[R-pkg-devel] Fwd: [CRAN-pretest-archived] CRAN submission huxtable 2.0.2

| Hi guys,
| 
| I've been having some problems with updating my 'huxtable' package on CRAN.
| The latest issue is that I run a test which renders a rmarkdown document to
| PDF, using rmarkdown::render. This passes R CMD check fine on my machine,
| on travis and on win-builder, but fails on the CRAN machines as follows:
| 
| Last 13 lines of output:
|   BEGIN failed--compilation aborted at
| D:\Compiler\texmf\scripts\latexmk\perl\latexmk.pl line 122.
|   File::Path version 2.08 required--this is only version 1.08 at
| D:\Compiler\texmf\scripts\latexmk\perl\latexmk.pl line 122.
| 
| This looks like something coming via pandoc, and it also looks as if it is
| a function of a very old version of Perl: File::Path version 2.08 came out
| in 2009. Version 1.08 is included with Perl 5.8 which came out in 2002.
| 
| What's going on here, and can anyone suggest a fix?

Maybe don't run the test if it always fails, and you cannot control the setting?

Rcpp has had conditional tests for (I guess at least) 5+ years -- I just test
if the version number has four components (ie 0.1.2.3, a test version) as
opposed to three (ie 0.1.3, a release) and test only on the former (my box,
Travis, ... but not CRAN). That started because CRAN had its time limit, but
the idea is generally valid.

| In general, is there any info available about the CRAN check servers? I
| keep running into problems that are specific to them and cannot be
| replicated on any other machine available to me.

You can ask, and eventually someone will answer, but "those in charge" are
busy too.

Dirk

http://dirk.eddelbuettel.com | @eddelbuettel | edd at debian.org

Hi Dirk

Not running the test on CRAN would fix the problem, but it is kind of an
admission of failure. Part of CRAN's point is quality control, so switching
off tests just to pass seems perverse. I'll do it if that is the only
option.

Yes, I bet the CRAN guys are highly busy, and full respect to them. It
might be nice, though, if there were a public document about what is on
those servers. Just e.g. a list of software and versions. Or the name of
the OS and release. That would presumably help
travis/win-builder/individual R developers to create environments closer to
the CRAN one, and that would reduce the number of CRAN-only bugs.

Cheers,
David

Maybe don't run the test if it always fails, and you cannot control the
setting?

Rcpp has had conditional tests for (I guess at least) 5+ years -- I just
test
if the version number has four components (ie 0.1.2.3, a test version) as
opposed to three (ie 0.1.3, a release) and test only on the former (my box,
Travis, ... but not CRAN). That started because CRAN had its time limit,
but
the idea is generally valid.

| In general, is there any info available about the CRAN check servers? I
| keep running into problems that are specific to them and cannot be
| replicated on any other machine available to me.

You can ask, and eventually someone will answer, but "those in charge" are
busy too.

Dirk

--
http://dirk.eddelbuettel.com | @eddelbuettel | edd at debian.org