For closure, I was advised by Professor Ripley to both pass "-std=f008" in Makevars and add "SystemRequirements: A version of gfortran supporting Fortran 2008" to the DESCRIPTION. CRAN accepted the package with those two changes. I also note that the R Daily News of 2023-02-03 [1] has the following entry under "R-devel": ?check?'s ?checking compilation flags in Makevars? has been relaxed to accept the use of flags such as ?-std=f2008? in ?PKG_FFLAGS?. Hopefully, this means future packages will have an easier time relying on more modern features, albeit they are 5 or 15 years old by now :). Thank you, Avi [1]: https://developer.r-project.org/blosxom.cgi/R-devel/NEWS/2023/02/02#n2023-02-02
On Mon, Jan 30, 2023 at 12:12 AM Avraham Adler <avraham.adler at gmail.com> wrote:
I would like to convert some of my hand-rolled Fortran code to Fortran intrinsics which entered the language in Fortran 2008, which should be extremely widely supported at this time. Specifically, I would like to use the log_gamma function and ieee_arithmetic functions and values. If it helps, the current package is at [1] and devel branch with new code is at [2]. If I understand it correctly, WRE allows for dependence on Fortran 2008 by explicitly passing -std=f2008 in PKG_FFLAGS (WRE section 1.2.3 Using F9x code [3]). However, checking a package as-cran with that addition delivers a warning (further confirmed by win-builder and rhub). My question is is it appropriate to submit a package with that warning and note it in the submission or is there a better way to express reliance on Fortran 2008. The implication from WRE 1.2.4 is that passing standards in Makevars is preferable to stating them in DESCRIPTION, at least as regards C++. What somewhat complicates matters is that if I build my package (Rtools43 on Windows 10) without passing "-std=f2008" it still builds properly and passes all the tests because Rtools43 is based on GCC12.2 which supports Fortran 2008 intrinsics. So I can probably submit it without the flag, but someone with a very old Fortran installation may suffer. I understand being told I should just address CRAN directly, but I thought I would try the collected institutional memory here first. Thank you, Avi [1] https://github.com/aadler/Delaporte [2] https://github.com/aadler/Delaporte/tree/devel [3] https://cran.r-project.org/doc/manuals/r-devel/R-exts.html#Using-F9x-code