I'm trying to test this, however, even with the server running I get:

$ ./blp-client 
Command not found: blp-client

Is this expected?

The correct command binary for the client is ./bloop the blp-client is the unwrapped version. Maybe it should not be in the bin directory. I do not know if there are any best practices around this.

I tried to update it on my side as well, and I have obtained some different sha values...

For example, to get the nailgunCommit, I got 7dfdad8c from https://github.com/scalacenter/bloop/releases/download/v1.3.0/install.py

Maybe it has changed since?

The sha256 for client should change as well, I get 0bchc59dmdxqg5vpdsnhagw0374jr19xg4l5rydcp5ljvv2dp6d4.

For zshCompletion, I get 09qq5888vaqlqan2jbs2qajz2c3ff13zj8r0x2pcxsqmvlqr02hp.

Caveat: I'm not very advanced in Nix, so may I miss something...

PS. Thank you @Tomahna for adding bloop, very useful!

Thank you for testing this ! Yes you're perfectly right. As I did not change the sha256, it used the value in my cache so it worked for me, but of course not for you.

It should be fixed.

I've made a few changes. That should prevent both problems from happening again.

As the client has its own derivation it will not appear in the bin directory of the final derivation (only the wrapper will appear), so it should avoid the confusion.

As both the client and the zsh completion have their own derivation, they will have to be rebuilt on version changes and the cache should not be hit.

Please let me know, if there is a better/simpler way to do this.

 > bloop about
bloop v1.3.2

Using Scala v2.12.8 and Zinc v1.3.0-M4+22-4704d479
Running on Java v1.8.0_212 (/nix/store/i9kl79ch4id866b3xvi3rwr963a4cz9h-openjdk-8u212-ga-jre/lib/openjdk/jre)                                                                
Maintained by the Scala Center (Martin Duhem, Jorge Vicente Cantero)

however:

 > bloop
Command not found: wfid76hlhxc91nk3j37a4sivkbvcg942-bloop-client-9327a60a

If you want to launch the server from the cli you have to pass the location manually, I'm not sure how to fix that and I honestly don't intend to try as I do not use it this way. However you can use the systemd service with services.bloop.install = true or just run the command blp-server.

For the second part, right now I don't know what causes it. It's not really a blocker because running the cli without a command is not really the way you want to use bloop anyway but it is definitely annoying.

Agreed. Plus I'm not sure how it worked before the upgrade and as both are nonblockers for usage, I'd prefer the new even if somewhat quirky bloop vs nice old one even if these are new issues :)

I checked the old version, it behaves the same. I'd need to check on a non-nix machine to see how it is actually supposed to behave.

@FRidh maybe we can remove the work-in-progress tag ?

btw. I've been using this PR in my manually ported nixpkgs for a while now - not quite sure who could merge it would be great if anyone did before we need to update it again :)

(I've only tested bloop itself, not zsh completions)

I would like that too.

So I've made a few changes to get closer to the way bloop is packaged on other distributions.

This fix a few problems:

• bloop without argument now work properly
• bloop server work transparently without having to give it the --server-location
• it is possible to add java options the way bloop expects them bloop server -J-Xmx2G

Thanks to @svalaskevicius for helping me improve this !

just had a chance to test it (except systemd config, as I use my own setup for systemd user unit) - seems like I'm able to pass options to bloop server w/o any issues now :)

Yay! My local nixpkg can now just point to the master :) thanks!