Fixed flappers and test_SSLHandshakeFirst by skipping if server is < 2.10.0 #784
Conversation
Signed-off-by: Ivan Kozlovic <[email protected]>
Codecov ReportAll modified and coverable lines are covered by tests ✅
Additional details and impacted files@@ Coverage Diff @@
## main #784 +/- ##
==========================================
+ Coverage 68.71% 68.85% +0.14%
==========================================
Files 39 49 +10
Lines 15207 15219 +12
Branches 3143 3133 -10
==========================================
+ Hits 10449 10479 +30
+ Misses 1700 1692 -8
+ Partials 3058 3048 -10 ☔ View full report in Codecov by Sentry. |
Added missing `-a 127.0.0.1` in some tests to prevent the server from sending other IP addresses to the client. Signed-off-by: Ivan Kozlovic <[email protected]>
kozlovic
changed the title
| @@ -18352,7 +18352,7 @@ void test_GetClientID(void) | |||
| testCond(true); | |||
| return; | |||
| } | |||
| pid1 = _startServer("nats://127.0.0.1:4222", "-cluster nats://127.0.0.1:6222 -cluster_name abc", true); | |||
| pid1 = _startServer("nats://127.0.0.1:4222", "-a 127.0.0.1 -p 4222 -cluster nats://127.0.0.1:6222 -cluster_name abc", true); | |||
There was a problem hiding this comment.
@kozlovic is this (extra server flags) a windows-specific fix?
There was a problem hiding this comment.
Not really. When we just specify the port, the server uses 0.0.0.0 for the TCP listen address, which then can resolve to multiple IPs that are sent to the client through gossip protocol. What I have seen (in my Windows VM, and probably what is happening on Windows CI) is that the server would return multiple addresses (say 192.168.x.y but also 10.5.x.y, etc..), which "interfere" with the tests expectations. On Linux or other CIs that we are running it on, there was probably nothing returned (maybe it is 127.0.0.1).
So the "fix" is not really Windows specific. It is simply making the test more deterministic.
…irst by skipping if server is < 2.10.0 (#784)
…irst by skipping if server is < 2.10.0 (nats-io#784)
Signed-off-by: Ivan Kozlovic [email protected]