Fixed element collapsing in Safari

Since we are using the "beforeload" event on Safari to block content, neiher the "load" nor "error" event is dispatched, and therfore the code that removes blocked elements from the DOM isn't triggered. So I had to dispatch the "error" event manually, in the "beforeload" event handler.

[Wladimir Palant, 2014-01-16 09:21:24]

http://codereview.adblockplus.org/5641434644348928/diff/5629499534213120/safa...
File safari/ext/content.js (right):

http://codereview.adblockplus.org/5641434644348928/diff/5629499534213120/safa...
safari/ext/content.js:330: event.target.dispatchEvent(errorEvent);
a) Does Safari really allow dispatching the same event multiple times? That's
non-standard and would likely be a security issue.

b) We are expecting a load event for frames, not error.

c) Dispatching an event visible to the webpage really doesn't seem to be the
right solution (in addition to being wasteful). How about an ext.onBlocked
callback that will be called with a node that *might* have been blocked? This
can be implemented via load/error event listeners on Chrome and triggered
directly on Safari.

[Sebastian Noack, 2014-01-16 13:20:56]

http://codereview.adblockplus.org/5641434644348928/diff/5629499534213120/safa...
File safari/ext/content.js (right):

http://codereview.adblockplus.org/5641434644348928/diff/5629499534213120/safa...
safari/ext/content.js:330: event.target.dispatchEvent(errorEvent);
On 2014/01/16 09:21:24, Wladimir Palant wrote:
> a) Does Safari really allow dispatching the same event multiple times? That's
> non-standard and would likely be a security issue.
Apparently it does. But if that is non-standard, I agree that we shouldn't do
that. I uploaded a new patchset that fixes that.

> b) We are expecting a load event for frames, not error.
Also fixed.

> c) Dispatching an event visible to the webpage really doesn't seem to be the
> right solution (in addition to being wasteful).
Actually it is important that the event is visible to the webstie. Do you
remember the SiteCatalyst issue
(https://adblockplus.org/blog/links-not-working-on-a-website-you-can-fix-that)?
The solution you suggested there required to handle the "error" event for a
blocked resource. This wouldn't be possible if that event isn't dispatched to
the website.

> How about an ext.onBlocked
> callback that will be called with a node that *might* have been blocked? This
> can be implemented via load/error event listeners on Chrome and triggered
> directly on Safari.
I would prefer to keep the same behavior for Chrome and Safari. And while this
event would be what you would expect from an onBlocked event in Safari, since it
can be triggered by the code that does the actual blocking. On Chrome it would
be rather an aggregation of the "load" and "error" event.

[Wladimir Palant, 2014-01-16 16:14:41]

LGTM