[R-pkg-team] Bug#980809: rmatrix: breaks autopkgtest of r-cran-glmmtmb on s390x

[Dirk Eddelbuettel]

On 14 February 2021 at 12:00, Graham Inggs wrote:
| glmmTMB upstream came up with a minimal reproducer (see forwarded bug):
| 
| stopifnot(require("testthat"),
|           require("glmmTMB"),
|           require("lme4"))
| data("Orthodont", package="nlme")
| fm1 <- glmmTMB(distance ~ age + (age|Subject), data = Orthodont)

(It seems clear you do not even need "testhat" for the two lines that
follow. On the other hand it uses "nlme" for the data. I guess that might get
pulled in from "lme4"....)
 
| I can confirm this fails on the s390x porterbox using our packages
| from unstable.

Ok. And I said all along I suspect that _one_ of the many packages along the
chain is likely lacking a rebuild. It is a long one, but a number of them are
required ones leaving TMB, lme4, nlme, RcppEigen, MASS, minqa, nloptr,
statmod, Rcpp.

  > db <- tools::CRAN_package_db()   # retrieve package info
  > res <- tools::package_dependencies("glmmTMB", recursive=TRUE) # defaults to 'hard depends'
  > res[[1]]
   [1] "methods"   "TMB"       "lme4"      "Matrix"    "nlme"     
   [6] "RcppEigen" "stats"     "graphics"  "grid"      "splines"  
  [11] "utils"     "parallel"  "MASS"      "lattice"   "boot"     
  [16] "minqa"     "nloptr"    "statmod"   "Rcpp"      "grDevices"
  > 
 
| Is it possible to try this using packages from R / CRAN?

Sure. Just fire up R on the box, don't install anything beyond r-base-core
(and maybe r-recommended accounting for a few of the ones above) and then say

  > install.packages("glmmTMB")

and then the two lines to test:

  > data("Orthodont", package="nlme")
  > fm1 <- glmmTMB(distance ~ age + (age|Subject), data = Orthodont)

I do not have access to any weird or exotic hardware.

Dirk

-- 
https://dirk.eddelbuettel.com | @eddelbuettel | edd at debian.org