settings not WAI in package sub-directory?

[MichaelChirico]

In Rdatatable/data.table@5e36615, I tried to put the .lintr config into a non-standard location and point to it in CI with the env approach mentioned in ?read_settings: R_LINTR_LINTER_FILE: .ci/.lintr. Also tried with .ci/.lintr.R (the exact file, not relying on auto-guessing the R extension).

However, this failed to get picked up, as seen by the 1000s of lints picked up in the associated job: https://github.com/Rdatatable/data.table/actions/runs/7523183440/job/20476211212?pr=5908

Am I doing something wrong?

I've also tried running options(lintr.linter_file = ".ci/.lintr") (and .ci/.lintr.R) to no avail. I've also print-debugged that file.exists(".ci/.lintr.R"), so the file is not in the wrong location / omitted for being in a hidden directory.

Currently, I'm needing to run lintr:::read_settings(".ci/lintr"); lint_package(parse_settings = FALSE) to accomplish what I intended. Really not pretty.

PS it might help for verbose=TRUE to report some info in read_settings() as well, e.g. which file was chosen, maybe even which settings are read as which values.

[AshesITR]

Any reason for using basename(...) here?

find_local_config(path, basename(linter_file))

[MichaelChirico]

I assume you're referring to:

lintr/R/settings_utils.R

Line 78 in bad1632

find_local_config(path, basename(linter_file)),

IINM that entered the package long ago before the current form, then survived in #2205 without being re-examined:

1a7114b

But in short I don't know for sure what the aim is there...

[AshesITR]

IINM removing the call to basename() should fix this issue.