Elliott Sprehn esprehn

## Simulate Click Events.html
<!doctype html>
<title>Firefox Click Event</title>

<a href="http://www.yahoo.com/" onclick="return clickRedirect('google');">Test!</a>
<a href="http://www.google.com/" id="google">Google.com</a>

<script>
  /**
   * Redirect a click to another element.
   *

## datoken
#!/usr/bin/ruby

# Tool for getting and managing deviantART auth tokens.
#
# Copyright 2009, Elliott Sprehn. (Zeros-Elipticus.deviantart.com)
#
# Licensed under the GPLv3.
# http://www.gnu.org/licenses/quick-guide-gplv3.html
#

## dAToken.rb
# DeviantART Auth Token
# Zeros-Elipticus
# 4-12-2010

require 'net/https'
require 'uri'

# Get a token, optionally asking dA to generate a brand new one
# if reusetoken is false.
#

## apt-get dist-upgrade
aptitude dist-upgrade
Reading package lists... Done
Building dependency tree
Reading state information... Done
Reading extended state information
Initializing package states... Done
Building tag database... Done
The following packages are BROKEN:
  libapache2-mod-php5 mysql-client-5.1 php5-cgi php5-cli php5-dev php5-gd
  php5-pgsql php5-snmp php5-xsl

## overflow-attachment.patch
diff --git a/Source/WebCore/css/CSSComputedStyleDeclaration.cpp b/Source/WebCore/css/CSSComputedStyleDeclaration.cpp
index c43a48f..ca14687 100644
--- a/Source/WebCore/css/CSSComputedStyleDeclaration.cpp
+++ b/Source/WebCore/css/CSSComputedStyleDeclaration.cpp
@@ -2743,6 +2743,11 @@ PassRefPtr<CSSValue> CSSComputedStyleDeclaration::getPropertyCSSValue(CSSPropert
         case CSSPropertyUnicodeRange:
             break;

+        // FIXME: Implement these for getComputedStyle.
+        case CSSPropertyWebkitOverflowAttachmentX:

## gist:10334499

(function() {
  var requestAutocomplete = HTMLFormElement.prototype.requestAutocomplete;
  HTMLFormElement.prototype.requestAutocomplete = function() {
     var form = this;
     return new Promise(function(resolve, reject) {
         requestAutocomplete.call(form);
         function cleanup() {
            form.removeEventListener(onSuccess);
            form.removeEventListener(onError);

## gist:1f80d17709b34eb66187

<polymer-test name="my-widget-tests">
  <template>
    <my-widget object="{{ object }}"></my-widget>
  </template>
  <script>
    PolymerTest("my-widget-tests", {
      setup: function(data) {
        // Setup data binding for each test.
        data.object = {someValue: 2};

## gist:80be9cf37d4f112c7229
I profiled the starry night benchmark for 4 minutes, below is what I learned. There's a summary at the top, and analysis inline between each section of the trace.

Major take aways from profiling the starry night benchmark:

- Scrollbars are crazy expensive on OS X. We spend a 12.4% of the layout time and 16.5% of the recording time dealing with scrollbars.
- 5.6% of the time is spent starting a base::Timer on OS X to update the preferred size for the green bubble on OS X. Why is base::Timer::Start() so slow?
- updateLayerPositionsAfterLayout strikes again at 14% of the layout costs.
- core layout is extremely fast, it seems we just do "too much stuff" in general.
- Text recording (18% main thread) and rasterization (38% impl thread) are major factors.
- Rastering and clipping rounded rects is still too expensive (21% impl thread).

## fx framework

/*
    Sends and observes notifications by broadcasting notify events.

    Usage:

    <fx-observer for="example" on-notify="{{ handleHashChange }}"></fx-observer>

    FxObserver.notify("example", relatedData);

## README.txt
See the attached test case for an example where Firefox, Chrome/Safari and
old Opera (12) all differ.

1. cd to a directory of files.
2. Run `ruby fouc-server.rb`.

Now the directory will be served on localhost:8000 on your machine,
the directory is also served on localhost:8000/slow and supports a
?delay={seconds} query parameter to cause artificial network delay on that
resource.
	<!doctype html>
	<title>Firefox Click Event</title>

	<a href="http://www.yahoo.com/" onclick="return clickRedirect('google');">Test!</a>
	<a href="http://www.google.com/" id="google">Google.com</a>

	<script>
	/**
	* Redirect a click to another element.
	*
	#!/usr/bin/ruby

	# Tool for getting and managing deviantART auth tokens.
	#
	# Copyright 2009, Elliott Sprehn. (Zeros-Elipticus.deviantart.com)
	#
	# Licensed under the GPLv3.
	# http://www.gnu.org/licenses/quick-guide-gplv3.html
	#
	# DeviantART Auth Token
	# Zeros-Elipticus
	# 4-12-2010

	require 'net/https'
	require 'uri'

	# Get a token, optionally asking dA to generate a brand new one
	# if reusetoken is false.
	#
	aptitude dist-upgrade
	Reading package lists... Done
	Building dependency tree
	Reading state information... Done
	Reading extended state information
	Initializing package states... Done
	Building tag database... Done
	The following packages are BROKEN:
	libapache2-mod-php5 mysql-client-5.1 php5-cgi php5-cli php5-dev php5-gd
	php5-pgsql php5-snmp php5-xsl
	diff --git a/Source/WebCore/css/CSSComputedStyleDeclaration.cpp b/Source/WebCore/css/CSSComputedStyleDeclaration.cpp
	index c43a48f..ca14687 100644
	--- a/Source/WebCore/css/CSSComputedStyleDeclaration.cpp
	+++ b/Source/WebCore/css/CSSComputedStyleDeclaration.cpp
	@@ -2743,6 +2743,11 @@ PassRefPtr<CSSValue> CSSComputedStyleDeclaration::getPropertyCSSValue(CSSPropert
	case CSSPropertyUnicodeRange:
	break;

	+ // FIXME: Implement these for getComputedStyle.
	+ case CSSPropertyWebkitOverflowAttachmentX:

	(function() {
	var requestAutocomplete = HTMLFormElement.prototype.requestAutocomplete;
	HTMLFormElement.prototype.requestAutocomplete = function() {
	var form = this;
	return new Promise(function(resolve, reject) {
	requestAutocomplete.call(form);
	function cleanup() {
	form.removeEventListener(onSuccess);
	form.removeEventListener(onError);

	<polymer-test name="my-widget-tests">
	<template>
	<my-widget object="{{ object }}"></my-widget>
	</template>
	<script>
	PolymerTest("my-widget-tests", {
	setup: function(data) {
	// Setup data binding for each test.
	data.object = {someValue: 2};
	I profiled the starry night benchmark for 4 minutes, below is what I learned. There's a summary at the top, and analysis inline between each section of the trace.

	Major take aways from profiling the starry night benchmark:

	- Scrollbars are crazy expensive on OS X. We spend a 12.4% of the layout time and 16.5% of the recording time dealing with scrollbars.
	- 5.6% of the time is spent starting a base::Timer on OS X to update the preferred size for the green bubble on OS X. Why is base::Timer::Start() so slow?
	- updateLayerPositionsAfterLayout strikes again at 14% of the layout costs.
	- core layout is extremely fast, it seems we just do "too much stuff" in general.
	- Text recording (18% main thread) and rasterization (38% impl thread) are major factors.
	- Rastering and clipping rounded rects is still too expensive (21% impl thread).

	/*
	Sends and observes notifications by broadcasting notify events.

	Usage:

	<fx-observer for="example" on-notify="{{ handleHashChange }}"></fx-observer>

	FxObserver.notify("example", relatedData);
	See the attached test case for an example where Firefox, Chrome/Safari and
	old Opera (12) all differ.

	1. cd to a directory of files.
	2. Run `ruby fouc-server.rb`.

	Now the directory will be served on localhost:8000 on your machine,
	the directory is also served on localhost:8000/slow and supports a
	?delay={seconds} query parameter to cause artificial network delay on that
	resource.