Details
-
Test
-
Status: Open
-
Major
-
Resolution: Unresolved
-
None
-
None
-
None
-
None
Description
Although our testing is generally pretty good about not having hard coded ports, we still have some places in our code where we can encounter "port roulette" bugs – expecting that a particular network port will be available to use to Bind to at some point in the test.
But instead of classic port bugs – basing that assumption off of hardcoded ports – the problematic portions of our test code expect that since a port was available to us previously (via code asking the OS to assign a port by binding to '0') it will still be available to us later even if we release it. But there is no guarantee the OS won't have already re-allocated that port to a different process/purpose.
I'm creating this issue to serve as a parent for Sub-Tasks targeting the specific instances where this problematic practice comes up. Some of these can be solved easily, others may be more challenging.
This issue can also serve as a reference point for citating/linking from other jiras, in case there is any general discussion/concern about the overall problem, or in case new instances are discovered later.