I would like to know if the function
Process.exit(pid, :kill)
Is synchronous : once the function has returned, are we sure that no more code from the process pid
will be executed ? If not how could we kill a process in a synchronous way ?
Best
I would like to know if the function
Process.exit(pid, :kill)
Is synchronous : once the function has returned, are we sure that no more code from the process pid
will be executed ? If not how could we kill a process in a synchronous way ?
Best
I am not sure that the function is synchronous, but to be sure, you could first set a monitor on the process with Process.monitor
, then kill it, then wait for the :DOWN
message from the monitor.
Because the Elixir documentation (as well as the Erlang documentation of :erlang.exit/2
, which this function delegates to) only mention “sends an exit signal”, I am fairly certain that the function is asynchronous.
The approach mentioned by @lud works. However, this will not (always) prevent other code from running on pid
before it exits. It will only prevent code running on the process calling Process.exit
to wait with continuing before pid
has exited.
However, another question would be: what are you trying to do? It seems like a rather odd request. Maybe your actual problem can be resolved in a wholly different manner.
Here is what I try to achieve : I want to run a process and be able to kill it if needed. When the process is either killed or terminates normally, there is a piece of code that I always want to run.
The issue I see with @lud
's solution is the following : To monitor a process, you need a pid. So when you call the monitor
function the process is normally already running. It may even be over by the time you call the monitor
function, in which case you will never receive the DOWN
message you expect (this DOWN
message would trigger the piece of code you always want to execute when a process finishes).
Am I making sense ?
From the documentation of Process.monitor/1:
If the process is already dead when calling
Process.monitor/1
, a:DOWN
message is delivered immediately.
So you will receive the :DOWN
message in that case as well.
Well, then it seems to be a good way to implement what I want
Also if you call Process.monitor
on a pid that is not alive, the down message will contain :noproc
as the reason, so you will know your process was not alive.
if you use spawn_monitor/3 you will remove the race condition on starting the process