[
https://issues.apache.org/jira/browse/MINIFICPP-1985?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17635292#comment-17635292
]
Ádám Markovics commented on MINIFICPP-1985:
-------------------------------------------
This random port that the OS assigns must be an unused one, right? I think it's
even more important than being random, that might affect naming functions or
documentation.
> Use asio related tests on random ports
> --------------------------------------
>
> Key: MINIFICPP-1985
> URL: https://issues.apache.org/jira/browse/MINIFICPP-1985
> Project: Apache NiFi MiNiFi C++
> Issue Type: Improvement
> Reporter: Martin Zink
> Assignee: Martin Zink
> Priority: Minor
>
> The tests in question are PutTCPTests, PutUDPTests, ListenTCPTests,
> ListenUDPTests.
> We shouldn't use hard coded ports to run the tests on.
> I've tried creating a test utility that creates a socket on port 0. (so the
> OS assigns a random port to it), then close the socket and return the port to
> bind to.
> Unfortunately this could leave the port on TIME_WAIT (which could be
> workaround by setting reuse_address (which we may very well want to use
> besides this, so restarting the processor on the same port can work without
> hiccups), but due to race conditions sometimes the port is not even in the
> TIME_WAIT state when the processor wants to bind to it.
>
> The proper solution would be to directly set the processor (or utility) to
> port 0. And get the listening port out after the socket/acceptor is already
> up and running, this would require a larger refactor (both tests and prod
> code as well), it would make sense to do this during the MINIFICPP-1979
> refactor.
--
This message was sent by Atlassian Jira
(v8.20.10#820010)
