CLI test failure: Underlying Transport Error

[abitmore]

https://travis-ci.org/bitshares/bitshares-core/builds/406910050

921622ms th_a application.cpp:287 reset_websocket_serv ] Configured websocket rpc to listen on 127.0.0.1:205
921623ms th_a application.cpp:447 startup ] 13 N11websocketpp9exceptionE: Underlying Transport Error

IMHO we need to avoid listening to a port which is less than 1024.

https://travis-ci.org/bitshares/bitshares-core/builds/408586482

189434ms th_a application.cpp:287 reset_websocket_serv ] Configured websocket rpc to listen on 127.0.0.1:681
189434ms th_a application.cpp:447 startup ] 13 N11websocketpp9exceptionE: Underlying Transport Error

[clockworkgr]

Why? It's a system admin choice. (and def wouldn't mark it as a bug)

I wouldn't recommend it of course but it really is a user preference to run as root or not (running as root would allow binding to privileged ports)

[pmconrad]

IMO that's only an indication of travis being broken, not a bug in the cli_test. See ip_local_port_range in http://man7.org/linux/man-pages/man7/ip.7.html .

Nevertheless we might want to change the unit test code to avoid unnecessary failures.

Edit FTR - it's not a travis bug.

[abitmore]

Perhaps not due to port number too small, or 33153 is still too small?

• https://travis-ci.org/bitshares/bitshares-core/builds/408601340

3551897ms th_a application.cpp:209 reset_p2p_node ] Configured p2p node to listen on 0.0.0.0:40944
3551898ms th_a application.cpp:287 reset_websocket_serv ] Configured websocket rpc to listen on 127.0.0.1:33153
3551898ms th_a application.cpp:447 startup ] 13 N11websocketpp9exceptionE: Underlying Transport Error
Underlying Transport Error:
{"what":"Underlying Transport Error"}
th_a application.cpp:290 reset_websocket_server

[abitmore]

This (https://travis-ci.org/bitshares/bitshares-core/builds/408573531) succeeded:

1489215ms th_a application.cpp:209 reset_p2p_node ] Configured p2p node to listen on 0.0.0.0:38730
1489216ms th_a application.cpp:287 reset_websocket_serv ] Configured websocket rpc to listen on 127.0.0.1:28573

[abitmore]

I think I've found the real reason. Uninitialized port number in test cases, e.g. 8db6ca9#diff-df4f0e7f90194894dbd17f9cb2af9d48 .

Fix added to PR #1187.

[pmconrad]

Uninitialized port number in test cases

I think that's not it. server_port_number gets set as a side effect of start_application. Its original value is not used.