I’m not sure how to deal with this where I care about continued compatibility with versions of R before C++11 became the default (v3.6.0). If (minimum) C++11 support is needed it must be specified for 3.5.x, but must *not* be specified for 4.3.0 🤔 #rstats #cpp
Dirk Eddelbuettel @eddelbuettel
Jan 28
Most interesting. Just how I was musing (in the quoted tweet) that (too ?) many packages still hard-wire C++11 I just noticed that one of mine now gets warnings for this on the Vienna and Oxford @debian and @fedora r-devel systems of CRAN. #rstats #rcpp #cpp
1
2
Replying to @jonclayden
It's a rich topic. Did you see the thread on r-pkg-devel in the last few days? One could use configure as needed. What I like in the change I commented on and highlighted here is that the NOTE no longer leads people to blindly do s/CXX11/CXX17/ which is not a panacea either.

Feb 10, 2023 · 1:54 PM UTC

1
1
Replying to @eddelbuettel
Yes, indeed. I’d missed the thread, but have located it now – thanks for the pointer. I had a feeling more configure-fu was going to be the answer…
1
That is an option. I am leaning towards the "it is 2023 and R 4.3.* (soon)" of removing the compilation standard requirement for most of my packages -- but also just sent a PR to another project to roll from C++14 to C++17. So as always: it depends, and we have choices.
1