nodeos crashing on occupied listener port of net_plugin

[ghost]

Run the following command twice from different terminals:
$nodeos -d ./data --config-dir ./config --contracts-console

Problem:
warn 2019-09-13T20:01:45.870 thread-0 controller.cpp:432 initialize_blockchai ] Initializing new blockchain with genesis state
info 2019-09-13T20:01:46.402 thread-0 chain_plugin.cpp:982 plugin_startup ] starting chain in read/write mode
info 2019-09-13T20:01:46.402 thread-0 chain_plugin.cpp:986 plugin_startup ] Blockchain started; head block is #1, genesis timestamp is 2018-06-01T12:00:00.000
./02-nodeos.sh: line 17: 31060 Segmentation fault: 11 $vg $nodeos -d ./data --config-dir ./config --contracts-console

Solution:
Proper error handling if bind of listener socket fails.

Test:
Run nodeos twice and on second time it should exit with proper error message.

[heifner]

What version of eosio? I'm unable to reproduce with 1.8.3.

[ghost]

Saw it on latest release/1.8.x d87164f
as well as on master 7116e88

[heifner]

I tested that commit of 1.8.3, what platform?

[ghost]

MACOS

Are we using the SO_REUSEADDR socket param on that listener? With that enabled it should not throw the exception. Perhaps this is already enabled per default on other platforms?

[heifner]

Yes, using acceptor->set_option(tcp::acceptor::reuse_address(true));
I tested on Ubuntu 18.10. I'll test mac on Monday.

[ghost]

If it's of any help, I've sent a valgrind.log on Telegram.

[heifner]

Fixed by #7931