Issue 4211 - Improve detection of uninitialized documents for means of element hiding

Issue 4211 - Improve detection of uninitialized documents for means of element hiding

Repository: hg.adblockplus.org/adblockplus

[Thomas Greiner, 2016-07-18 15:58:14]

LGTM

While I haven't found any documentation on the "uninitialized" state, I found it
being used throughout the Firefox code so that should be fine for this purpose.

https://codereview.adblockplus.org/29347936/diff/29347937/lib/child/elemHide.js
File lib/child/elemHide.js (right):

https://codereview.adblockplus.org/29347936/diff/29347937/lib/child/elemHide....
lib/child/elemHide.js:331: // it doesn't fire when changing from "uninitialized"
state.
While the "unitialized" state is not mentioned in
https://www.w3.org/TR/html5/dom.html#documentreadystate (the listed states are
considered non-normative so no issue), it clearly states that "When the value is
set, the user agent must fire a simple event named readystatechange at the
Document object." (see
https://www.w3.org/TR/html5/dom.html#current-document-readiness) so seems to be
a browser issue.

[Wladimir Palant, 2016-07-18 18:07:18]

https://codereview.adblockplus.org/29347936/diff/29347937/lib/child/elemHide.js
File lib/child/elemHide.js (right):

https://codereview.adblockplus.org/29347936/diff/29347937/lib/child/elemHide....
lib/child/elemHide.js:331: // it doesn't fire when changing from "uninitialized"
state.
On 2016/07/18 15:58:14, Thomas Greiner wrote:
> While the "unitialized" state is not mentioned in
> https://www.w3.org/TR/html5/dom.html#documentreadystate (the listed states are
> considered non-normative so no issue), it clearly states that "When the value
is
> set, the user agent must fire a simple event named readystatechange at the
> Document object." (see
> https://www.w3.org/TR/html5/dom.html#current-document-readiness) so seems to
be
> a browser issue.

Given that "uninitialized" means "too early to be seen by any content JavaScript
code" - it's not surprising that Gecko devs cut this corner and don't fire an
event for the state change here. You can only observe this state from an
extension, and even here it might be considered a bug. Arguably,
content-document-global-created notification shouldn't fire before the document
is initialized.