Package `Rmosek` on CRAN is actually a stub, essentially:
```r
mosek <- function(...) message("Download binary package Rmosek from
www.mosek.com")
```
And "www.mosek.com" provides a meaty `Rmosek::mosek(...)` that does
actual work.
Package CVXR "Enhances" `Rmosek` and uses, for "special situations":
```r
if (requireNamespace("Rmosek", quietly = TRUE)) {
? Rmosek::mosek(x)
} else {
? not.as.good::mosek_equivalent(x)
}
```
There is no way for package `CVXR` to ascertain which version of
`Rmosek::mosek()` is available at runtime, the stub, or the actual one
from "www.mosek.com".
This would cause `CVXR` to fail CRAN checks during submission since
"Enhances" packages get loaded during checks. But we took care to avoid
tests that would exercise "special situations", so it is on CRAN.
But some examples in our reverse dependencies trigger those "special
situations", causing problems with their CRAN submission.
Does anyone have any suggestions on how to handle this?
PS. `Rcplex` is different, as it gets linked to commercial libraries
during **install**, unlike `Rmosek`.
PPS. I recall that the author Rmosek requested that this package be
removed from CRAN via personal email, but never followed up my
suggestion to directly contact CRAN. Unfortunately, I can't seem to find
that email.
[R-pkg-devel] "Enhances" issue with stub package Rmosek
1 message · Balasubramanian Narasimhan