525
|
|
|
William Reade |
12 years ago
|
|
|
524
|
|
|
William Reade |
12 years ago
|
|
|
523
|
|
|
William Reade |
12 years ago
|
|
|
522
|
|
|
William Reade |
12 years ago
|
|
|
521
|
|
|
William Reade |
12 years ago
|
|
|
520
|
|
|
William Reade |
12 years ago
|
|
|
519
|
|
|
William Reade |
12 years ago
|
|
|
518
|
|
|
William Reade |
12 years ago
|
|
|
517
|
|
|
William Reade |
12 years ago
|
|
|
516
|
|
|
William Reade |
12 years ago
|
|
|
515
|
|
|
William Reade |
12 years ago
|
|
|
514
|
|
|
William Reade |
12 years ago
|
|
|
513
|
|
|
William Reade |
12 years ago
|
|
|
512
|
|
|
William Reade |
12 years ago
|
|
|
511
|
|
|
William Reade |
12 years ago
|
|
|
510
|
|
|
William Reade |
12 years ago
|
|
|
509
|
|
|
William Reade |
12 years ago
|
|
|
508
|
|
|
William Reade |
12 years ago
|
|
|
507
|
|
|
William Reade |
12 years ago
|
|
|
506
|
|
All Constraints construction in "real" code now goes via a provider
Specifically, the base provider's get_constraint_set method calls through to the existing get_constraint_set function in constraints (so, specific providers don't yet create the ConstraintSets themselves); but the state methods and command line tools which need to use constraints now use only constraints as mediated by the provider.
Constraint.from_strs, and .__init__(None, data), continue to work, due to the sheer weight of tests using them (and the fact that the diff is big enough already); expect a followup tidying this up.
The vast majority of this diff is, however, quite easy to process: because we need a provider to construct a Constraints, and we need a Constraints to construct a service, a very large number of tests came to require an /environment node; I think that now all the random ad-hoc config hackery is gone, and most test cases need no more than a self.push_default_config(), which itself replaces a large amount of very repetitive code scattered across the tests.
R=hazmat CC= https://codereview.appspot.com/5936048
|
William Reade |
12 years ago
|
|
|