pre_exit.hpp File Reference

Register and run operations that must execute before program exit. More...

Go to the source code of this file.

Typedefs
using	stlab::pre_exit_handler
	Function type invoked during pre_exit() (must not throw; noexcept with C++17 and later).

Functions
void	stlab::stlab_pre_exit ()
	An extern "C" vector for pre-exit() to make it simpler to export the function from a shared library.
void	stlab::stlab_at_pre_exit (pre_exit_handler f)
	An extern "C" vector for at_pre-exit() to make it simpler to export the function from a shared library.
void	stlab::pre_exit ()
	Invoke all registered pre-exit handlers in the reverse order they are registered. It is safe to register additional handlers during this operation. Must be invoked exactly once prior to program exit. Handlers needed to unblock running core work must be registered after the first timer or default-executor use and before invoking pre_exit(). On threadless Emscripten, returns after initiating retirement; the next ordinary main task runs only after executor work and deferred handlers finish. Default/high/low submission remains available during the drain and is a precondition violation after it completes.
void	stlab::at_pre_exit (pre_exit_handler f)
	Register a pre-exit handler. The pre-exit-handler may not throw. With C++17 or later it is required to be noexcept.

Detailed Description

Register and run operations that must execute before program exit.

Register handlers with at_pre_exit(); call pre_exit() once before normal process termination (before std::exit() or when leaving main()). Handlers run in reverse registration order. Required when using the default executor so detached or canceled tasks do not overlap global teardown (see default_executor.hpp). std::quick_exit() is an alternative when appropriate.

Timer and default-executor resources share a teardown handler registered on their first use. Register application handlers that release running work after that first use and before calling pre_exit(), so they execute before core teardown. Core teardown cancels pending timers, waits for running timer callbacks, then drains the default executor.

On threadless Emscripten, pre_exit() initiates asynchronous retirement instead of blocking. Default/high/low work, including descendants and capture cleanup, drains through the event loop while ordinary main tasks are deferred. Earlier-registered handlers resume in LIFO order after that drain. A main task submitted after pre_exit() is the completion fence: producer main submissions and the remaining handlers precede it. Drain work must not depend on deferred main tasks for progress. Final process exit must follow that fence.