So, if Flash Builder freezes trying to load your workspace what you need to do to fix it is to navigate to .metadata/.plugins/org.eclipse.core.resources and delete the file named .snap. That’s it. No more rearranging views and reimporting projects!

TypeError: Error #1009: Cannot access a property or method of a null object reference.

This task looks like a great exercise for my epic Scala skills I got trying to understand famous Joa‘s code.

Thanks to apparat and Scala I managed to code a small app which rewrites SWF files and moves all constructor code to a private method which is later subscribed to ADDED_TO_STAGE event.

usage: java -jar initInjector.jar [-c] <from.swf> [to.swf]“)
-c — check for references to stage in constructor
If [to.swf] is not specified resulting SWF is saved to <from.swf>

The 0.1 version fails to move exceptions though. Trying to grasp that bytecode exceptions magic atm.

Here’s the application: initInjector.zip

One of the entries in the ABC file is an array of script_info entries (see the next chapter). Each of these entries contains a reference to an initialization method for the script and a set of traits to be defined in the script’s environment. The last entry in that array is the entry point for the ABC file; that is, the last entry’s initialization method contains the first bytecode that’s run when the ABC file is executed.

That’s apparently is not true. I keep getting different classes in the last entry except the main document class which I need. Thanks to @joa for pointing me to SymbolClass tag which contains an array to map Classes to Library Symbols in SWF files. And the very first class there seems to be the Document class I need.

What’s more, the actual game is not in Silverlight, it is in Flash. And it fails with errors on some timer event which don’t go away.

Microsoft “can” make Flash. OLOLOLOLOLOL!!!!

3.3.1 Applications may only use Documented APIs in the manner prescribed by Apple and must not use or call any private APIs. Applications must be originally written in Objective-C, C, C++, or JavaScript as executed by the iPhone OS WebKit engine, and only code written in C, C++, and Objective-C may compile and directly link against the Documented APIs (e.g., Applications that link to Documented APIs through an intermediary translation or compatibility layer or tool are prohibited).

I’m sorta surprised they didn’t wait until Tuesday (after CS5 drops) to drop this bomb after everyone bought flash CS5. But then again, that’s likely why they did this show and tell today, to make sure people didn’t buy flash.

Looks like Apple totally doesn’t want Flash on their devices. In any form.

R.I.P CS5 iPhone publishing.
It was good while it lasted. Hell, it’s not even available yet.

[Embed(source="text.txt",mimeType="application/octet-stream")]
private var TextClass:Class;

I bet this site gets a ton of visits from people like me.

I just don’t know why but I can’t remember this [Embed] construct. I even have several snippets app helpers installed on my mac but all the time it’s faster to google “actionscript embed text file” instead of launching some app.

Same thing is with HTML CDATA syntax. I just can’t remember it. Every time I google “CDATA” just to get this:

<![CDATA[
 ...
 ]]>

OK, I don’t know where to start, so the order here means nothing. Let’s start.

1. LOL! ME BROKED TEH FLASH VERIFEIRZ!!!

for ( 	var i:uint = reversed ? array.length-1 : 0;
    reversed ? i >= 0 : i < array.length;
    reversed ? i-- : i++ ) {
}

This code compiles perfectly but results in a huge bytecode dump and a weird error — VerifyError: Error #1030: Stack depth is unbalanced. 1 != 0.

Don’t ask me how I got to this code.

2. Object vs. Object

var a:Object = {};
a["a"] = 1;
a["b"] = 2;
a["c"] = 3;
for ( var v:String in a ) {
    a["a"] = 2;
    trace(v, a[v]);
}

How do you think what this code will result into? You would guess something like

a 2
b 2
c 3

But it’s actually

a 2
a 2
c 3

for me.

If you change the first line to

var a:Object = new Object();

you’ll get

b 2
a 2
c 3

Of course Adobe didn’t promise order in collections which Objects are. Also, modifying a collection while working with it looks wrong anyway. But how come new Object() and {} differ so much, as far as I remember everyone says that both these constructs create an object, the later one just can handler properties to set. Apparently it’s not right.

Actually {} results into just one opcode NewObject while new Object() results in 2 opcodes: look for class, create class (sorry, will change for actual opcode names later). Anyway, you should be aware of this fact.

3. Where did this Security Error came from?

Imagine that we have a Socket object. We want to connect to some server but it’s down. We got IO Error and want to remove all listeners from Socket events and BOOOOM! we get a Security Error because Flash couldn’t get socket policy file (undoubtedly).

Who knew that right after IO Error we’ll get a Security Error?

4. Who called this .load() ?

Imagine we have a class extending Sound.

public class S extends Sound {
  public function S() {
    super();
  }
  public override function load(stream:URLRequest, context:SoundLoaderContext=null):void {
    throw new Error( 'ZOMG PEW PEW!!!1' );
  }
}

How do you think, what will we get after executing var s:S = new S(); ?

Right, we’ll get ZOMG PEW PEW!!!1 because despite of Flash help saying “If you pass a valid URLRequest object to the Sound constructor, the constructor automatically calls the load() function for the Sound object.”, constructor actually calls .load() anyway.

5. Super()? What super()?

Continuing #4, what if we could block parent constructor of being executed… But we can’t, if you don’t specify super() call compiler will add it automatically. Well, we could add it somewhere where it won’t be executed.

if ( false ) {
  super();
}

But damn, compiler is smart. But not so smart to overcome this:

if ( !true ) {
  super();
}

Or it knows something we don’t! Apparently !true isn’t always false. But we can skip executing super constructor until Adobe fixes it.

6. GET MY CHILD BACK YOU BIATCH!!!

Imagine that we have a Container extends DisplayObjectContainer which overrides removeChild() method to throw exception so bad people couldn’t remove children from it.

Let child:DisplayObject be a child of container:Container, now look at the code

var s:Sprite = new Sprite();
s.addChild( child );
s.removeChild( child );

Apparently, it bypasses our overridden method and STEALS OUR CHILD!!!!1

7. Wait, so many loaders… I’m confused…

All the time we use loaders of different kinds: Loader, Sound, URLLoader, URLStream, FileReference. So why are they so different?

1. Loader.

• First of all it extends DisplayObject, maybe that’s why everything loading related is in contentLoaderInfo (this is what makes me mad every time).
• load(request:URLRequest, context:LoaderContext = null):void
• bytesLoaded:uint (read-only)
• bytesTotal:uint (read-only)
• url:String (read-only)

2. Sound.

• load(stream:URLRequest, context:SoundLoaderContext = null):void
• bytesLoaded:uint (read-only)
• bytesTotal:int (read-only)
• url:String (read-only)
• Doesn’t have httpStatus event

3. URLLoader.

• load(request:URLRequest):void
• bytesLoaded:uint (read-write)
• bytesTotal:uint (read-write)
• Also, why all URLLoader’s properties are public variables not accessors?

4. URLStream.

• load(request:URLRequest):void
• No bytesLoaded or bytesTotal, but at least we get a progress event

5. FileReference.

• We get 3 methods to load data: load, download, upload
• No bytesLoaded or bytesTotal, but at least we get a progress event

And last, ProgressEvent has properties bytesLoaded (read-write) and bytesTotal (read-write). Both of type NUMBER.

So if you want to implement your own universal loader you must know all these differences.

8. Missing TextField.

Flash says that inheritance chain for TextField is the following: TextField -> InteraftiveObject -> DisplayObject -> EventDispatcher -> Object. Apparently there’s DisplayObjectContainer missing because we can do the following:

public function test() {
  var t:TextField = new TextField();
  t.addEventListener( Event.ADDED, this.handler_addedToStage )
  t.width = 400;
  t.height = 200;
  t.htmlText = ' <img src="http://www.google.ru/logos/olympics10-hockey-hp.png" />';
  super.addChild( t );
}
private function handler_addedToStage(event:Event):void {
  if ( event.target == event.currentTarget ) return;
  trace( event.target );
}

The code results in

[object Loader]
[object Sprite]

It means that we can grab Loaders for HTML images and preload them. Also we can reparent them using the trick above (8

9. Hi, I am addedToStage! Hi, I am addedToStage!

Here’s the code:

public function test() {
  var s1:Sprite = new Sprite();
  s1.addEventListener( Event.ADDED_TO_STAGE, handler );
  super.addChild( s1 );
}

private function handler(event:Event):void {
  var s2:Sprite = new Sprite();
  s2.addEventListener( Event.ADDED_TO_STAGE, trace );
  ( event.target as Sprite ).addChild( s2 );
}

Do you expect to see only one trace? Nope, you’ll get TWO!

And here’s why. We know that addedToStage event is dispatched in the addChild() method. After executing it must tell all children that they were added to stage too. So, when we add a new object to display list in handler() it first gets its own event and after that execution flow returns back to super.addChild() where it tells all its children about them being added to stage. And there’s where the second event comes from.

Of course if we add more children objects we’ll get more misleading events.

10. __AS3__.vec

The Vector class is kind of mysterious. Flash help says that it is top level class but actually it relies in virtual package __AS3__.vec which doesn’t actually exists but is made by compiler.

Looks like a patch which is there since some early beta of Flash Player 10.

HOORAY!

That’s all for now. If you have something to add to the list feel free to mail me or leave a comment here.