Error: SETLENGTH() cannot be applied to an ALTVEC object.

[HughParsonage]

On Linux I get an error in a package which imports data.table for R-dev only:

# Minimal reproducible example

out <- data.table(tex_group = c(0L, 1L, 1L))
out[, z1 := .GRP, by = "tex_group"]
#> Error in `[.data.table`(out, , `:=`(z1, .GRP), by = "tex_group") : 
#>   SETLENGTH() cannot be applied to an ALTVEC object.
#> Calls: [ -> [.data.table
#> Execution halted

$ Rscript -e 'sessionInfo()'
R Under development (unstable) (2018-05-02 r74679)
Platform: x86_64-pc-linux-gnu (64-bit)
Running under: Ubuntu 14.04.5 LTS
Matrix products: default
BLAS: /home/travis/R-bin/lib/R/lib/libRblas.so
LAPACK: /home/travis/R-bin/lib/R/lib/libRlapack.so
locale:
 [1] LC_CTYPE=en_US.UTF-8       LC_NUMERIC=C              
 [3] LC_TIME=en_US.UTF-8        LC_COLLATE=en_US.UTF-8    
 [5] LC_MONETARY=en_US.UTF-8    LC_MESSAGES=en_US.UTF-8   
 [7] LC_PAPER=en_US.UTF-8       LC_NAME=C                 
 [9] LC_ADDRESS=C               LC_TELEPHONE=C            
[11] LC_MEASUREMENT=en_US.UTF-8 LC_IDENTIFICATION=C       
attached base packages:
[1] stats     graphics  grDevices utils     datasets  methods   base  

Likely culprit? wch/r-source@e0b8474

[mattdowle]

Yes exactly: that commit made to R-devel yesterday. CRAN maintainers let me know too. Luke made that change at the same time as data.table was accepted on CRAN so we made it on just in time. Will need to liaise with Luke.
Expected 1.11.2 to need to be on CRAN soon anyway.

[mattdowle]

Your repex doesn't seem quite right. Column tex_group isn't in that DT. And I'd expect to see a sequence vector like 1:10 or maybe rep(1L,10L) somewhere. c(01, 1L, 1L) would be a regular length-3 vector as far as I know and unaffected by ALTREP.

[mattdowle]

All PRs will fail now because of this due to all PRs testing against latest nightly R-devel snapshot. One option is to turn off r-devel testing on AppVeyor for now. Let's see in a few days, depending on how hard it turns out to be.

Looks like AppVeyor has updated to R 3.5.0 now and is passing for r-release, so that's good.

[HughParsonage]

@mattdowle Sorry for the sloppy copy paste. Fixed now. Here's the travis build failure: https://travis-ci.org/HughParsonage/r-dev-vs-release/jobs/373807347

[HughParsonage]

I suspect it's .GRP rather than c(0L, 1L, 1L) that's affected.

[HughParsonage]

Great! I know it's more work for you, but could I request that you re-release to CRAN asap? This issue will prevent all packages that import data.table (including 4 of mine :D) from being updated. (Unless you know there are likely to be future commits to R-dev that might cause failures.)

[mattdowle]

@HughParsonage For completeness of issue log, done: v1.11.2 is now on CRAN. I believe they are accepting updates and new submissions now that use data.table. Let me know if not.