Arvados

Reviewing e102efba. The fix is good, but I'm surprised at how involved the test code is.

• I'm not sure I understand what value we get out of testing ten cases of this. It seems like we get the most value just by testing a simple case, a case that spans the AM/PM boundary, and a case that spans the midnight boundary. If there are other cases that have some unique property, then let's definitely include those. But ten seems like an arbitrary number, and that's a lot of overhead given that we're firing up a whole browser session to test datetime formatting.
  • I now limited to two tests, one with 0 run time and one with run time that spans between AM and PM

• You can use DateTime::strptime to parse these strings, rather than doing all the splitting and array indexing yourself.
  • Thanks for this tip. It is fantastic

• Rather than trying to parse the "elapsed time" string, let's use fixtures whose elapsed time we know ahead of time, and compare against that. That will save us a bunch of parsing and math that aren't directly relevant to what we're trying to test (the datetime formatting).
  • Adjusted this

BTW, there's also trailing whitespace on some of the lines in the current branch.

Thanks.

Thanks for following through on this. 99ad159 is good to merge, thanks.