[Rd] S3 generic/method consistency issue with R CMD check

[Drew Schmidt]

More on this.  It looks to me like there is some kind of issue
specifically with the qr and qr.* functions in R CMD check, as my last
email suggests.  Additionally, by changing the dispatch of the qr.*
functions from y to x (in this example), such as:


setGeneric(name="qr.Q",
  function(x, ...)
    standardGeneric("qr.Q"),
)

setMethod("qr.Q", signature(x="bar"),
  function(x, ...) { }
)


we avoid the WARNING.  So too does:


setGeneric(name="qr_Q",
  function(y, ...)
    standardGeneric("qr_Q")
)

setMethod("qr_Q", signature(y="bar"),
  function(y, ...) { }
)


avoid the WARNING.

This suggests to me that R CMD check is incorrectly using "qr"'s dispatch
for "qr.Q", even when we specify "qr.Q" as S4 generic with dispatch y. 
More to the point, it looks to me like R CMD check is applying S3 rules to
S4 methods for this specific set of base:: functions.

Not sure how else to interpret what I'm seeing here, but I still have
doubts as to my understanding of this issue.  Would very much appreciate
some insight.

Apologies in advance if I botch the reply threading; I am using a very
limited webmail client and I'm not sure if this will work properly.

--
Drew Schmidt
National Institute for Computational Sciences
University of Tennessee