Bloop should explain how to start itself when it cannot connect to the server

[Duhemm]

Currently, running bloop when there's no server started prints:

Could not connect to server 127.0.0.1:8212

It would be better to add something along the lines of:

Have you forgotten to start the server? It can be started with blp-server

Moreover, if we're changing the name of the server executable, that will make it harder to discover, and more users risk hitting that.

This should be improved in the python script.

[propensive]

I have wrapped bloop in a script which checks whether the server is running, and if not, starts it in a screen session, something like,

(netstat -ln | grep -q `:8212 ' || (screen -X -S bloop quit ; screen -S bloop -d -m ~/.bloop/blp-server && sleep 1)) && bloop "$@"

There's a lot of ugly stuff in there, and it relies on screen and netstat and (worst of all) sleep, but it has meant I haven't had to think much about starter the server when it's not running.

Might be useful as the basis for something less hacky...

[olafurpg]

On OSX, I've found launch agents to work really well. I set it up once in December and the server has always been up when I've run bloop since then. To restart the server I run launchctl stop Bloop.

[jvican]

For now, I lean towards not focusing on this issue and letting users deal with it manually. In the future, it would be nice if bloop starts the server on its own, but this is not something we should focus on in the long term.

The very notion of a client-server should be ingrained in people's mind when they use bloop, and I find this kind of user error to be beneficial in the long-term: it prevents users from running two servers.

Related ticket upstream: facebookarchive/nailgun#55.

[jvican]

Fixed in #293.