Due to a known V8 memory leak[1] we need to perform extra cleanup to make it easier
for GC to collect this scope object.
V8 leaks are due to strong references from optimized code (fixed in M34) and inline
caches (fix in works). Inline caches are caches that the virtual machine builds on the
fly to speed up property access for javascript objects. These caches contain strong
references to objects so under certain conditions this can create a leak.
The reason why these leaks are extra bad for Scope instances is that scopes hold on
to ton of stuff, so when a single scope leaks, it makes a ton of other stuff leak.
This change removes references to objects that might be holding other big
objects. This means that even if the destroyed scope leaks, the child scopes
should not leak because we are not explicitly holding onto them.
Additionally in theory we should also help make the current scope eligible for GC
by changing properties of the current Scope object.
I was able to manually verify that this fixes the problem for the following
example app: http://plnkr.co/edit/FrSw6SCEVODk02Ljo8se
Given the nature of the problem I'm not 100% sure that this will work around
the V8 problem in scenarios common for Angular apps, but I guess it's better
than nothing.
This is a second attempt to enhance the cleanup, the first one failed and was
reverted because it was too aggressive and caused problems for existing apps.
See: #6897
[1] V8 bug: https://code.google.com/p/v8/issues/detail?id=2073Closes#6794Closes#6856Closes#6968
ddb8081982 refactors jqLite, and removes support
for IE8. This patch extends this to prevent the CL from breaking IE8, and to
remain suitable for the 1.2.x branch.
Closes#6963
Previously, the jqLite constructor was limited and would be unable to circumvent many of the HTML5
spec's "allowed content" policies for various nodes. This led to complicated and gross hacks around
this in the HTML compiler.
This change refactors these hacks by simplifying them, and placing them in jqLite rather than in
$compile, in order to better support these things, and simplify code.
While the new jqLite constructor is still not even close to as robust as jQuery, it should be more
than suitable enough for the needs of the framework, while adding minimal code.
Closes#6941Closes#6958
Previously, constant numbers with a unary minus sign were not treated as constants. This fix corrects
this behaviour, and may provide a small performance boost for certain applications, due to constant
watches being automatically unregistered after their first listener call.
Closes#6932
Due to a known V8 memory leak[1] we need to perform extra cleanup to make it easier
for GC to collect this scope object.
The theory is that the V8 leaks are due to inline caches which are caches
built on the fly to speed up property access for javascript objects.
By cleaning the scope object and removing all properties, we clean up ICs
as well and so no leaks occur.
I was able to manually verify that this fixes the problem for the following
example app: http://plnkr.co/edit/FrSw6SCEVODk02Ljo8se?p=preview
Given the nature of the problem I'm not 100% sure that this will work around
the V8 problem in scenarios common for Angular apps, but I guess it's better
than nothing.
[1] V8 bug: https://code.google.com/p/v8/issues/detail?id=2073Closes#6794Closes#6856
Makes xhr status text accessible is $http success/error callback.
See www.w3.org/TR/XMLHttpRequest/#dom-xmlhttprequest-statustext
Closes#2335Closes#2665Closes#6713
If a JS animation is run before a CSS animation then the JS animation may end up writing style
data to the element. If any transition or animation style data is written then it may end up
being accidentally inherited into the CSS animation hanlder that ngAnimate uses. This may result
in an unexpected outcome due to the tweaks and hacks that the CSS handler places on the element.
If the CSS animation is run before the JS animation then, if there are no transitions on the style
attribute nor within the global CSS on the page then nothing will happen and the JS animation can
work as expected.
Closes#6675
These were apparently entirely undocumented. I'm not sure if they're intended
to be private, but in case they're not, I've written some initial docs for them
Previously, we had problems with code that contained symbols that looked
like jsdoc directives. This has now been fixed so we can convert these
HTML character codes back to @ signs.
Closes#6822Closes#6826
Running html5-validation immediately after model-value is updated is incorrect, because the view
has not updated, and HTML5 constraint validation has not adjusted.
Closes#6796Closes#6806
The $document docs are pretty empty, and this fills them out a bit. The example itself may not be
particularly useful, but it could be improved or removed later. Works for me.
Closes#6757
The Android 2.x browser is not ES5-compatible in that it does not allow
use of reserved words as property names. This docs fix adds Android to the
note to the `$q` docs which already make it known that string property
notation should be used when using the `finally` method on `$q`.
$watchCollection checks if oldValue !== newValue which does not work for NaN. This was causing
infinite digest errors, since comparing NaN to NaN in $watchCollection would always return false,
indicating that a change was occuring on each loop.
This fix adds a simple check to see if the current value and previous value are both NaN, and
if so, does not count it as a change.
Closes#4605
The orderBy filter now allows string predicates passed to the orderBy filter to make use property
name predicates containing non-ident strings, such as spaces or percent signs, or non-latin
characters.
This behaviour requires the predicate string to be double-quoted.
In markup, this might look like so:
```html
<div ng-repeat="item in items | orderBy:'\"Tip %\"'">
...
</div>
```
Or in JS:
```js
var sorted = $filter('orderBy')(array, ['"Tip %"', '-"Subtotal $"'], false);
```
Closes#6143Closes#6144
Previously, non-string values stored in $cookies would be removed, without warning the user, and
causing difficulty debugging. Now, the value is converted to string before being stored, and the
value is not dropped. Serialization may be customized using the toString() method of an object's
prototype.
Closes#6151Closes#6220
Fix click busting of input click triggered by a label click quickly
following a touch event on a different element, in desktop
and mobile WebKit
To reproduce the issue fixed by this commit set up a page with
- an element with ng-click
- a radio button (with hg-model) and associated label
In a quick sequence tap on the element and then on the label.
The radio button will not be checked, unless PREVENT_DURATION has passed
Closes#6302
This is hard to test as a unit-test, since it involves the actual loading
of angular, but it turns out that it is easy to test using a protractor
e2e test.
Closes#5863Closes#5587
When the example for `ngAnimate` was added in commit:3344396, the `@param name` annotation was unintentionally duplicated. Remove this duplicate.
Closes#6720
If dealing with a document fragment node with a host element, and no parent, use the host
element as the parent. This enables directives within a Shadow DOM or polyfilled Shadow DOM
to lookup parent controllers.
Closes#6637
In Firefox, hovering over an option in an open select menu updates the selected property of option
elements. This means that when a render is triggered by the digest cycle, and the list of options
is being rendered, the selected properties are reset to the values from the model and the option
hovered over changes. This fix changes the code to only use DOM elements' selected properties in a
comparison when a change event has been fired. Otherwise, the internal new and existing option
arrays are used.
Closes#2448Closes#5994Closes#6769
Originally we destroyed the oldValue by incrementaly copying over portions of the newValue
into the oldValue during dirty-checking, this resulted in oldValue to be equal to newValue
by the time we called the watchCollection listener.
The fix creates a copy of the newValue each time a change is detected and then uses that
copy *the next time* a change is detected.
To make `$watchCollection` behave the same way as `$watch`, during the first iteration
the listener is called with newValue and oldValue being identical.
Since many of the corner-cases are already covered by existing tests, I refactored the
test logging to include oldValue and made the tests more readable.
Closes#2621Closes#5661Closes#5688Closes#6736
PR #5547 introduced conversion of all 0 status codes to 404 for cases
where no response was recieved (previously this was done for the
file:// protocol only). But this mechanism is too eager and
masks legitimate cases where status 0 should be returned. This commits
reverts to the previous mechanism of handling 0 status code for the
file:// protocol (converting 0 to 404) while retaining the returned
status code 0 for all the protocols other than file://
Fixes#6074Fixes#6155
