It timed out after 25 minutes :(.
We've gotten a few reports of problems with String#shellescape: * https://twitter.com/jcoglan/status/278464520245092353 * #728 (comment) However, I can't repro errors. Our travis build is green on 1.8.7. I realized that String#shellescape is only available if `shellwords` hasn't been loaded, but Rake loads that library: jimweirich/rake@d000a3d ...so I'm not sure how the reported errors are occurring. Regardless, the changes here are an improvement, I think: * Explicitly require `shellwords` -- it's bad to rely on Rake requiring it as there is no guarantee it will always do so. * Use feature detection (e.g. `"".respond_to?(:shellwords)`) rather than version detection. This is helpful if a user runs a 1.8.6 w/ shellescape backported (e.g. using the backports gem) and also if there are any early patch levels of 1.8.7 for which shellwords does not define String#shellescape (not sure if that is the case, but it's one potential source of the reported breakage).
Configuration options are loaded from files, this docstring clarifies which order they are loaded in and where they are loaded from (hopefully!) Closes #739
…scription. Fixes #733.
This commit fixes a problem that I introduced in pull request #569, where RSpec won't run any specs if someone explicitly called exit() or if they were using both Test::Unit and RSpec together and they had loaded RSpec before loading Test::Unit. In particular, exit() called in Test::Unit's at_exit handler would raise SystemExit, thereby defining $!, whose presence inhibits RSpec's at_exit handler from running specs.