CRAN v1.6.12

Is there a problem with the following situation ?

> QupstrUnit <- NULL
> match.arg(arg = QupstrUnit, choices = c("mm", "m3", "m3/s", "l/s", "L/s"))
[1] "mm"

It's maybe exaggerated but we should prevent that a NULL value make a choice be default that it is not known by the user. I suggest to adopt one of these solutions:

• if(is.null(QupstrUnit)) stop("QupstrUnit should be one of \"", paste(choices, collapse = "\", \""), "\"")
• if(is.null(QupstrUnit)) warning("'QupstrUnit' has been redefined to \"mm\"")

Using tolower(QupstrUnit) fixed it and it allows to be flexible in the respect of the case of characters.

> match.arg(arg = tolower(NULL), choices = c("mm", "m3", "m3/s", "l/s"))
Error in match.arg(arg = tolower(NULL), choices = c("mm", "m3", "m3/s",  : 
  'arg' doit être un de “mm”, “m3”, “m3/s”, “l/s”
> match.arg(arg = NULL, choices = c("mm", "m3", "m3/s", "l/s"))
[1] "mm"

I wasn't shocked by the previous behavior, because that's how the match.arg() function usually works.

And it is a normal behavior in R to return the default value if the NULL value is used.

The check crashes with this message:

> if (length(devtools::check_failures(path = Sys.getenv("BUILD_LOGS_DIR"), note = FALSE)) > 0) stop()
Erreur : 'check_failures' n'est pas un objet exporté depuis 'namespace:devtools'
Exécution arrêtée

Really? WTF!

I have seen. Has the version of R changed on the CI?

Nope. Nothing has changed. But it really seems to be an issue with the R version installed on the server.

I'm currently updating "patched version" because it still was v4.0.0 that I had compiled myself last year. I added the cran repository on my debian and update to v4.0.5 to not depend on a compiled version of R from source which is always a real challenge to manage...

I keep you in touch.

Something also seems to have affected airGRdatassim.

I said bullshit, airGRdatassim CI runs on the Lyon's runner...

Found why in https://cran.r-project.org/web/packages/devtools/news/news.html:

devtools 2.4.0

Breaking changes and deprecated functions

The check_results() function has been removed. It was not used by any CRAN package, and much better alternatives are available in the rcmdcheck package.

I'm searching for an alternative in rcmdcheck package...

I really didn't need that at this moment...

Temporarily, the easiest way is probably to install the previous version of 'devtools' on the server, isn't it?

It seems that I have solved this issue (and by the way again simplified the CI process): https://gitlab.irstea.fr/HYCAR-Hydro/airgr/-/commits/86-gitlab-ci-correctly-handle-warning-and-note-in-checks/.gitlab-ci.yml

Like the French people say: "Encore une victoire de canard!"

Also change it in the README file! #114 (closed)

merged

mentioned in commit 795efbb6

I'm not sure that that test should be remove.

It shall still have a test describing the error that should have the user if he enters a wrong unit.