R CMD build

Prof Brian D Ripley writes:

On Wed, 29 Mar 2000, Paul Gilbert wrote:

Well, are there \alias entries in the Rd files?

No. I guess that would be it. Should I have an empty \alias entry if the
object does not have an alias, or should should I just live with the
warnings?

You should have an alias for each function documented there, even if
there is only one. The object name is itself an alias.  The \name
field is the help file name, not the object name (although 90+% of
them are the same), and it is the \alias's that gives the object
names,

(Not sure about the 90+ percent :-))

The convention is to put a \alias for the method in the help file for

the generic.
Ok, but I guess this won't work if the generic is the base and my method
is in my library?

Sure (and that is part of the debate). Either ignore them, or have a
.Rd with just a very brief description of these (e.g. just their
usages).

4/ Checking for undocumented objects fails with an ERROR when

checking

an (undocumented) assignment object ( "input.series.names<-" ).

I think it works if you document it, though, and the error is correct,
surely?  Do you mean it gives ERROR and then stops rather than warns?

Yes, it should warn me about this, but it gives ERROR and stops. (If I
run it with my makefile it returns code 1.) The message is:
in eval(expr, envir, enclos) : couldn't find function
"input.series.names<-"

Looks like a bug, then.

Hmm.  Can you (Paul) perhaps give a small example I could reproduce?  I
just tried the following:  let the R code contain 

	"foo<-" <- function(x, value) NULL

Then if an Rd file has \alias{foo<-}, everything is o.k., if not,
"foo<-" is shows as undoc object.  How do you get the error?