[R-pkg-devel] Strange timings reported from R CMD check

[Benjamin Hofner]

Dear all,

When we prepared a release candidate form package mboost (see 
https://github.com/hofnerb/mboost/tree/master/pkg/mboostPatch for the 
code) we we found a strange issue regarding the timings. When running

R CMD check --as-cran

on the package mboost we found the following timings:

Linux (x86_64-pc-linux-gnu):
----------------------------

Examples with CPU or elapsed time > 5s
          user system elapsed
mboost 14.791  9.274   11.75


If I run on the Linux machine in R --vanilla the following command:

   system.time({ library("mboost"); example("mboost", ask = FALSE) })

i.e., load the package AND run the example I get a timing of

   user system elapsed
  2.840  1.374   3.051

To check the timings, I also used a Mac.

MacOS (x86_64-apple-darwin10.8.0):
----------------------------------

Examples with CPU or elapsed time > 5s
         user system elapsed
mboost 2.629  2.644     2.9

The MacOS note is very strange as it complains that time > 5s which it 
isn't. The Linux note is also strange as I cannot reproduce these timing 
at all if I run the code separately.

We have now shortened the example to comply with the CRAN policies (in 
any case) but wonder where these very different timings come from.

Is it a bug in R CMD check? Or does anyone know what the reason for 
these strange timings is?

Best,
Benjamin