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.
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
.
Raises Failure
if fork was unsuccessful.
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.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 behavior whereby both of these descriptors are connected to /dev/null. daemonize_wait
, however, will not redirect stdout/stderr to /dev/null if they are already redirected to a regular file by default, i.e., default redirection is `Dev_null_skip_regular_files
. This is to preserve behavior from earlier versions.)
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 asynchronously, 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 shut down. On non-Linux platforms, using ~allow_threads_to_have_been_created:true
eliminates the protection daemonize
and daemonize_wait
have regarding threads.
Raises Failure
if forking was unsuccessful.