[R-pkg-devel] My test passes in OSx on Travis, but not in OSx on CRAN

Dear all,

I have a test in my R package that passes in OSx on Travis-CI, but it does
not pass in OSx on CRAN. What does pass on both CRAN and Travis are the
Linux tests. Also, on CRAN the Windows flavor passes. So, there seems to be
an issue specific to the combination OSx and CRAN.

The easiest solution for me would be to add 'skip_on_cran()' at the top of
the test, but that would blind me for serious issues (if any). I already
made sure that the seeds are set before random number generators and that
value checks are limited to 3 decimal places to deal machine precision
issues. Does anyone have suggestions for me on what else I can do to
investigate this?

Link
<https://github.com/wadpac/GGIR/blob/master/tests/testthat/test_read.myacc.csv.R>
to the specific test on GitHub. The function that is being tested first
creates dummy csv-files with random data and then checks that my wrapper
function around data.table() is able to read and interpret those dummy
files.
Link <https://cran.r-project.org/web/checks/check_results_GGIR.html> to
CRAN results
Link <https://travis-ci.org/wadpac/GGIR> to Travis-CI page for the package

I have a local Windows and Linux machine, but for OSx testing I depend on
Travis-CI.

Thanks,
Vincent

	[[alternative HTML version deleted]]

The error seems to be because you are trying to write to the user's home
directory. This is not allowed. Probably by accident your test creates
"~/testcsv4.csv" (the same problem is present for other test files).

Best
Tomas

On 9/13/19 1:21 PM, Vincent van Hees wrote:

Dear all,

I have a test in my R package that passes in OSx on Travis-CI, but it

does

not pass in OSx on CRAN. What does pass on both CRAN and Travis are the
Linux tests. Also, on CRAN the Windows flavor passes. So, there seems to

be

an issue specific to the combination OSx and CRAN.

The easiest solution for me would be to add 'skip_on_cran()' at the top

of

the test, but that would blind me for serious issues (if any). I already
made sure that the seeds are set before random number generators and that
value checks are limited to 3 decimal places to deal machine precision
issues. Does anyone have suggestions for me on what else I can do to
investigate this?

Link
<

https://github.com/wadpac/GGIR/blob/master/tests/testthat/test_read.myacc.csv.R

to the specific test on GitHub. The function that is being tested first
creates dummy csv-files with random data and then checks that my wrapper
function around data.table() is able to read and interpret those dummy
files.
Link <https://cran.r-project.org/web/checks/check_results_GGIR.html> to
CRAN results
Link <https://travis-ci.org/wadpac/GGIR> to Travis-CI page for the

package

I have a local Windows and Linux machine, but for OSx testing I depend on
Travis-CI.

Thanks,
Vincent

      [[alternative HTML version deleted]]