[anyNA()] now resolving to backports::anyNA (etc)

[MichaelChirico]

Our CI to check man/ comments are in sync with the source code just started failing:

https://github.com/r-lib/lintr/actions/runs/17746754124/job/50433547555?pr=2919

-\code{\link[=anyNA]{anyNA()}} exists as a replacement for \code{any(is.na(x))} which is more efficient
+\code{\link[backports:anyNA]{backports::anyNA()}} exists as a replacement for \code{any(is.na(x))} which is more efficient

It seems something changed in {roxygen2} to cause this -- our most recent run on main 3 weeks ago didn't have this issue:

https://github.com/r-lib/lintr/actions/runs/17204746502/job/48802781565

The passing job, per the logs, was 7.3.2, while the failing one is 7.3.3.

We have {backports} in Imports:, but we don't import anything (e.g. any methods); currently, it's only used to import one object in .onLoad:

https://github.com/r-lib/lintr/blob/33efecf7f4aee58fbc9d4f2e1f3fda5ae908ed42/R/zzz.R#L311

So it seems to me it shouldn't resolve to the ?backports::anyNA page.

[MichaelChirico]

From the NEWS, #1612 looks likely as a culprit

[MichaelChirico]

Looks like I was only reporting the first such instance as our CI fails there instead of gathering all the differences at once; it also affects other {backports}-{base} collisions, e.g.

-functions \code{\link[=paste]{paste()}} and \code{\link[=paste0]{paste0()}}. This is because
+functions \code{\link[backports:paste]{backports::paste()}} and \code{\link[backports:paste0]{backports::paste0()}}. This is because