Dots not passed to install.packages

[infotroph]

This is a new report for attention, but I'm re-reporting the same issue as #2154, which was closed as supposedly fixed but I still see the reported buggy behavior when running the latest devtools (4cbed16).

I've checked that my versions of remotes (3fd1845e) and ellipsis (77387d25) are up to date as well; am I missing something or was it not in fact fixed?

[jimhester]

It was fixed, and I cannot reproduce it, sorry.

[infotroph]

Looks like @danarmak nailed a key point: apparently this only happens when all deps are up to date:

docker run --rm -it -e PASSWORD=foo rocker/tidyverse:devel R

remotes::install_github(
	c("r-lib/devtools", "r-lib/remotes", "r-lib/ellipsis"),
	upgrade = "always")

usethis::create_package("demo")
install.packages("plotrix") # usethis refuses to add an uninstalled dep
usethis::use_package("plotrix")

# fails
devtools::install_deps(Ncpus = 2, upgrade = "always")
# succeeds, nothing to do
devtools::install_deps(upgrade = "always")
# succeeds, nothing to do
remotes::install_deps(Ncpus = 2, upgrade = "always")

remove.packages("plotrix")

# succeeds, reinstalls plotrix
devtools::install_deps(Ncpus = 2, upgrade = "always")

# fails again
devtools::install_deps(Ncpus = 2, upgrade = "always")

[jimhester]

Sorry, this docker file does not reproduce the error, are you restarting R after you install devel devtools? With the devel version of devtools this is a warning, not an error, as intended.

[jimhester]

Hmm, it seems 674b3f8 was not included in the 2.2.1 release, I will do a new release shortly to include it

[jimhester]

CRAN is not currently accepting packages until the 6th of January, so I cannot submit an update of devtools until then.

[infotroph]

OK, yes, I can confirm I see a warning not an error when using master, and I admit I wasn't checking carefully for warning vs. error before.

And I understand that y'all added ellipsis to try to mitigate a very real general usage hazard (that arguments unexpectedly swallowed by ... are insidious and hard to debug).

But:

1. It still seems buggy to accept ... when dependencies need updating but complain about it when they're up to date.

2. I don't think "a warning not an error" resolves the issue here, because the only purpose of ... in this function is to be passed onward for ultimate use by install.packages. If ellipsis can't check that properly, then either ellipsis should skip checking here or install_deps should stop claiming to accept .... It seems counterproductive to issue warnings against arguments that are used exactly as documented.

[jimhester]

If you don't like the behavior you can set options(devtools.ellipsis_action = rlang::signal) and you will receive no output as detailed in NEWS.md

[lock]

This old issue has been automatically locked. If you believe you have found a related problem, please file a new issue (with reprex) and link to this issue. https://reprex.tidyverse.org/