[tapx-dev] What a wonderfully horrible thought.

[Michael G Schwern]

Andy Armstrong wrote:
> On 14 Mar 2007, at 22:17, Michael G Schwern wrote:
>> PS  That's not a great way to tell the user they've run too many  
>> tests and put
>> things out of order.
> 
> It's clear that the long term solution is to push structured  
> diagnostics out through STDOUT and ignore STDERR (for our purposes)  
> altogether.
> 
> In the mean-time I think we can still support stream merging but it  
> has to be a distinct option and it has to carry a gurt big warning  
> about the dangers of pushing random STDERR junk through the parser.

My PS was actually in reference to the fact that runtests didn't print out
anything that told the user what actually went wrong.  It just sort of said
"something failed!"  It even thought the test file passed.

Yes, something not about stream merging. :)