Help packages transition away from C++11

[eddelbuettel]

As of a reverse depends run started 2025 Sep 09 the following 226 packages fail when switching to Armadillo 15.0.2:

abclass, acebayes, adaHuber, ADMM, ahMLE, alpaca, ANN2, AntMAN, ARCokrig, autoRasch, bama, bayescopulareg, bayeslm, BayesReversePLLH, BayesRGMM, bayMDS, BClustLonG, biClassify, bife, bisque, bmabasket, BNPmix, BOSSreg, bpgmm, brmsmargins, BSPBSS, Buddle, BVSNLP, catSurv, CB2, CERFIT, CircSpaceTime, cIRT, clhs, clusterHD, cmenet, combinIT, COMIX, communication, CoOL, corrcoverage, covdepGE, covTestR, CovTools, CoxPlus, coxrt, cPCG, cpss, curveDepth, dcortools, DCSmooth, dcurver, deforestable, DepthProc, dga, DGM, didimputation, dina, DriftBurstHypothesis, DSSP, DtD, DySS, edina, elmNNRcpp, em, errum, evgam, ExhaustiveSearch, expSBM, FamilyRank, FarmTest, fasterElasticNet, fastGLCM, fastpos, fctbases, FGLMtrunc, filling, fipp, flamingos, FMCCSD, fmerPack, fmf, fourPNO, FRESHD, FSSF, gbp, GCSM, geoFKF, geoFourierFDA, GiRaF, gjam, glamlasso, GMKMcharlie, GofKmt, GPM, gridOT, gsynth, GxEScanR, HACSim, hdbm, higlasso, HMB, HMMEsolver, HTLR, iccbeta, ILSE, imbalance, IncDTW, infinitefactor, interflex, intkrige, intsurv, IrishDirectorates, ivsacim, ivx, JFM, JMbayes, jmcm, jmotif, KernelKnn, KoulMde, LassoGEE, latentgraph, ldsep, ldsr, linconGaussR, Lmoments, lodr, lowmemtkmeans, ltsspca, lvmcomp, manifold, MAPITR, matchingR, mclustcomp, mcmcse, mDAG, meshed, meteorits, milr, mirtjml, mixedCCA, modeLLtest, multivar, netClust, netcmc, netcontrol, NetworkDistance, networkR, nlmeVPC, NPCirc, ohoegdm, parallelDist, parglm, pdSpecEst, penalized, penPHcure, pgsc, pleiotest, PLSiMCpp, PPMR, PQLseq, pqrfe, PRDA, qris, ratematrix, rbridge, RcppCensSpatial, RcppHMM, Rdimtools, registr, Riemann, RiemBase, Rlinsolve, RLumCarlo, RLumModel, rmarchingcubes, robustcov, rocTree, rofanova, roptim, ROptSpace, rotations, rrpack, rSpectral, rsvddpd, SAGMM, samurais, sasfunclust, SBmedian, scanstatistics, scINSIGHT, sdpdth, sfcr, SFS, ShrinkCovMat, slasso, SMMA, SMME, smmR, SNPknock, spFW, SPlit, spnn, sprintr, Sshaped, SSOSVM, SuperpixelImageSegmentation, surbayes, symphony, T4cluster, tabularMLC, TAG, tci, TDAkit, telefit, textmineR, tfarima, tmvtnsim, tree.interpreter, treenomial, tvR, UtilityFrailtyPH12, VMDecomp, VUROCS, WeibullR

In most cases this is due to C++11 being forced, which leads to Armadillo 15.0.2 erroring out. From the first package:

using C++ compiler: ‘g++ (Debian 14.3.0-5) 14.3.0’
using C++11
ccache g++ -std=c++11 -std=gnu++11 -I"/usr/share/R/include" -DNDEBUG -I../inst/include -I'/usr/local/lib/R/site-library/Rcpp/include' -I'/usr/local/lib/R/site-library/RcppArmadillo/include'    -fopenmp -fpic  -\
g0 -O3 -Wall -pipe -Wno-unused  -fext-numeric-literals -DBOOST_NO_AUTO_PTR -DARMA_USE_CURRENT   -c RcppExports.cpp -o RcppExports.o
In file included from /usr/local/lib/R/site-library/RcppArmadillo/include/RcppArmadillo.h:29,
                 from ../inst/include/abclass/Abclass.h:21,
                 from ../inst/include/abclass.h:22,
                 from RcppExports.cpp:4:
/usr/local/lib/R/site-library/RcppArmadillo/include/RcppArmadillo/version/arma.h:40:6: error: #error "Do not select 'current' with C++11 (or older) as 'current' requires C++14 or newer."
   40 |     #error "Do not select 'current' with C++11 (or older) as 'current' requires C++14 or newer."
      |      ^~~~~
In file included from /usr/local/lib/R/site-library/RcppArmadillo/include/current/armadillo:27,
                 from /usr/local/lib/R/site-library/RcppArmadillo/include/RcppArmadillo/interface/RcppArmadilloForward.h:57,
                 from /usr/local/lib/R/site-library/RcppArmadillo/include/RcppArmadillo.h:32:
/usr/local/lib/R/site-library/RcppArmadillo/include/current/armadillo_bits/compiler_check.hpp:82:4: error: #error "*** C++14 compiler required; enable C++14 mode in your compiler, or use an earlier version of A\
rmadillo"
   82 |   #error "*** C++14 compiler required; enable C++14 mode in your compiler, or use an earlier version of Armadillo"
      |    ^~~~~

This shows one error from RcppArmadillo anticipating the following error from Armadillo. We will try to confirm which 226 package have that issue and see about contacting maintainers.

Edit 2025-09-11: One package identified as misclassified, it was lacking a dependency. The remaining 226 are confirmed.

[coatless]

Excellent, off to test out some of my new toys! (maybe develop newer toys... 😃 )

[eddelbuettel]

If/when we both have a moment one of these days may might sense to gab. There are a few moving pieces. Kurt very kindly put a check into r-devel which they have enabled on the incoming checker 'r-devel' (see here for the commit) and Uwe now includes a short snippet I once wrote in reply (and edited once or twice, below 'details'). This seems to work for at least a few packages. When I check the 'cran' organization here for string 'DARMA_USE_CURRENT' the number of hits increases steadily day by day. That is step one: getting packages to migreate to 'current'.

Another step is this issue. The list above is from a reverse-depends check I did where I forced '-DARMA_USE_CURRENT' globally so those stuck with C++11 will fail. I am trying to shell script something now to ascertain that there we no other failures.

Slowly but steadily we will get there. New ideas and suggestions welcome!

Text being sent out, originally from an email:

Yes, please see the top-level issue #475 (and its sub-issues) at

 https://github.com/RcppCore/RcppArmadillo/issues/475

For your package all you should hopefully need is to add a line to
src/Makevars and src/Makevars.win (or expand one if you have it) so
PKG_CXXFLAGS also contains the term -DARMA_USE_CURRENT so that the
current Armadillo is used via RcppArmadillo.

Most packages do not need changes, a few may now get warnings about
deprecated code. Please get in touch (via an issue at the
RcppArmadillo repo, preferably) if you need help. You could also
choose to define -DARMA_USE_LEGACY but we want to (eventually) move
away from this version.

Let me know if you need other help. A new version 15.0.2-1 of package
RcppArmadillo is now at CRAN and should work for many packages, yours
included. Until all CRAN binaries are rebuilt consider installing a
binary from https://rcppcore.r-universe.dev/RcppArmadillo -- and
thank you in advance for moving to (Rcpp)Armadillo 15.0.* which we
plan to make the default once all (or most) packages needing touchups
have done so.

[eddelbuettel]

In the meantime we refined the analysis further from both the CRAN side thanks particularly to @kurthornik who also wrote the following in email

I will shortly reach out to the maintainers of all packages asking for
exactly C++11 and linking to RcppArmadillo to stop doing the former.

There are currently 221 such packages: computing the numbers of
recursive revdeps for these finds

  0   1   2   3   4   7   8  10  11  13  15  16  32  41 
180  20   5   2   4   1   1   1   1   1   2   1   1   1 

so removing the legacy fallback would make the 221 packages and these
revdeps uninstallable, which is not so good.  Ideally we'd get the 41
packages with revdeps updated asap, but that's not realistic, so we
should perhaps aim for those with more than two:

       GiRaF       rrpack    ROptSpace     WeibullR       parglm    textmineR 
           3            3            4            4            4            4 
   Rdimtools         ADMM        evgam   mclustcomp       roptim parallelDist 
           7            8           10           11           13           15 
   penalized       mcmcse     Lmoments      dcurver 
          15           16           32           41 

and I will plan to work on PRs or emailed patched for these sixteen packages over the next few days.

[eddelbuettel]

The next (six today, not five) with two recursive reverse depends are

                  DepthProc                      NPCirc SuperpixelImageSegmentation 
                          2                           2                           2 
                  T4cluster                        clhs               lowmemtkmeans 
                          2                           2                           2 

and I will add those next.

[eddelbuettel]

And now added a few more with one reverse dependency.