On Wed, Feb 23, 2005 at 06:42:57PM -0800, chromatic wrote: > The TAP documentation in 2.47_01 says: > > A harness must only read TAP output from standard output and not from > standard error. > > The way Test::Builder works, diagnostics always go to STDERR. Is there > a reason for this beyond "It's tricky to correlate diagnostics to the > appropriate test numbers"? (I agree with that, but I'm willing to take > my chances on certain occasions.)
More to the point is its completely unstructured output that comes not as part of the test but between them. Add to this the comment about not thinking of it in terms of STDOUT and STDERR but two arbitrary, parallel data streams and you rapidly run into nasty syncronization problems. So I'd just leave it as a Test::Harness feature. Similarly, if TH wants to attempt to associate comments (or whatever "lines starting with #" are called this week) with certain tests it can do so but because its unstructured output I would not put it in the protocol.
