@jadeallenx, I'd like to have #452 approved and merged before I go for #454, then #455, then this one, as I'm sure there'll be conflicts.

@x0id, thanks for this. A couple of comments:

• would you mind porting this to fish and csh?
  • if you do mind, would you create a task for this port to be done at a later date? (then maybe I can tackle it)?
• maybe you'll want to wait on the previous action I mentioned - approve + merge
• I think this solution is acceptable, though I wonder what'll happen to e.g. plugins that use rebar3's cache directly (I didn't look into rebar3 to see how the path to the folder is exposed - but I know for sure there are plugins that depend on it, and it'd be a pity to break local/CI compatibility)
  • in that case, if rebar3's public functions don't take this variable into account (which I would say is oversight, if it happens) maybe we could open an issue there, too?
• this change is lacking tests - one could e.g. call the activate script on two different installations and make sure the $absdir/.cache/rebar3 folders are populated, have different content

Yeah this is a great patch, thanks - does need a rebase but I'd hold off until those other two PRs are merged so you don't have to constantly rebase... thank you! ❤️

@jadeallenx, I'd like to have #452 approved and merged before I go for #454, then #455, then this one, as I'm sure there'll be conflicts.

@x0id, thanks for this. A couple of comments:

• would you mind porting this to fish and csh?

what do you mean by that? I added lines to all variants of activate/deactivate including fish and csh I believe

• if you do mind, would you create a task for this port to be done at a later date? (then maybe I can tackle it)?
• maybe you'll want to wait on the previous action I mentioned - approve + merge
• I think this solution is acceptable, though I wonder what'll happen to e.g. plugins that use rebar3's cache directly (I didn't look into rebar3 to see how the path to the folder is exposed - but I know for sure there are plugins that depend on it, and it'd be a pity to break local/CI compatibility)

What I could suggest - making that variable optional. I'll think about it.

• in that case, if rebar3's public functions don't take this variable into account (which I would say is oversight, if it happens) maybe we could open an issue there, too?

Not sure I'm aware of these details, if you have an example let me know, but I agree with the point.

• this change is lacking tests - one could e.g. call the activate script on two different installations and make sure the $absdir/.cache/rebar3 folders are populated, have different content

Is there any part of kerl code that is for testing? Sorry, I didn't look too deep so far, so I can't say I understand what is the testing strategy in this project.

what do you mean by that? I added lines to all variants of activate/deactivate including fish and csh I believe

Right you are. The GitHub diff UI is so small I only saw the bash changes... good stuff!

What I could suggest - making that variable optional. I'll think about it.

I'm Ok as-is, though, as long as rebar3 honours it where it exposes the global cache folder... (later) I checked and it seems it does.

    State5 = case os:getenv("REBAR_CACHE_DIR") of
                false ->
                    State4;
                CachePath ->
                    rebar_state:set(State4, global_rebar_dir, CachePath)
            end,

then later...

global_cache_dir(Opts) ->
    Home = home_dir(),
    rebar_opts:get(Opts, global_rebar_dir, filename:join([Home, ".cache", "rebar3"])).

which I'd say is the "official" exposed function.

Is there any part of kerl code that is for testing? Sorry, I didn't look too deep so far, so I can't say I understand what is the testing strategy in this project.

Look in the GitHub actions file, under .github. I can help you set something up if you're stuck...

So what's missing now is:

• rebase on master once Warn on activation script staleness #454, and Remove no-longer-supported OTP version-specific code #455 get merged
• add tests to GitHub CI

Once we get this in I'll do a new release <3

I have rebased the PR onto the latest master after adding a couple of cleanup-improving updates.
Will work on the tests...

I added tests/activate_test.sh script for testing sh-based activate and kerl_deactivate cleanup. This is a basic test. It may be improved/extended later.
I have also tried to work on the fish version and suddenly realized that the fish-activate/deactivate is far from perfect. The cleanup code just didn't work out of the box - maybe nobody tested it.
My suggestion is to run CI, verify what has been done so far, and merge/release it as is if no problems are found. I can work on validating/improving/testing fish (and csh) versions later in a separate branch/PR.

I have also tried to work on the fish version and suddenly realized that the fish-activate/deactivate is far from perfect. The cleanup code just didn't work out of the box - maybe nobody tested it.

Undoubtedly true - unless it’s zsh or bash it doesn’t get a lot of love.

My suggestion is to run CI, verify what has been done so far, and merge/release it as is if no problems are found.

ok that seems reasonable

I can work on validating/improving/testing fish (and csh) versions later in a separate branch/PR.

that would be very welcome- thanks!

Looking good. ❤️ 2 minor changes requested, and a code update suggestion added.

I can work on validating/improving/testing fish (and csh) versions later in a separate branch/PR

Agreed. Would you open an issue for this, referencing your comment, in that case? Thanks. Maybe we can even try do see what an AI would do for "convert this bash to that fish" and then adapt? For tests too, but that's "for the future".

Tests are failing and so is the Lint action. I can help with the lint, if you want, by doing it locally and proposing changes. For the tests, there's very little information, so I'm actually not sure what's breaking: maybe in case of error we could also output the diff?

I have fixed lint errors, activate-test integration in CI, and corrected activate/deactivate code slightly.
Now all tests went green (in my account).
https://github.com/x0id/kerl/actions/runs/6121381052

Looks good to me. I'll wait for @jadeallenx.

@x0id, is the stuff you plan to do identified in the Issues? Or not interesting to have there (e.g. for pre-discussion before a pull request)?

Let's ship it :) Thanks!

@x0id, is the stuff you plan to do identified in the Issues? Or not interesting to have there (e.g. for pre-discussion before a pull request)?

Tried to briefly describe in #458.
I have a branch for the fish version already, still WIP.

Perfect, stuff, thanks much.