Red squiggly in source buffer is gone with Clojure 1.10

[slipset]

Expected behavior

Whenever I do C-c C-c with a form that has an error, I get a red squiggly in my source buffer indicating where the error is.

Actual behavior

I’m not getting the red squiggly, but I get a rather friendly message in the *cider-repl..* buffer telling me that I have an error somewhere in my source file

Steps to reproduce the problem

Create a project with Clojure 1.10, jack-in, create a form with an error in it, evaluate it with C-c C-c. You will not have a red squiggly under the form with error, but you will get a message in the *cider-repl ...* buffer

Environment & Version information

Any recent version of Cider, Clojure 1.10.

Additional information

The messages I get in the repl are: for 1.10

Syntax error compiling at (core.clj:10:3).
      Unable to resolve symbol: somthing in this context

Wheras for 1.9.0 it’s

CompilerException java.lang.RuntimeException: Unable to resolve symbol: somthing in this context, compiling:(/Users/erik/Documents/github.com/refactor-nrepl-ns-kw-bug/src/foo/core.clj:10:3)

So a good starting point would be to look at this regex here:
https://github.com/clojure-emacs/cider/blob/master/cider-eval.el#L300

[bbatsov]

Yeah, you're right. That's broken because of the new message format.

[slipset]

So it seems like

(defvar cider-compilation-regexp '("Syntax error compiling at (\\(.*?\\):\\([[:digit:]]+\\):\\([[:digit:]]+\\))"
				 1 2 3))

works (at least for me (on my computer)). This does, however pose the question of backwards compatibility. If we simply change cider-compilation-regexp to this regex, then people working with older clojure versions will loose the red squiggly.

This leaves us with (at least) three options

1. Simply use this regex, and hope the population of devs using versions earlier than 1.10 is small and diminishing quickly
2. Make a monstrosity of a regex to cater for both
3. Make several entries in compilation-error-regexp-alist-alist and compilation-error-regexp-alist

I feel that my emacs-lisp-fu leaves me at 1), and I'd need very clear guidance if you want to go for 2) or 3)

[bbatsov]

Hmm, seems we dropped the ball on the original issue #2443 I'll put it here for posterity as I shared some thoughts about solving the problem there.