Issue #762: Remove calls to log.Fatal

[danomagnum]

The function that used log.Fatal in config.go isn't exported and is only called by functions that return errors so I added an error return then pass it up.

The first log.Fatal in server.go followed the same approach as #616 to delay the error till the start() is called on the server.

The second log.Fatal in server.go should never be possible, so I changed it to a panic (which should never happen).

Fixes #762

[magiconair]

Other than that LGTM

[magiconair]

LGTM

[danomagnum]

@magiconair

Did you see what happened with the checks on this? It failed the first time (couldn't open port 4840 in one of the conn_test tests). It does pass on my machine, so I pushed an empty commit and when it ran again, it passed. Some kind of race condition must exist in terms of one test closing the port fast enough before another can open it? Pretty strange.

On the python integration tests I've started getting a different error sometimes too. I'm sure it's something with how my network is setup, but I'm not 100% sure why it doesn't try to go out the gateway anyway.

=== RUN   TestClientTimeoutViaOptions
Endpoints other than open requested but private key and certificate are not set.
Listening on 0.0.0.0:4840
    timeout_test.go:55: 
                Error Trace:    /home/dan/programs/opcua/tests/python/timeout_test.go:55
                                                        /home/dan/programs/opcua/tests/python/timeout_test.go:28
                Error:          got %#v, wanted net.timeoutError
                Test:           TestClientTimeoutViaOptions
                Messages:       connect: network is unreachable
--- FAIL: TestClientTimeoutViaOptions (1.45s)

[magiconair]

I've seen this and it happens sporadically but not often enough for me to be concerned. Would be great to understand why though.

Released as v0.6.3

[magiconair]

@danomagnum I wonder where all the thumbs-up on the issue came from. People from your company? If this was such an issue why didn't anyone raise this sooner.

[danomagnum]

@magiconair I don't know. Not sure who any of the people are. I saw the issue about log.fatal come in and it involved the server so I went ahead and fixed it. That issue got a bunch of attention really quickly as well. Maybe all those people are at the same company?