Fork me on GitHub
#cider
<
2018-08-21
>
plexus08:08:37

I'm having trouble with sesman-link-with-project. I can pick a connection, and I can see the project link in sesman-browser, but the modeline still says "not connected", and I can't eval.

plexus08:08:52

Doing sesman-link-with-directory from the same buffer does work...

plexus08:08:33

I think this may have been brought up in the past but could there be room for a discussion around CIDER's release process? Currently I'm guessing most people essentially run off master, since that's how MELPA works, but that does mean that every year or two everyone is taken along for a rollercoaster ride of instability. The docs do mention pinning to melpa-stable, but for most people it really should be the default.

4
plexus08:08:14

I'm not sure which model would be ideal, but a relatively small change with big impact could be to develop on a develop branch, and use master for pre-release testing.

plexus10:08:08

I've been having an issue where evaluating certain forms will cause Emacs to hang. It seems it's blocking on a call to (accept-process-output nil 0.01) in nrepl-send-sync-request, which is strange because apparently that 0.01 is supposed to be a timeout, so it's not expected to block

plexus10:08:26

actually scratch that, seems it's not breaking out of the loop around it, that's what's locking it up... investigating.

plexus10:08:50

ok, figured it out. I'll file an issue, the problem was that I had a clj and cljs repl open, but the browser hadn't opened yet, so the cljs repl wasn't functional, causing it to be seen as a "clj" repl instead. CIDER tries to send it the eval, and fails to do so, so it's waiting forever on a response

bozhidar14:08:10

> I’m not sure which model would be ideal, but a relatively small change with big impact could be to develop on a develop branch, and use master for pre-release testing.

bozhidar14:08:10

The problem with this approach is that we’d simply be postponing the discovery of certain bugs until things are merged to master. Very few people outside the core devs actually would run from this develop branch, which would make it pretty useless for testing purposes.

4
bozhidar14:08:17

Obvious the real problems are: • not enough to time to work on CIDER and related projects • which leads to cutting corners everywhere • and making it very hard to reliably execute big changes If the project had an extensive test suite (for instance), that would have helped with many of the problems people are failing on a daily basis. Unfortunately when we were starting out it was pretty hard to develop such a test suite and after a while it was hard to backfill it. Ultimately it was a comprise of either making some progress with the humble amount of time available to work on the project vs freeze everything for a very long time until we clean up all the internal mess.

bozhidar14:08:51

Long story short - I really don’t think that there’s some magic recipe to stability that won’t entail a lot of extra leg work that I don’t see anyone volunteering to do. 🙂

plexus14:08:55

> Very few people outside the core devs actually would run from this develop branch By that logic why does anyone ever release alpha/beta/RC versions?

bozhidar14:08:34

@plexus In my experience those only make sense for really big projects (in terms of users). Most of the time alphas, betas, etc are simply ignored by the majority of the users. The only alphas I’ve used in the Clojure world are the “forever” alphas by Cognitect (e.g. core.async).

bozhidar14:08:55

I can certainly tag more releases down the road, but I’m skeptical that would amount to much, as almost all the time I asked people to try some CIDER branch I would get between 0 and 3 people willing to do so. 🙂

bozhidar14:08:17

TL;DR; I understand very well the theoretical purpose of pre-releases, but for them to work in practice there should be people willing to provide feedback and testing with respect to them.

bozhidar15:08:50

Anyways, I’m always open to improve the release process. I really think that decoupling the release cycles of CIDER and cider-nrepl was a major step forward, as now at least users of MELPA won’t be running cider-nrepl snapshots all the time.