Investigate why the change in R-devel on 18 Sep 2015 caused 2 of data.table's non-ASCII tests (1092 & 1093) to fail on R-devel on Windows. data.table v1.9.4 on CRAN was very stable in 'OK' state for nearly a year before this fail. Decided it's good to have non-ASCII tests in a separate package just to take a belt-and-braces approach.
See https://developer.r-project.org/blosxom.cgi/R-devel/NEWS
The change to match() is most likely I'd guess.