Feature/module load #353
Feature/module load #353
Conversation
|
Looks good to me. @kevinushey ? |
|
There is still some headscratchery in other places. Eg We need a better tool chain to run (and report) these tests. Is Gabor done yet? ;-) |
| @@ -67,7 +67,9 @@ loadModule <- function( module, what = character(), loadNow, | |||
| loadNow <- !is(loadM, "error") | |||
| } | |||
| if(loadNow) { | |||
| .botched <- isBotchedSession() | |||
| ## .botched <- isBotchedSession() | |||
| .botched <- FALSE | |||
There was a problem hiding this comment.
Do we know why we had this check?
The comment above isBotchedSession():
## the following items are to get around some insanity in the
## CMD check of packages using Rcpp that dies in loadModule()
## because some code somewhere can't find the methods package
isBotchedSession <- function()
! ("package:methods" %in% search())
This, at least to me, implies that there is code within Rcpp modules that requires methods not only to be loaded but also attached; do we have some way of verifying that Rcpp modules work when methods is loaded but not attached?
There was a problem hiding this comment.
That is why I brought this to the rcpp-core list and John -- AFAIK he added it.
This contains code emailed in by Kurt Hornik to catch up with changed on the R side.
It has been tested against reverse dependencies on both r-release and r-devel and it appears that it is not creating any new issues. There are one or two breakages we have not seen before but they may be tickled by something else as well.