libdill: Structured Concurrency for C


bundle_mem - create an empty coroutine bundle


#include <libdill.h>

int bundle_mem(
    void* mem);


Coroutines are always running in bundles. Even a single coroutine created by go gets its own bundle. A bundle is a lifetime control mechanism. When it is closed, all coroutines within the bundle are canceled.

This function creates an empty bundle. Coroutines can be added to the bundle using the bundle_go and bundle_go_mem functions.

To wait until coroutines in the bundle finish use bundle_wait function.

When hclose() is called on the bundle, all the coroutines contained in the bundle will be canceled. In other words, all the blocking functions within the coroutine will start failing with an ECANCELED error. The hclose() function itself won't exit until all the coroutines in the bundle exit.

This function allows to avoid one dynamic memory allocation by storing the object in user-supplied memory. Unless you are hyper-optimizing use bundle instead.

mem: The memory to store the newly created object. It must be at least BUNDLE_SIZE bytes long and must not be deallocated before the object is closed.


In case of success the function returns handle of the newly create coroutine bundle. In case of error it returns -1 and sets errno to one of the values below.



int b = bundle();
bundle_go(b, worker());
bundle_go(b, worker());
bundle_go(b, worker());
/* Give wrokers 1 second to finish. */
bundle_wait(b, now() + 1000);
/* Cancel any remaining workers. */


bundle(3) bundle_go(3) bundle_go_mem(3) bundle_wait(3) go(3) go_mem(3) hclose(3) yield(3)