Elisp bugfix/docs

[simonjwinwood]

Nothing too exciting, although using ad-advice-activate doesn't seem to work for me, while using after-save-hook does.

[DanielG]

I didn't realize anyone other than me was using this branch :)

I've been having problems with the after-save advice as well recently and switched to the hook too, just haven't committed that yet.

The problem with after-save-hook is that it only runs if the buffer was actually saved, so if nothing is modified it doesn't run check again which can be annoying with the new jump-to-error stuff. So eventually I'd like to investigate why the advice isn't working when it used to.

[DanielG]

One minor difference to my local version is that I add the hook buffer-locally and then don't bother checking if haskell-mode is t. I think with your version making the hook buffer local anyways might be a good idea too? Not sure though.

[simonjwinwood]

I was thinking of writing the minor-mode stuff myself, and then noticed the elisp branch. With the hook stuff, I imagine that the case where you have made no changes but still want ghc to run would be pretty rare, maybe just add a binding for ghc-reload or whatever it is called. The only reason I didn't add the hook locally is that after-save-hook is marked as risky (although I notice haskell-mode makes it local anyway) --- having it be local makes more sense to me apart from that.

Do you have any other TODOs? I use ghc-mod in Emacs a fair bit, so I am happy to help.

[DanielG]

I dunno I do use jump-to-error quite a bit and wanting to jump back to the first error location does come up. I'll see if I can get the advice to work again at some point before this gets released, I mean it's been working for like 4+ year or something so .. ;)

As far as TODOs go yeah lots, the elisp is pretty neglected and really needs a good cleanup. Some immediatley useful things I can think of is removing legacy uses of ghc-* overlay properties (other than ghc-info): https://github.com/DanielG/ghc-mod/blob/dev-elisp/elisp/ghc-check.el#L212

Those used to be useful because Kazu decided to use defstruct for the info structure but I'm using pcase-let now to make that easier. Part of my ongoing mission to make ghc-mod's elipsp bits feel a bit more haskelly. Anyways you should be able to find all the relevant references by looking for overlay-get.

Then there is rewriting/extending ghc-with-process to support the unmap-file ghc-mod command which would probably get rid of such nasty bugs as #609 and maybe #652 but that probably needs a lot of prior discussion.

Speaking of which if you want to you should hang out in #ghc-mod which is where I can usually be found for discussions too tangential for a GH issue.

Other than that the component: elisp label is useful, for example #705 might be easy to do (thought I don't remember the exact quoting rules we ended up with off hand).

Lots to do in Elisp land in any case.