Fresh CRAN issue on recent clang++ version

[eddelbuettel]

Via two emails from CRAN:

Dear maintainer,

Please see the problems shown on
https://cran.r-project.org/web/checks/check_results_RcppFastAD.html.

Please correct before 2024-09-27 to safely retain your package on CRAN.

The CRAN Team

and

This looked like clang being picky, but it flagged a bigger problem in a
submission.

The overdue but supposedly imminent LLVM 19 has a different wording and
reports a different flag:

../inst/include/fastad_bits/forward/core/forward.hpp:107:63: warning:
passing no argument for the '...' parameter of a variadic macro is a C++20
extension [-Wc++20-extensions]

I will try to reproduce on the weekend and look into what may need fixing. @JamesYang007 if you have moment, take a peek too. We should be able to get clang18 in Ubuntu 25.04 containers (see here) and likely in Debian unstable. (Yup, clang-18 at version 18.1.8 in Debian unstable, easily accessible via r-base Rocker container.)

[eddelbuettel]

Can confirm that a) under clang++-18 (as in Debian) we see the warnings and that b) it goes away when we make it C++20:

root@37586cbd9d9d:/work# ./cleanup 
root@37586cbd9d9d:/work# R CMD INSTALL .
* installing to library ‘/usr/local/lib/R/site-library’
* installing *source* package ‘RcppFastAD’ ...
** using staged installation
** libs
using C++ compiler: ‘Debian clang version 18.1.8 (9)’
using C++20
clang++-18  -std=gnu++20 -I"/usr/share/R/include" -DNDEBUG -I../inst/include -I'/usr/lib/R/site-library/Rcpp/include' -I'/usr/lib/R/site-library/RcppEigen/include'     -fpic  -O2 -Wall -pipe -pedantic -Wno-unused-but-set-variable -fdiagnostics-color=always -Wformat    -c RcppExports.cpp -o RcppExports.o
clang++-18  -std=gnu++20 -I"/usr/share/R/include" -DNDEBUG -I../inst/include -I'/usr/lib/R/site-library/Rcpp/include' -I'/usr/lib/R/site-library/RcppEigen/include'     -fpic  -O2 -Wall -pipe -pedantic -Wno-unused-but-set-variable -fdiagnostics-color=always -Wformat    -c black_scholes.cpp -o black_scholes.o
clang++-18  -std=gnu++20 -I"/usr/share/R/include" -DNDEBUG -I../inst/include -I'/usr/lib/R/site-library/Rcpp/include' -I'/usr/lib/R/site-library/RcppEigen/include'     -fpic  -O2 -Wall -pipe -pedantic -Wno-unused-but-set-variable -fdiagnostics-color=always -Wformat    -c linear_regression.cpp -o linear_regression.o
clang++-18  -std=gnu++20 -I"/usr/share/R/include" -DNDEBUG -I../inst/include -I'/usr/lib/R/site-library/Rcpp/include' -I'/usr/lib/R/site-library/RcppEigen/include'     -fpic  -O2 -Wall -pipe -pedantic -Wno-unused-but-set-variable -fdiagnostics-color=always -Wformat    -c quadratic_expression.cpp -o quadratic_expression.o
clang++-18 -std=gnu++20 -Wl,-S -shared -L/usr/lib/R/lib -Wl,-z,relro -o RcppFastAD.so RcppExports.o black_scholes.o linear_regression.o quadratic_expression.o -L/usr/lib/R/lib -lR
installing to /usr/local/lib/R/site-library/00LOCK-work/00new/RcppFastAD/libs
** R
** inst
** byte-compile and prepare package for lazy loading
** help
*** installing help indices
** building package indices
** testing if installed package can be loaded from temporary location
** checking absolute paths in shared objects and dynamic libraries
** testing if installed package can be loaded from final location
** testing if installed package keeps a record of temporary installation path
* DONE (RcppFastAD)
root@37586cbd9d9d:/work# 

(This also adds -Wno-unused-but-set-variable to suppress some noise we get from RcppEigen.)

[eddelbuettel]

I am planning to make a new CRAN release, simply setting C++20 as the compilation standard. That forward path code is nifty and reasonably complex so best to leave it as is for now.

[JamesYang007]

Thanks for resolving this issue! Sorry I've been away on vacation.

[eddelbuettel]

No problem. It's not 'really resolved' more circumvented. If you need a little puzzle take a look at if / how we could make it C++17 compliant. Otherwise C++20 is an easy-ish way out. (And I proceeded because of the deadline, the change to C++20 is of course not meant to be final or definitive.)

[JamesYang007]

This looked like clang being picky, but it flagged a bigger problem in a
submission.

The overdue but supposedly imminent LLVM 19 has a different wording and
reports a different flag:

../inst/include/fastad_bits/forward/core/forward.hpp:107:63: warning:
passing no argument for the '...' parameter of a variadic macro is a C++20
extension [-Wc++20-extensions]

Grr this is annoying. I have a quick fix (though not as elegant as I would like). I'll submit a PR and link it.