[Rd] Add a new environment variable switch for the 'large version' check

[Jan Gorecki]

For the same reason, handling false positive in CRAN checks, there are
other places that could be improved.
Like "size of tarball" NOTE.
If one could control this size with an environment variable. Similarly
to the proposal made by Jim. It would be useful as well.

On Thu, Apr 16, 2020 at 5:06 PM Henrik Bengtsson
<henrik.bengtsson using gmail.com> wrote:
>
> I'd second Jim's feature request - it would be useful to be able to
> disable this in CI and elsewhere.    The concept of using an "unusual"
> version component such as a very large number does a nice job of
> indicating "unusual" and serves as a blocker for submitting
> work-in-progress to CRAN by mistake (hence the validation in 'R CMD
> check').
>
> Another point, which I don't think Jim made, is that this would make
> it possible to run R CMD check --as-cran on your work-in-progress and
> get all OKs.  This in turn would allow us to trigger a non-zero exit
> status also for NOTEs (not just ERRORs and WARNINGs).  Currently, the
> warning on -9000 is a false positive in this sense.  This will allow
> developers to be more conservative without risking to treat NOTEs as
> something to expect as normal.  CI services are typically configured
> to alert the developer on ERRORs and WARNINGs but, AFAIK, not on
> NOTEs.
>
> On the topic of unusual version numbers: I'd like to suggest that
> CRAN(*) makes an unusual version bump whenever they orphan a package,
> e.g. to suffix -00001. CRAN already updates/modifies the package
> tarball for orphaned packages by setting 'Maintainer: ORPHANED' in the
> DESCRIPTION file. By also bumping the version of orphaned packages it
> would it stand out in sessionInfo(), which helps in troubleshooting
> and bug reports, etc.  But more importantly, the most recent stable
> CRAN release remain untouched, which I think has a value by itself for
> scientific purposes.
>
> /Henrik
>
> (*) Yes, I should email CRAN about this, but I think it's worth
> vetting it here first.
>
> On Thu, Apr 16, 2020 at 7:44 AM Dirk Eddelbuettel <edd using debian.org> wrote:
> >
> >
> > Or you use a fourth component to signal a development version as Rcpp has
> > done for years (and, IIRC, for longer than devtools et al used '9000').
> >
> > There is no functional difference between 1.2.3.1 and 1.2.3.9000. They are
> > both larger than 1.2.3 (in the package_version() sense) and signal an
> > intermediate version between 1.2.3 and 1.2.4.
> >
> > But one requires a patch. ¯\_(ツ)_/¯.
> >
> > Dirk
> >
> > --
> > http://dirk.eddelbuettel.com | @eddelbuettel | edd using debian.org
> >
> > ______________________________________________
> > R-devel using r-project.org mailing list
> > https://stat.ethz.ch/mailman/listinfo/r-devel
>
> ______________________________________________
> R-devel using r-project.org mailing list
> https://stat.ethz.ch/mailman/listinfo/r-devel