[MERGE] Improve tests for the behaviour of Tree.iter_changes for missing paths that are only present in one tree, and fix found bugs. (Robert Collins)

Vincent Ladeuil v.ladeuil+lp at free.fr
Wed Aug 13 19:08:21 BST 2008


>>>>> "aaron" == Aaron Bentley <aaron at aaronbentley.com> writes:

    aaron> Robert Collins wrote:
    >> (Or to put it another way, I don't like TestNotApplicable all that
    >> much, because 'raise' and 'not a problem' just don't fit in my head).

    aaron> So then, consider it a problem that a test was run
    aaron> which it did not make sense to run.

Thanks for that Aaron.

I wrote a mail earlier today and threw it away because I was
unhappy with the result.

I agree 90% here.

The missing 10% are because I think that there should be a way to
*not* run these tests *at all*. Since they are parametrized tests
they should just not be generated[2]. 

And I'm sure there is a high value in coding what Robert
expressed with both "not tree1.path2id('file')" and
"InterDirstateTree falls back to InterTree if the basis is not a
DirstateRevisionTree, and revision trees cannot have missing
files"[0].

        Vincent

Shameless pseudo-review[0]: Missing paren for the "e.g." here
Robert :-D

Since I'm in review mode now, BB:tweak, I'm ok with the test as
is *iff* a FIXME saying 'there should a better way to avoid
running non-sensical tests than using a bare return while
TestNotApplicable "just don't fit in my head" -- <nick><date>'[1]
is added.

[1] I know the feeling, I'm sure I can find tests where I used
TestNotApplicable even if it didn't fit in my head either.

[2] I'm not asking to do the needed refactoring in our
    parametrization framework, just linking subjects so that this
    case is not forgotten.



More information about the bazaar mailing list