| Age | Commit message (Collapse) | Author |
|
|
|
If an app uses HTML5 mode and we open an html5 url on IE8 or 9 which
don't support location href, we use location.replace to reload the page
with the hashbang equivalent of the url but this fails with infinite
digest. This is because location.replace doesn't update location.href
synchronously on IE8 and 9.
Closes #2802, #3305, #1417
|
|
|
|
$window may be mocked out in tests causing those tests to fail. So
don't use $window.
|
|
ngScenario expects an ngApp directive to be used, and doesn't work for
manually bootstrapped apps. The failure mode is to hang on navigation.
Trying to make this wont-fix bug less obscure by documenting it.
Eventually Protractor will replace ngScenario and fix this.
|
|
|
|
|
|
In commit 6820322db562382fac903be35831275948825317 of Karma-Jasmine, the
dependency on angular.dump was removed. This caused two undesirable side
effects in the angular.js project. 1) Tests for presence of mock dump were failing,
and 2) the default window.dump was not outputting valuable angular-aware info. This
simple fix adds window.dump in testabilityPatch, to preprocess dumped input prior
to passing it to the global dump method.
|
|
This code is not being used any more and the test is now failing
due to Karma changes. Karma used to expose window.dump but that
changed recently and that's why our build is now failing.
I'm removing the code and test, but we still need to figure out
how to route window.dump through angular.mock.dump, but that will
have to be a separate commit.
|
|
Previously it was possible to get into a situation where child controller
was being instantiated before parent which resulted in an error.
Closes #2738
|
|
Previously any $evalAsync task scheduled from a isolate scope or a child of an isolate scope
would never execute because we never flushed this queue
|
|
The input [number] error spans did not show on the example, as they were
relying on an non-existing property (myForm.list.$error) vs the working
property (myForm.input.$error)
|
|
|
|
|
|
|
|
|
|
This allows us to use minErr in other modules, such as resource and sanitize.
|
|
|
|
|
|
|
|
Ref: 1adf29af13890d61286840177607edd552a9df97
BREAKING CHANGE: img[src] URLs are now sanitized via a separate
whitelist regex instead of sharing the whitelist regex with a[href].
With this change, img[src] URLs may also be data: URI's matching
mime types image/*. mailto: URLs are disallowed (and do not make
sense for img[src] but were allowed under the a[href] whitelist used
before.)
|
|
|
|
Added directives for focus and blur events.
Closes #1277
|
|
|
|
Fix closing parenthesis, quotes around string literal and remove
trailing whitespace.
Closes #3250
|
|
|
|
|
|
not attached to the DOM
|
|
Google chrome (when tested using karma) spits out a log file called libpeerconnection.log
|
|
|
|
It is now possible to notify a promise through deferred.notify() method.
Notifications are useful to provide a way to send progress information
to promise holders.
|
|
|
|
Add an alternative way to define a mapping for ng:pluralize using
attributes instead of the `when` attribute
Closes #2454
|
|
|
|
|
|
caching the version number speeds up the build and preserves resources.
this also fixed EMFILE error that now occurs on some macs.
|
|
Regular expression objects didn't used to be considered to be equal when using
'angular.equals'. Dirty checking therefore failed to recognize a
property modification.
Closes #2685
|
|
|
|
Closes #2770
|
|
|
|
Return early in `angular.toJson` if the object to be stringified is `undefined`.
IE8 stringifies `undefined` to `'undefined'` whereas other browsers return
`undefined`. This normalizes behavior and passes currently broken unit tests
in IE8.
|
|
|
|
The colon character is used to identify parameters in $resource.
This meant that we had to escape the colon used in a port.
It turns out that this is not necessary if we assume that parameter
names cannot consist of only digits.
If the parameter consists only of numbers, then it's a port.
Closes #2778
|
|
With select(...).option(val) it previously would select the first node
which contains the value, even if an exact match was available.
This fix prefers exact matches if available, otherwise it reverts
to the previous 'contains' behaviour for backwards compatibility.
Closes #2856
|
|
According to RFC 3986 (http://tools.ietf.org/html/rfc3986#section-3.1)
schemes such as http or mailto are case-insensitive. So links such as
http://server/ and HTTP://server/ are valid and equivalent.
Closes #3210
|
|
|
|
|
|
- add tests
- the link points to the gruntUtil.getVersion().number tree on github
|
|
Closes #2915
|
|
|