Travis-CI Build Errors unsure why

[trinker]

I get no such buid errors locally: https://travis-ci.org/trinker/pacman

  Running ‘testthat.R’
 ERROR
Running the tests in ‘tests/testthat.R’ failed.
Last 13 lines of output:
  24: getNamespace(ns)
  25: tryCatch(loadNamespace(name), error = function(e) stop(e))
  26: tryCatchList(expr, classes, parentenv, handlers)
  27: tryCatchOne(expr, names, parentenv, handlers[[1L]])
  28: value[[3L]](cond)

  testthat results ================================================================
  OK: 56 SKIPPED: 0 FAILED: 2
  1. Error: p_install works 
  2. Error: p_install_version returns FALSE package doesn't exist on CRAN and min. not met 

  Error: testthat unit tests failed
  Execution halted
* checking for unstated dependencies in vignettes ... OK
* checking package vignettes in ‘inst/doc’ ... WARNING
Package vignettes without corresponding PDF/HTML:
   ‘Introduction_to_pacman.Rmd’
   ‘pacman_functions_quick_reference.Rmd’
* checking running R code from vignettes ...
   ‘Introduction_to_pacman.Rmd’ ... OK
   ‘pacman_functions_quick_reference.Rmd’ ... OK
 OK
* checking re-building of vignette outputs ... OK
* checking PDF version of manual ... OK
* DONE
Status: 1 ERROR, 1 WARNING

[Dasonk]

My guess is that we are expecting warnings in our tests but instead of using expect_warning we just try to suppress them but aren't doing it properly. and travis is using the warning -> error setting.

I can't modify anything on the repo at the moment but I think the following should fix one of the problems - change test-p_install.R to this:

context("Checking p_install")

test_that("p_install works",{
    require(pacman)
    expect_warning(tmp <- p_install("iDontExistAnywhere"))
    expect_false(tmp)
})

and change the last test_that in test-p_install_version.R to

test_that("p_install_version returns FALSE package doesn't exist on CRAN and min. not met",{

    expect_warning(out2 <- p_install_version("MadeUpPackage", "3.4.1"))
    expect_false(out2[[1]])
})

[trinker]

In R CMD (Build/Check in RStudio) this line:

tmp <- p_install("iDontExistAnywhere")

produces an error not a warning. If I do:

context("Checking p_install")

test_that("p_install works",{
    require(pacman)
    #expect_warning(tmp <- p_install("iDontExistAnywhere"))
    expect_error(tmp <- p_install("iDontExistAnywhere"))
    #expect_false(tmp)
})

All works. That means that the R CMD uses of testthat is somehow different than running devtools::test(). I'm not sure what's going on. Again running devtools::test() results in:

> devtools::test()
Testing pacman
Checking p_author : ..
Checking p_base : ..
Checking p_citation : .
Checking p_data : ..
no data sets found
...
Checking p_exists : ........
Checking p_extract : ....
Checking p_functions : ......
Checking p_install : Warning in install.packages :
  package ‘iDontExistAnywhere’ is not available (for R Under development)
.
Checking p_install_gh : 

which is expected. We pass the checks and devtools makes me feel good stating:

Way to go!

I suspect this is to mock me.

Dason said:

travis is using the warning -> error setting

I think this is pertinent. This is how the dev version of R seems to be handling the warnings as well.

[Dasonk]

I'll be honest - I've never really investigated the travis stuff as much as you have. Is this still an issue or should we close this?