Skip to content

HTTPS clone URL

Subversion checkout URL

You can clone with HTTPS or Subversion.

Download ZIP

Loading…

Testing #119

Open
piccolbo opened this Issue · 1 comment

1 participant

@piccolbo
Owner

Testing is becoming a problem

  1. too slow: we need more and bigger tests but it slows down development
  2. doesn't provide enough info on failure.
  3. not automated: need to bring up cluster, start, check, on different platforms
  4. not isolated. I suspect that at least on my laptop I may have interfered with a test by launching it and then checking out a different branch to do some work.

ideas welcome

@piccolbo
Owner

As far as point 1, as the errors happen later and later in the test set, it get more more costly to retry the testing. I normally execute the failing test by with R --vanilla < test.R but when it passes I call R CMD check rmr for convenience and it starts from scratch, whereas if the changes are not far reaching I would rather continue with the next test. As far as point 2 above, one problem is that I haven't switched to my unit testing system consistently. On failure unit.test gives a lot of information. The other problem is that most test are wrapped in a loop for(be in c("local", "hadoop") which means that R CMD check always prints the last lines of the loop on failure, not the test that failed.

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Something went wrong with that request. Please try again.