QA for latest changes

[vemv]

I've got a couple reports that cljr-slash is not working for the latest versions of refactor-nrepl + clj-refactor.el.

When I tried to debug with those users, the backend part seemed in order e.g. (refactor-nrepl.ns.libspecs/namespace-aliases) returns a proper map. There's a good deal of unit-testing in refactor-nrepl anyway.

I don't know so much about clj-refactor.el though. I have barely performed changes to it yet. I see in the commit log cljr-slash got bit of a revamp earlier this year.

I use an old version of clj-refactor.el and have it hard to use $latest I believe. So whenever I QA things, what I'm trying out is refactor-nrepl (official, latest) + my clj-refactor.el (old, divergent). So it's not exactly realistic QAing.

I wonder if someone in the team would have the time to give the latest refactor-nrepl + clj-refactor.el versions a go? @expez @magnars @yuhan0

🍻 - V

[vemv]

As a pereliminary diagnostic, this is part of the user error report I got (using clj-refactor.el master, latest)

Debugger entered--Lisp error: (wrong-type-argument arrayp nil)
  substring(nil 0 0)
  replace-regexp-in-string("\n" " " nil)
  cljr--keywordp(nil)
  cljr--in-keyword-sans-alias-p()
  ...

It seems to have to do with (cider-symbol-at-point). In #493 (WIP) we found out (thing-at-point 'symbol) is more desirable?

Would love if someone could just give a QAing / debugging session and weed out various unforeseen cases for this conditional:

clj-refactor.el/clj-refactor.el

Lines 2018 to 2027 in 9e1f920

	(when-let (aliases (and cljr-magic-requires
	(not (cljr--in-map-destructuring?))
	(not (cljr--in-ns-above-point-p))
	(not (cljr--in-reader-literal-p))
	(not (cider-in-comment-p))
	(not (cider-in-string-p))
	(not (cljr--in-keyword-sans-alias-p))
	(not (cljr--in-number-p))
	(clojure-find-ns)
	(cljr--magic-requires-lookup-alias)))

I'd say it's easier than it sounds!

[yuhan0]

I can give the latest 3.0.0 refactor-nrepl a go next week, but from the traceback it just looks like a missing type check in cljr--in-keyword-sans-alias-p.
Both cider-symbol-at-point and thing-at-point will return nil when not in a symbol.

(defun cljr--in-keyword-sans-alias-p ()
  "Checks if thing at point is keyword without an alias."
  (let ((sym (cider-symbol-at-point)))
    (and
+        (stringp sym)
         (cljr--keywordp sym)
         (not (string-match-p "::.+" (cljr--symbol-prefix sym))))))

I wonder if this will solve it.

[vemv]

Thanks!

Sounds likely, looking forward to your session

[vemv]

👋 Any luck with the QAing?

[vemv]

Update, the error two users facing was clojure-emacs/parseclj#35 so there isn't necessarily a bug on clj-refactor/refactor-nrepl's end

[yuhan0]

Yup, I've been using the latest refactor-nrepl and clj-refactor.el all week (including cljr-slash) without any issues.

[vemv]

Thanks! Between that and clojure-emacs/cider#3072 I don't fear anymore about these features working properly.