Make rextendr packages CRAN compatible by default

[JosiahParry]

As a followup from #393 and comments from @CGMossa and @Ilia-Kosenkov, this PR makes adds --print=native-static-libs to any existing RUSTFLAGS ensuring that they are not overwritten.

[JosiahParry]

Portable Makefiles do not use GNU extensions such as +=, :=,

welp! back to the drawing board!

[Ilia-Kosenkov]

Can't you do something like RUSTFLAGS = $(RUSTFLAGS) --whatever?
UPD: ChatGPT suggests this (did not test it myself)

RUSTFLAGS := $(RUSTFLAGS) --print=native-static-libs
export RUSTFLAGS

[JosiahParry]

@Ilia-Kosenkov @CGMossa

I've taken this as an opportunity to refactor the way in which Makevars are done.

Summary

This PR now removes use_cran_defaults() and instead, uses a single Makevars and Makevars.in to support CRAN environment and regular development environment. This is done in the following way:

• Makevars & Makevars.win are now Makevars.in and Makevars.win.in
• The configure script generates the Makevars and Makevars.win files substituting the @CRAN_FLAGS@ this approach is taken from data.table see https://github.com/Rdatatable/data.table/blob/master/src/Makevars.in. It also deletes the src/Makevars and src/Makevars.win before generating the respective files
• Now src/Makevars{.win} are in the .Rbuildignore and .gitignore when using use_extendr() ensuring that the file is always rebuilt
• When NOT_CRAN=false the vendor.tar.xz is unbundled and the flags -j 2 --offline are set

To ensure all of this is done correctly, two new tests are added which run rcmdcheck::rcmdcheck() and rcmdcheck::rcmdcheck(env = c("NOT_CRAN"="false")) to check both scenarios.

The function use_cran_defaults() has been removed.

Next steps

If you both approve of this direction, I would like to make the following changes

Within this PR

• remove vignettes/articles/cran-compliance.qmd as an alternative i am happy to write a standalone help doc e.g. ?rextendr::cran
• remove mentions of use_cran_defaults() from the NEWS.md as the function will have been introduced and removed before a new CRAN release

Follow up PRs

• write a function to generate an inst/AUTHORS file as was done by @DavZim in {rtiktoken} see https://github.com/DavZim/rtiktoken/blob/master/inst/AUTHORS
• write a page in the user guide on publishing to CRAN

[Ilia-Kosenkov]

I still believe checking NOT_CRAN for anything other than true is invalid. NOT_CRAN is an optional env var that you can set to, usually, true to signify that you are not on CRAN. In all other scenarios, you treat is as if you are on CRAN. What does NOT_CRAN=false mean? When do you expect to set this?

[JosiahParry]

Thanks @Ilia-Kosenkov, that's fine. That can be addressed. But are you okay with this plan (assuming we change the NOT_CRAN to check based on a true value)? Having one ring Makevars to rule them all?

[DavZim]

What happens when NOT_CRAN is true but I don't have vendor/ only vendor.tar.xz? Ie I installed the package from GitHub or from local.

Is the archive unpacked correctly?
Btw, once this is done, im happy to upgrade the other blog accordingly.

[JosiahParry]

@DavZim the Makevars is only ran when building the package from source. This template will unpack the archive if NOT_CRAN is unset and the vendor.tar.xz file exists. If NOT_CRAN=true the case when using devtools etc then the archive is not unpacked.

[JosiahParry]

I've moved this as a draft so I can adjust the configure scripts to set CRAN_FLAGS when NOT_CRAN is unset and a vendor.tar.xz folder is present.

[codecov]

Codecov Report

All modified and coverable lines are covered by tests ✅

Project coverage is 80.12%. Comparing base (d58284b) to head (d2db22e).
Report is 1 commits behind head on main.

Additional details and impacted files

Files with missing lines	Coverage Δ
R/cran-compliance.R	78.37% <ø> (+2.72%)	⬆️
R/use_extendr.R	98.98% <100.00%> (+0.05%)	⬆️