Module Core.Daemon

This module provides support for daemonizing a process. It provides flexibility as to where the standard file descriptors (stdin, stdout and stderr) are connected after daemonization has occurred.

module Fd_redirection : sig ... end
val daemonize : ?redirect_stdout:Fd_redirection.t ‑> ?redirect_stderr:Fd_redirection.t ‑> ?cd:string ‑> ?umask:int ‑> ?allow_threads_to_have_been_created:bool ‑> unit ‑> unit

daemonize makes the executing process a daemon.

See Chapter 13 of Advanced Programming in the UNIX Environment Second Edition by Stephens and Rago for more details.

The optional arguments have defaults as per daemonize_wait, below.

By default, output sent to stdout and stderr after daemonization will be silently eaten. This behaviour may be adjusted by using redirect_stdout and redirect_stderr. See the documentation for daemonize_wait below.

See daemonize_wait for a description of allow_threads_to_have_been_created.

val daemonize_wait : ?redirect_stdout:Fd_redirection.t ‑> ?redirect_stderr:Fd_redirection.t ‑> ?cd:string ‑> ?umask:int ‑> ?allow_threads_to_have_been_created:bool ‑> unit ‑> (unit ‑> unit) Core__.Import.Staged.t

daemonize_wait makes the executing process a daemon, but delays full detachment from the calling shell/process until the returned "release" closure is called.

Any output to stdout/stderr before the "release" closure is called will get sent out normally. After "release" is called, stdin is connected to /dev/null, and stdout and stderr are connected as specified by redirect_stdout and redirect_stderr. The default is the usual behaviour whereby both of these descriptors are connected to /dev/null.

Note that calling release will adjust SIGPIPE handling, so you should not rely on the delivery of this signal during this time.

daemonize_wait allows you to daemonize and then start async, but still have stdout/stderr go to the controlling terminal during startup. By default, when you daemonize, toplevel exceptions during startup would get sent to /dev/null. With daemonize_wait, toplevel exceptions can go to the terminal until you call release.

Forking, especially to daemonize, when running multiple threads is tricky, and generally a mistake. daemonize and daemonize_wait check that the current number of threads is not greater than expected. daemonize_wait and daemonize also check that threads have not been created, which is more conservative than the actual requirement that multiple threads are not running. Using ~allow_threads_to_have_been_created:true bypasses that check. This is useful if Async was previously running, and therefore threads have been created, but has since been shutdown. On non-Linux platforms, using ~allow_threads_to_have_been_created:true eliminates the protection daemonize and daemonize_wait have regarding threads.