[Rd] Re: stop() and Rcmd (PR#910)

Prof Brian Ripley ripley@stats.ox.ac.uk
Thu, 12 Apr 2001 17:47:42 +0100 (BST)

On Thu, 12 Apr 2001, Paul Gilbert wrote:

> >> > "Rcmd check"  on NT does not catch errors generated by stop() in files
> >> > in the tests directory. Thus the fact that all the tests complete does
> >> > not mean they completed succesfully.
> >In my experiments it *does* put out an error but continues.  This is
> >intentional, as some Perls are rather flaky.  If you want to change it,
> >remove the leading `-' in line 29 of etc/Makeconf-tests.
> Knowing little about Windows I won't try to suggest what the default should be for
> Windows. However, I do rely on make stopping when there is an error signalled by "R
> CMD build". There is too much output to examine manually every time I test a package.
> In general I think errors may go unnoticed for a long time if make does not stop on
> errors, so I think the idea of continuing is a bit risky. But perhaps I should be
> using something other than "R CMD build" to do testing. I thought things had recently
> changed so that this was the preferred way to test a package?

1) On Unix only R CMD check exists.

2) On Windows, we did hope to make Rcmd check the standard way.
Unfortunately, a whole series Perl builds were broken, so make
pkgcheck-foo is still preferred.

3) There is a perfectly clear error message in the output on the console
from the run (and there is very litle from correct tests, so little to
look at), and I think it is hard to overlook, if you look at all.
But if you want, you can change it.

> >On the other hand, Rcmd check dse fails for me because the scripts
> >are asking for interactive input (the number of singular values) and
> >chewing up the rest of the script.  That *is* a bug in dse.
> That is a bug in some examples, thanks for pointing it out. Is there a strategy for
> handling interactive input to examples?

You can't have any at all.  It will read the next part of the script,
after the current buffer.

> I don't seem to get the "chewing up the rest of the script" though. I also don't get
> any failure in the sense of an error signalled to R CMD build or to my make. For the
> reasons mentioned above I think that might be a good thing to do.

What happens depends on buffer size and so is platform-specific.  Windows
seems to have much smaller buffers than Solaris or Linux, so this seems to
bite there much more often.

Brian D. Ripley,                  ripley@stats.ox.ac.uk
Professor of Applied Statistics,  http://www.stats.ox.ac.uk/~ripley/
University of Oxford,             Tel:  +44 1865 272861 (self)
1 South Parks Road,                     +44 1865 272860 (secr)
Oxford OX1 3TG, UK                Fax:  +44 1865 272595

r-devel mailing list -- Read http://www.ci.tuwien.ac.at/~hornik/R/R-FAQ.html
Send "info", "help", or "[un]subscribe"
(in the "body", not the subject !)  To: r-devel-request@stat.math.ethz.ch