>>>>> "Tom" == Tom Tromey <[EMAIL PROTECTED]> writes:
>>>>> "Akim" == Akim Demaille <[EMAIL PROTECTED]> writes:
Akim> * automake.in (&handle_single_transform_list): Simplify
Akim> computation of $object and $this_obj_ext. * tests/lex3.test:
Akim> Merge into... * tests/lex.test: here. * tests/pr19.test:
Akim> Improve and rename as... * tests/lex3.test: this.
Tom> How about we agree not to remove tests, and only change them if
Tom> they are obviously, blatantly wrong? I try to treat the test
Tom> suite as append-only as much as possible. Changing them is just
Tom> asking for trouble. Also it seems pointless: it doesn't
Tom> substantially help with long term maintenance of automake, nor
Tom> does it make useful progress towards a new release.
It looks like my answer never went out.
My point in changing some parts of the test suite is:
1. make it easier to move to another testing framework
2. group together related tests so that it is easier to see whether
all the possibilities are covered (it is fairly common to have
a pseudo Cartesian product of options to test, and having them
split across several test files does not help improving the test
suite).
The cons is mainly that before knowing that foobar.test fails can be
enough to spot what is actually going wrong. Now if foobar.test holds
several tests it is no longer doable.
But in practice we always need at least a VERBOSE run, so this point,
IMHO, does not hold.
