apply it a tell me if it fix with your compiler?
I am happy to confirm the fix, really appreciate so quick response from you. To clarify: I was using boost_1_54_0_beta1_rc1.7z from beta announcement.
Btw: I am using a workaround for packaged_task to be able store it in std/boost::function:
typedef boost::packaged_task
Task; typedef boost::shared_ptr<Task> TaskPtr; TaskP task(new Task(f)); addTask(boost::bind(&Task::operator (), task)); where:
void addTask(boost::function
task); not a problem of course but with 1.54 beta I was still not able to just move packaged_task to function directly:
boost::packaged_task
t1; boost::function
f2(boost::move(t1)); gives:
Error1error C2248: 'boost::packaged_task<<unnamed-symbol>>::packaged_task' : cannot access private member declared in class 'boost::packaged_task<<unnamed-symbol>>'d:\devel\boost_1_54_0_beta1\boost\function\function_template.hpp1070ConsoleApplication2
same error with std::function.
Is this an issue with packaged_task or a std/boost function implementations. I think boost::functions are not movable so they probably try to copy packaged_task but not sure why it also doesn't work with std::function.
IMO boost::function doesn't supports move semantics. I will take a look at std::function. Could you post the error with std::function?
I don't find a constructor of function from a rvalue functor.
From the standard: template<class F> function(F f); template
function(allocator_arg_t, const A& a, F f); 7 Requires: F shall be *CopyConstructible*. f shall be Callable (20.10.11.2) for argument types ArgTypes and return type R. The copy constructor and destructor of A shall not throw exceptions. 8 Postconditions: !*this if any of the following hold: --- f is a NULL function pointer. --- f is a NULL pointer to member. --- F is an instance of the function class template, and !f 9 Otherwise, *this targets a copy of f initialized with std::move(f). [Note: Implementations are encouraged to avoid the use of dynamically allocated memory for small callable objects, for example, where f's target is an object holding only a pointer or reference to an object and a member function pointer. --- end note ] 10 Throws: shall not throw exceptions when f is a function pointer or a reference_wrapper<T> for some T. Otherwise, may throw bad_alloc or any exception thrown by F's copy or move constructor. But I find an assignment
template<class F> function& operator=(F&& f); 18 Effects: function(std::forward<F>(f)).swap(*this); 19 Returns: *this
This seems confusing. What am I missing? I've a clear answer from Howard. See below.
Le 01/06/13 18:29, Howard Hinnant a écrit :
The reason F has to be CopyConstructible is not because of the signatures of the constructor and assignment operator. Both could be F, or F&&, and we appear to have split the difference. :-)
The reason F has to be CopyConstructible is because of function's copy constructor, and because F is type-erased inside of function. Because F is type-erased, when function is copied, it must copy F using a virtual function call (or via a function pointer). And because F's copy constructor must be called via a virtual function, F's copy constructor is instantiated whether or not function's copy constructor is instantiated. And that is why F must be CopyConstructible.
If F was not type-erased inside of function, F would not have to be CopyConstructible until actually copied.
2013/6/1 Vicente J. Botet Escriba
I've a clear answer from Howard. See below.
Le 01/06/13 18:29, Howard Hinnant a écrit :
The reason F has to be CopyConstructible is not because of the signatures of the constructor and assignment operator. Both could be F, or F&&, and we appear to have split the difference. :-)
The reason F has to be CopyConstructible is because of function's copy constructor, and because F is type-erased inside of function. Because F is type-erased, when function is copied, it must copy F using a virtual function call (or via a function pointer). And because F's copy constructor must be called via a virtual function, F's copy constructor is instantiated whether or not function's copy constructor is instantiated. And that is why F must be CopyConstructible.
If F was not type-erased inside of function, F would not have to be CopyConstructible until actually copied.
Ah, perfectly clear now. Awesome of you to to share the info. Regards, Simon
participants (2)
-
Szymon Gatner
-
Vicente J. Botet Escriba