test: dynamic port in dgram bind shared ports

[sebastianplesciuc]

Removed common.PORT from test-dgram-bind-shared-ports in order to
eliminate the possibility of port collision with other tests.

Refs: #12376

Checklist

• make -j4 test (UNIX), or vcbuild test (Windows) passes
• tests and/or benchmarks are included
• commit message follows commit guidelines

Affected core subsystem(s)

test

[sebastianplesciuc]

I have a bit of trouble understanding why this works. If I were to set the socket1.bind() port to be the one that comes from worker1 explicitly, then I would get a EADDRINUSE from socket1 in worker2 even though exclusive is set to false.

While exclusive is false, socket1 gets the same port on both workers while using 0 and I can assert that to check it's true.

[santigimeno]

I have a bit of trouble understanding why this works. If I were to set the socket1.bind() port to be the one that comes from worker1 explicitly, then I would get a EADDRINUSE from socket1 in worker2 even though exclusive is set to false.

That's probably because the key by which the handle is identified in the cluster is different: the first worker would use 0 as port and the second the specific port.

[sebastianplesciuc]

@santigimeno Thank you for the explanation! Now I get it :)

[santigimeno]

Even if it's a corner case, I wonder if it should be handled correctly by the cluster module. /cc @bnoordhuis @cjihrig ?

[mscdex]

@santigimeno This is a long-standing issue and is why I did not bother "fixing" the cluster and some other tests when mass converting tests from common.PORT to 0.

See #7043, which was my attempt at trying to fix the cluster random port binding issue.

[santigimeno]

@mscdex thanks for the info. I've been looking at #7043, and instead of redefining what binding to port 0 means, I was thinking more about solving the specific issue @sebastianplesciuc was referring to and storing the real port number that was assigned to the port 0 query and return the same handle whether a worker queries for port 0 or that specific port. Does it make sense to you?

[sebastianplesciuc]

Closing this since it was moved to sequential by this PR