On 07/09/2018 2:18 AM, David Hugh-Jones wrote:
On Fri, 7 Sep 2018 at 01:16, Duncan Murdoch <murdoch.dun...@gmail.com
<mailto:murdoch.dun...@gmail.com>> wrote:
When packages delete tests just for CRAN, the quality of the repository
suffers. Users should be able to check an install by running the tests
that passed on CRAN and seeing them pass on their system as well.
In my limited experience there are usually tests that can't run on CRAN
because they take too long, rely on external software or configuration
that is absent, or just fail on CRAN only (which is naturally hard to
debug). This seems normal, and having the ability to turn off some tests
is useful. The fact that multiple workarounds have evolved to do this
suggests that the need is widespread.
I think it's useful to think of 3 groups who might run tests:
- authors
- CRAN
- other users of a package.
What Hadley was arguing for is that CRAN should identify itself to a
package, so that by default a package could run different tests for
CRAN than for other users. I am arguing that they should run the same
tests by default.
Clearly authors might want more extensive tests like the ones you mention.
I think it would be a good thing if a standard method evolved for others
to ask for those tests as well. That suggestion was made and support
for it was added to R back in 2014. So far standardization hasn't
happened, which could indicate that it isn't really needed, or just that
nobody has bothered to push for it.
Duncan Murdoch
______________________________________________
R-package-devel@r-project.org mailing list
https://stat.ethz.ch/mailman/listinfo/r-package-devel
