Fix proxy gw selection when a gw is down

[fholzhauser]

No description provided.

[coveralls]

Coverage increased (+0.2%) to 78.363% when pulling 1373f24 on fix/proxy_gw_selection_gw_down into d4d6a42 on master.

[vkatsuba]

Good catch! Maybe make sense add some test case for this fix?

[fholzhauser]

This is a fairly trivial/safe fix and also needed with some urgency. Extending CT to cover this previously missed scenario is planned in a follow up PR.

[ebengt]

Looks good

[RoadRunnr]

This is a fairly trivial/safe fix and also needed with some urgency. Extending CT to cover this previously missed scenario is planned in a follow up PR.

The interesting thing is that this line was actually hit by the tests, but the catch all a few lines above hide the problem.

In fact adding a when is_record(Socket, socket) to the catch all would have made the unit tests fail.

[vkatsuba]

The interesting thing is that this line was actually hit by the tests, but the catch all a few lines above hide the problem.

In fact adding a when is_record(Socket, socket) to the catch all would have made the unit tests fail.

Currently the checking of socket record already exist in choose_gw/4 however the test cases in proxy_lib_SUITE don't crashed by reason that in node_selection section provided one by one node. For raise an error like function_clause need add into config section node_selection two nodes where first node is invalid and next node is valid, using this configuration example CT in previous version can raise an error like function_clause for case if choose_gw/4 will be called as choose_gw(Next, NodeSelect, Socket, Version) instead of expected choose_gw(Next, NodeSelect, Version, Socket). PR #270 witch updating of CT was provided.