(potential?) false positive redundant-do when (do ,,,) contains one or more #_,,, ignored forms

[sumbach]

version

clj-kondo v2022.06.22

platform

Native clj-kondo on macOS 10.12 (Sierra)

problem

clj-kondo appears to ignore #_,,, ignored forms within the body of (do ,,,) forms for the purpose of determining if the (do ,,,) form is redundant.

repro

clj-kondo --lint - <<< '(do #_(println "42") 42)'
<stdin>:1:1: warning: redundant do
linting took 46ms, errors: 0, warnings: 1

expected behavior

Let me start by acknowledging this is an odd case and you could argue that leaving #_,,, ignored forms in the code just isn't good style. For better or worse, we have several instances like this in our codebases, and we are able to disable the linter warning with #_{:clj-kondo/ignore [:redundant-do]}.

I wanted to open a conversation here in case other people encounter the same behavior.

I suspect this may be tough to change in clj-kondo (assuming you even wanted to) because #_,,, is probably ignored/discarded relatively early in the process. Further, it might be necessary to distinguish between #_:clj-kondo/ignore/#_{,,, :clj-kondo/ignore [,,,] ,,,} forms (which probably should be ignored for redundant-do purposes) and other #_,,, forms that should not.

See also #1453 -- I think we have a couple places in our codebase like (doto (Foo.) #_(.doSomething)) that would be flagged redundant-doto without a change in the handling of #_,,, within (doto ,,,) forms.

I'm potentially interested in opening a PR, but I think the next step is to gauge interest and discuss what, if any, clj-kondo changes are desired.