On Fri, 30 Aug 2002, Ed Avis wrote: > Isn't it bad style to require a particular order for the tests? I
no. it's bad style to make test writers choose kludgey filenames. imagine that i name a file 0.load.t, and it tests isa_ok() for everything. that's fine. then, i have a problem with import, so i want to write a test for that, and i want to check it after i know everything loads. the best name for this test is "import" because that's what it tests. but i can't call it 0.import.t because it then happens after what i want my first test to be. alternately, i have to choose a not-so-good name to make the alphabetic order work out. > OTOOH, it is quite simple to make a single test which does two or more > things, if you need them to happen in that order. So there is no real > _need_ for MakeMaker or anything else to support requiring a > particular test order. that's not right. Schwern implemented test_harness to support running the tests in alphabetical order because that's what most people expect, historically, and MakeMaker wants things that worked before to still work. that's a good reason for a particular order. on the other hand, MakeMaker use to let the user override that, and i find that feature useful. even though BAIL_OUT isn't implemented, imagine that your first test script checks isa_ok() for everything. if all of those don't work, you don't want to run any other tests. in that case order matters. all sorts of other reasons exist too. i'm asking for flexibility, not policy. you don't have to do things in a way that you don't like, but you shouldn't keep people from doing the things that they would like to do either. i want loose coupling and freedom. -- brian d foy <[EMAIL PROTECTED]>
