0.2.14 Caused a breaking change

[OverZealous]

I (accidentally) just upgraded to the latest release, and suddenly my tests won't pass with this error:

 Error: Default value (false) for parameter 'error' is not an instance of Type (any)
          at $$getDefaultValue (./vendor/angular-ui-router/release/angular-ui-router.js:1628)
          at $value (./vendor/angular-ui-router/release/angular-ui-router.js:1643)
          at ./vendor/angular-ui-router/release/angular-ui-router.js:1686
          at forEach (./vendor/angular/angular.js:323)
          at ./vendor/angular-ui-router/release/angular-ui-router.js:1687
          at transitionTo (./vendor/angular-ui-router/release/angular-ui-router.js:3148)
          at go (./vendor/angular-ui-router/release/angular-ui-router.js:3069)
          at ./build/connect/js/app/routes/error/error.route.js:5:0 <- null:0:0
          at check (./vendor/angular-ui-router/release/angular-ui-router.js:2008)
          at update (./vendor/angular-ui-router/release/angular-ui-router.js:2020)
          at ./vendor/angular/angular.js:14712
          at ./src/connect/app/routes/routes.spec.js:30

I don't mind fixing the issue (it worked without error on 0.2.13), but this is a breaking change on a what should be a patch release.

The faulty code in question looks like this:

$stateProvider.state('error', {
    params: {
        error: false,
        fromState: {},
        fromParams: {},
        toState: {},
        toParams: {},
    },
   //...

---

Actually, on reading through the docs again, this seems like it should be an acceptable format, so this might actually be a legitimate bug.

[nateabele]

Yeah, seems like it might be a bug. I'll investigate shortly.

[eddiemonge]

.config ( $stateProvider ) ->
  $stateProvider

  .state 'packages',
    abstract: true
    parent: 'layout'
    template: '<ui-view />'
    url: '/packages/'

  .state 'packages.received',
    url: 'received?page&per_page&archived&q'
    reloadOnSearch: false
    params: 
      page:
        squash: true
        value: 1
      per_page:
        squash: true
        value: 10
      archived:
        squash: true
        value: false
      q: squash: true
    templateUrl: '/modules/packages/index.html'
    controller: 'PackagesIndexController'

sidenote squash never seemed to squash the query string part even when it matched. im guessing its because of type as well

[compact]

I got the same error after updating from 0.2.13 to 0.2.14. My state parameter is also a boolean that defaults to false, if that's relevant.

[philBrown]

I'd say the boolean param property is definitely causing the error.

Removing any boolean param properties in state config makes the error go away.

[lazar-vuckovic]

Same error here. Only shows up when setting the default parameter to "false".

[OtaK]

👍 Needs to be fixed ASAP. It took down our production momentarily.

[zacaj]

seems to be

if (defaultValue !== null && defaultValue !== undefined && !self.type.is(defaultValue))

(da63b27#diff-6de0577dce2b7e2e060fb6f1b36a33e3R1627)
, when self.type is 'any', self.type.is(x) simply returns x, and when x is false, or "" in my case, !self.type.is() is throwing the error

it seems strange that the 'any' type's is() maps to angular.identity(). Shouldn't this just return true or "typeof a !== 'undefined"?

[spencerwi]

This affected our production as well. It happened when the default parameters were false or 0, leading me to believe that this is a bad falsiness check.

[elado]

+1

Default value false for a param (params: { blah: false }) throws

Error: Default value (false) for parameter 'blah' is not an instance of Type (any)

[spencerwi]

I'm not one to be the semver police, but introducing breakages for something like this in a _._.x release strikes me as a dangerous move -- our production instance got hit by this because anyone who (like us) ran bower install ui-router --save when 0.2.13 was the newest version gets from bower a dependency on ~0.2.13 -- which will grab 0.2.14. Meaning that without any code changes, two consecutive builds may go from "working" to "broken".

The simple solution is to manually ensure that apps depend on exactly 0.2.13, but it takes debugging down (or googling and finding this GH issue) to figure out that's the fix.

[nateabele]

I'm not one to be the semver police, but introducing breakages for something like this [...]

@spencerwi Yup, trust me, didn't do it on purpose. 😉