On May 11, 2004, at 2:47 PM, Andy Lester wrote:
I've committed my t/41 and t/42 changes to subversion. They are not the
same as what I submitted to the list.
Yes, that slightly confused me, which is why I asked. I assumed there was a reason why you had chosen to not use skip_all, so I thought I would ask.
Is there any reason to now use skip_all? (Remember, you're wearing the official "Test Expert" hat now so you need to weigh up the issues and make recommendations :)
Personally I would like to stick as close to Test::More as we can, and as far away from home-grown solutions as we can (unless we have too of course).
I see part of the goal of the Phalanx project as not just improving the test suites, but also standardizing them, so that the code modules are all tested in a similar way, using similar standards of what should be tested, etc, etc. So by that line of thought, I think we should use 'skip_all' in these tests.
The reality is that they produce the same output and result, and are virtually identical in that respect (as far as I can tell). Which for me, just points toward using 'skip_all' all the more.
We can't use skip_all, because the t/zv_pp* files have to run their non-zv_pp counterparts as well. I put it back the way it was.
I wanted to use skip_all so that we aren't doing roll-your-own on things
that already are well-defined.
I have actually converted those 2 tests back to 'skip_all' and I think I have resolved the problems you were seeing. From what I can tell from your original patches you sent in the mail, the problem before was in where the call for 'skip_all' was, they were located outside of the BEGIN blocks, which meant that they would run too late. So I have moved them into BEGIN blocks and all seems to run fine now. We just need to be sensitive of load order at the top of the file with things like this. I will commit these changes by end of day today.
One concern just popped into my head... I'd like to not have to depend on very recent versions of Test::More. Can you look into that and make recommendations about what version of Test::More we should use as a minimum?
This brings up a point of how backwards compatible do you (Tim) want this all too be? I assume that DBI needs to support the widest possible audience, but is there any official "must be supported" list that we should know about?
Earliest on CPAN is 0.45, and that should be just fine. I don't see anything in the Changes file that tells me otherwise.
The most "very recent" version of Test::More is from August 2002.
I agree with this, I think the furthest back I would want to go is 0.40 (Dec 2001), since that was when cmp_ok was added, and I use that pretty heavily.
-
Steve
xoa
-- Andy Lester => [EMAIL PROTECTED] => www.petdance.com => AIM:petdance
