On 3 Jul 2006, at 13:56, Adam Kennedy wrote:
That seems like a problem too but the one I'm trying to get at is
4 no plan, with groups
If your script exits prematurely after one of the groups, the
harness
will not notice because everything looks just fine. The solution to
this is not to use "plan, with groups" because then you have to count
all the tests individually which goes aginst objective #2,
But then we've had this problem up till now anyway.
If it exists prematurely with a "good" return code now, it's a
correct ending, if it returns with a bad return code it's an error.
The addition of groups will not change that behaviour in unplanned
test space, because what you want is a simply unknowable.
If we don't have some way of signifying the end of a group in TAP
then it removes a chunk of the utility for the people writing things
that generate TAP - since everybody has to write their own checks
that groups actually output the number of tests that they should.
If we have an end-of-group marker the TAP::Parser can pick this up -
which seems much more sensible to me.
This, I think, is the same issue as the mixing grouped and non-
grouped tests that I wrote about yesterday. Without an end-of-group
marker a test script sending less/more than the specified number of
tests for the group cannot be detected.
Cheers,
Adrian
