On Apr 23, 2019, at 10:16 AM, I?aki Ucar <iucar at fedoraproject.org> wrote:
On Tue, 23 Apr 2019 at 15:57, David Blodgett via R-package-devel
<r-package-devel at r-project.org> wrote:
I'm working on some CRAN Debian test failures (https://win-builder.r-project.org/incoming_pretest/ncdfgeom_1.0.0_20190423_031452/). The tests all pass on rhub Debian, CRAN windows, local OSX, and Travis Ubuntu. The tests that are failing on CRAN Debian are ones where I check that changes got made to an existing file that is modified in place by my package code.
Are there nuances of the CRAN environment that prevent a `tempfile()` file being created then modified at run time? What is a suggested best practice for such a test?
According to the current master branch, the test that fails is not
using tempfile(), but "nc_file<-'data/test_output.nc'". Just change
that line with a call to tempfile and the test should pass.
--
I?aki ?car