BUG: --autocorrect finds lint/if-nil-else in .cljc file reader conditionals even if it's not shown in normal mode

[Cyrik]

Me written this time, sorry for last time (also my assertion that :enabled got ignored was wrong. I didn't realize that lint/if-else-nil and lint/if-nil-else both exist and are different):

Splints --autocorrect finds and fixes issues in .cljc files that aren't issues in normal mode (and shouldn't be).

Repro src/example.cljc:

(ns example.app)

(defn pick [d]
  (if (zero? d)
    #?(:clj :clj-branch :default :default-branch)
    d))

normal run:

bb -Sdeps '{:deps {io.github.noahtheduke/splint {:mvn/version "1.24.0"}}}' \
   -m noahtheduke.splint src/

Linting took 4ms, checked 1 files, 0 style warnings

--autocorrect run:

bb -Sdeps '{:deps {io.github.noahtheduke/splint {:mvn/version "1.24.0"}}}' \
   -m noahtheduke.splint --autocorrect src/

Replacing
(if (zero? d) nil d)
 with
(when-not (zero? d) d)
in src/example.cljc:4

[NoahTheDuke]

Thanks for the report!

[NoahTheDuke]

I took some time over my lunch to look at this, and I think I have a solution, but it's not ideal. Because of the way Splint is built, paired with how --autocorrect uses rewrite-clj, I don't know that I can avoid dataloss of some kind. I can treat a given #? reader conditional as the appropriate branch (.clj -> :clj, etc), but then if there is an autocorrection covering that code, the chosen branch will be used instead of the original reader-conditional.

I think the safest course of action is to skip .cljc files in autocorrect. That's not ideal either but the autocorrection system is already not great (overwrites comments/formatting), so here we can at least avoid data loss.

[Cyrik]

Agreed, it's tricky to get this correct.
I'm not sure you need to go as far as fully disabling it on cljc. It might be enough if you disable if-nil-else and if-else-nil autocorrections in cljc files.
I'm guessing some other rules would also interact badly with it, but these two would be the most common bad cases.