Is calling c.Check in TearDownTest supposed to be supported?

[rillig]

I have a suite of unit tests that check the output of my program. To ensure that there is no unchecked output, I use the following code:

type Suite struct {
  output bytes.Buffer
}
func (s *Suite) Output() string {
  defer s.output.Reset()
  return s.output.String()
}
func (s *Suite) TearDownTest(c *check.C) {
  if out := s.Output(); out != "" {
    c.Logf("Unchecked output; check with: c.Check(s.Output(), check.Equals, %q)", out)
  }
}

Using c.LogF hides the output though, since successful tests don’t log anything. Using c.ErrorF instead of c.LogF reports that the fixture panicked, which is also not what I intended when I wrote that code. I had expected that I can use c.ErrorF in the TearDownTest exactly as I can use it in the actual unit test, having it print the nice “obtained—expected” message.

The current API doesn’t allow me to query whether the test panicked, I can only ask c.Failed().

Is the above code something that check should support?

[rillig]

Related to #41 and #22.

[aronatkins]

Using either check.Equals against an empty string or check.HasLen against zero is fine.

func (s *Suite) TearDownTest(c *check.C) {
	// Either is fine.
	c.Check(s.Output(), check.HasLen, 0)
	c.Check(s.Output(), check.Equals, "")
}