R CMD check fails with R-devel_2019-03-11

[ressy]

Running R CMD check chiimp_0.2.2.tar.gz fails on recent development versions of R:

...
* checking for portable file names ... OK
* checking for sufficient/correct file permissions ... OK
* checking whether package ‘chiimp’ can be installed ... ERROR
Installation failed.
See ‘/data/home/jesse/dev/chiimp-r-ver/chiimp.Rcheck/00install.out’ for details.
* DONE

Status: 1 ERROR
See
  ‘/data/home/jesse/dev/chiimp-r-ver/chiimp.Rcheck/00check.log’
for details.

The file chiimp.Rcheck/00install.out shows:

* installing *source* package ‘chiimp’ ...
** 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
** testing if installed package can be loaded from final location
** testing if installed package keeps a record of temporary installation path
ERROR: hard-coded installation path: please report to the package maintainer and use ‘--no-staged-install’
* removing ‘/data/home/jesse/dev/chiimp-r-ver/chiimp.Rcheck/chiimp’

This is causing CI via Travis to fail for the "devel" and "bioc-devel" jobs. it's fine on the latest stable release, though. Is this a bug in the development version or a more stringent check that didn't previously exist?

[ressy]

It's a more stringent check. It now looks for evidence of absolute paths hardcoded into the byte-compiled objects, and that actually does exist in the form of test_data$f.locus_attrs. That isn't even necessary to keep once the data is loaded, so a simple fix may be to just remove the path from the list entirely.

[ressy]

That looks like it'll work fine, but there are other problems with that same object, notably that sample.int no longer produces the same output given the same set.seed(0) call beforehand, and affects the RNG state for anything that follows. A simple fix for now may be to use RNGversion to control the behavior, like RNGversion(3.5.3).

[ressy]

Fixed in 515f8ed.