* Ovid <[EMAIL PROTECTED]> [2006-04-19T04:02:31] > From a parser standpoint, there's no clean way of distinguishing that > from what the test functions are going to output. As a result, I > really think that "diag" and normal test failure information should be > marked differently (instead of the /^# / that we see).
I strongly agree. This came up when TAP was being named, and also during a previous Harness upgrade. Unfortunately, Harness and Straps discard comments and, as mentioned before, the STDOUT/STDERR makes it hard to associate diagonstic output with tests. How many things rely on which stream the various output methods in Test::Builder use? I know there was trouble last time that things changed, but wasn't that entirely because of the disconnect between Test::Builder and Test::Builder::Tester? Since they're now together, isn't that particular issue solved? There are other things that test test output, like Test::Tester. Will they break? To find out, I downloaded a pristene copy of the latest Test-Simple and Test-Tester and changed Test::Builder to use STDOUT for failure_output. The only test that failed in the whole set was one that checked whether failure_output defaulted to STDERR. I think that a real investigation into the impact of using one stream by default is in order. -- rjbs
signature.asc
Description: Digital signature
