ExceptionHandling Catch an Exception from a Running Taskflow ExceptionHandling_1CatchAnExceptionFromARunningTaskflow Catch an Exception from an Async Task ExceptionHandling_1CatchAnExceptionFromAnAsyncTask Catch an Exception from a Corun Loop ExceptionHandling_1CatchAnExceptionFromACorunLoop This chapters discusses how to handle exceptions from a submitted taskflow so you can properly catch or propagate exceptions in your workload. When a task throws an exception, the executor will store that exception in the shared state referenced by the tf::Future handle. You can catch that exception via calling the get method: tf::Executorexecutor; tf::Taskflowtaskflow; taskflow.emplace([](){throwstd::runtime_error("exception");}); try{ executor.run(taskflow).get(); } catch(conststd::runtime_error&e){ std::cerr<<e.what()<<std::endl; } As tf::Future is derived from std::future, it inherits all the exception handling behaviors defined by the C++ standard. An exception will automatically cancel the execution of its parent taskflow. All the subsequent tasks that have dependencies on that exception task will not run. For instance, the following code defines two tasks, A and B, where B runs after A. When A throws an exception, the executor will cancel the execution of the taskflow, stopping every tasks that run after A. In this case, B will not run. tf::Executorexecutor; tf::Taskflowtaskflow; tf::TaskA=taskflow.emplace([](){throwstd::runtime_error("exceptiononA");}); tf::TaskB=taskflow.emplace([](){std::cout<<"TaskB\n";}); A.precede(B); try{ executor.run(taskflow).get(); } catch(conststd::runtime_error&e){ std::cerr<<e.what()<<std::endl; } ~$exceptiononA #executionoftaskflowiscancelledafteranexecutionisthrown When multiple tasks throw exceptions simultaneously, the executor will only catch one exception and store it in the shared state. Other exceptions will be silently ignored. For example, the following taskflow may concurrently throw two exceptions from task B and task C. Only one exception, either B or C, will be propagated. tf::Executorexecutor; tf::Taskflowtaskflow; auto[A,B,C,D]=taskflow.emplace( [](){std::cout<<"TaskA\n";}, [](){ std::cout<<"TaskB\n"; throwstd::runtime_error("ExceptiononTaskB"); }, [](){ std::cout<<"TaskC\n"; throwstd::runtime_error("ExceptiononTaskC"); }, [](){std::cout<<"TaskDwillnotbeprintedduetoexception\n";} ); A.precede(B,C);//ArunsbeforeBandC D.succeed(B,C);//DrunsafterBandC try{ executor.run(taskflow).get(); } catch(conststd::runtime_error&e){ //caughteitherB'sorC'sexception std::cout<<e.what()<<std::endl; } Similar to std::future, tf::Executor::async will store the exception in the shared state referenced by the returned std::future handle. tf::Executorexecutor; autofu=executor.async([](){throwstd::runtime_error("exception");}); try{ fu.get(); } catch(conststd::runtime_error&e){ std::cerr<<e.what()<<std::endl; } Running the program will show the exception message on the async task: ~$exception On the other hand, since tf::Executor::silent_async does not return any future handle, any exception thrown from a silent-async task will be silently caught by the executor and (1) propagated to the its parent task if the parent task exists or (2) ignored if the parent task does not exist. tf::Taskflowtaskflow; tf::Executorexecutor; //exceptionwillbesilentlyignored executor.silent_async([](){throwstd::runtime_error("exception");}); //exceptionwillbepropagatedtotheparenttf::RuntimetaskandthenitsTaskflow taskflow.emplace([&](tf::Runtime&rt){ rt.silent_async([](){throwstd::runtime_error("exception");}); }); try{ taskflow.get(); } catch(conststd::runtime_error&re){ std::cout<<re.what()<<std::endl; } When you corun a graph via tf::Executor::corun or tf::Runtime::corun, any exception will be thrown during the execution. For example, the code below will throw an exception during the execution of taskflow1: tf::Executorexecutor; tf::Taskflowtaskflow1; tf::Taskflowtaskflow2; taskflow1.emplace([](){ throwstd::runtime_error("exception"); }); taskflow2.emplace([&](){ try{ executor.corun(taskflow1); }catch(conststd::runtime_error&re){ std::cout<<re.what()<<std::endl; } }); executor.run(taskflow2).get(); We can observe the same behavior when using tf::Runtime::corun: tf::Executorexecutor; tf::Taskflowtaskflow1; tf::Taskflowtaskflow2; taskflow1.emplace([](){ throwstd::runtime_error("exception"); }); taskflow2.emplace([&](tf::Runtime&rt){ try{ rt.corun(taskflow1); }catch(conststd::runtime_error&re){ std::cout<<re.what()<<std::endl; } }); executor.run(taskflow2).get(); For the above example, if the exception is not caught with tf::Runtime::corun, it will be propagated to its parent task, which is the tf::Runtime object rt in this case. Then, the exception will be propagated to taskflow2. tf::Executorexecutor; tf::Taskflowtaskflow1; tf::Taskflowtaskflow2; taskflow1.emplace([](){ throwstd::runtime_error("exception"); }); taskflow2.emplace([&](tf::Runtime&rt){ rt.corun(taskflow1); }); try{ executor.run(taskflow2).get(); } catch(conststd::runtime_error&re){ std::cout<<re.what()<<std::endl; } For the above example, if the exception is not caught with tf::Runtime::corun, it will be propagated to its parent task, which is the tf::Runtime object rt in this case. Then, the exception will be propagated to taskflow2.