Properly test clipr_available()

[mdlincoln]

After putting much effort into getting clipr to test properly on Travis (#14), there's still the issue of correctly testing whether clipr_available() actually does the job properly. Since we're using a build matrix on Travis, it's possible to explicitly test this function when we know that clipr ought to work (when DISPLAY=:90.0) and when it ought to fail (DISPLAY="").

This is the code I'm currently trying:

test_that("clipr_available fails when DISPLAY is not configured; 
succeeds when it is", {
  # Only run this test on Travis
  skip_if_not(identical(Sys.getenv("TRAVIS"), "true"))
  if (identical(Sys.getenv("DISPLAY"), ""))
    expect_false(clipr_available())
  if (identical(Sys.getenv("DISPLAY"), ":90.0"))
    expect_true(clipr_available())
})

Unfortunately, this test appears to fail when it ought to pass, clipr_available() returning FALSE when the Travis environment ought to be properly configured. Oddly, it fails even when incorporating the fix from #20

[krivit]

For what it's worth, I just tested the version from #20, and it returns the correct answer within X (TRUE) and on a bare shell (FALSE).

I don't know TravisCI well enough to speculate what's going on. Can the R process see the DISPLAY variables? Is R under Travis built with X capabilities? What does capabilities("X11") say?

[mdlincoln]

Turns out I had the DISPLAY variable set incorrectly - it should have been :99.0 instead of :90.0 🤕 The fix passes on Travis now.