Stop adding local variables in S3 generics to method evaluation.
This very strange feature make reasoning about method code very hard. Can it be dropped without causing too much collateral damage?
See this R-devel thread and SVN revisions 85312, 85323, 85341. I don't see the packages listed as affected failing CRAN checks or being updated to fix this yet, but the work is evidently underway.
Endorsed by Luke
Stop adding local variables in S3 generics to method evaluation. This very strange feature make reasoning about method code very hard. Can it be dropped without causing too much collateral damage?