Check with the oldest supported R version

[IndrajeetPatil]

The oldest R version we can check is R 3.4 (r-lib/pak#417 (comment)).

Given that we can't check on older R versions (or is there a way?) and thus can't confirm support for that R version either way, should we be bumping the minimum needed R version to 3.4? (Currently, we support R >= 3.2.)

[IndrajeetPatil]

Given that we can't check on older R versions (or is there a way?) and thus can't confirm support for that R version either way, should we be bumping the minimum needed R version to 3.4? (Currently, we support R >= 3.2.)

@MichaelChirico Any thoughts on this?

[MichaelChirico]

Is pak a requirement for GHA or one of our upstream dependencies?

[IndrajeetPatil]

It's a requirement for GHA, but, IINM, that's the only way we have to know if things are working as expected on the older versions. Since CRAN doesn't test past oldrel-1. So I am not sure how we can confirm that we indeed support R 3.2, in the sense that most features work as expected for this version.

[MichaelChirico]

It's not too difficult to install 3.2 locally and test :)

I am basically fine keeping the old version until it's proven not to work.

OTOH, for lintr, its main use case is package development, and people submitting to CRAN are technically supposed to be testing their packages with r-devel. So it seems more reasonable to force newer versions of R, and an alignment to GHA makes sense if folks are relying on GHA to test versions of R unavailable to them locally during development.

So, I'm fine either way. @AshesITR any thoughts?

[IndrajeetPatil]

I am basically fine keeping the old version until it's proven not to work.

Okay, I am fine with that.

I was just curious to hear your thoughts on this, apropos testing older R versions on GHA.