Mar 06

I have been meaning to add this to my blog for quite some time now. I have read a few postings related to this topic but I am surprised there aren’t a lot more people complaining about this. Personally I feel Adobe really dropped the ball with this bug. I guess I should explain what the bug is before I go on.

Lets pretend we’re creating a button, now I know most of my readers are developers and for the most part we would never create a simple button like this using multiple frames on the timeline. But we all know that designers love to use the timeline and create button states using frames and frame labels. I actually think this posting will help designers as much as it will help the developers who are working with said designers.

Now lets pretend our button has 3 states, up, over, and down and each state is represented on the timeline, 5 frames apart and the frame is labeled the same as each state. So the first frames label is “up”, the 5th frames label is “over”, and the 10th frames label is “down”. On each frame there is a MovieClip named button and each key frame has the same graphic just tinted differently to differentiate the states. Then on our code layer we set up MouseEvents for each state MOUSE_OVER, MOUSE_OUT, and MOUSE_DOWN, in each of the event handlers we move the playhead to each frame using the gotoAndStop method while passing in the frame label that we want the playhead to move to. Ok, so this is all straight forward. Very flash 4. The interesting part of this is when you try to access sibling displayObjects on the same frame right after we call gotoAndStop.

To illustrate this bug lets add a layer below the layer that has our button on it. On this layer add a keyframe at frame 5 and add a MovieClip, whatever you want, import a picture draw a shape, it doesn’t matter. Just as long as you give it an instance name of overClip. Right after we call gotoAndStop(’over’) try and trace (overClip). If you are not completely following me right now here is a picture of what my timeline looks like:

And our code will look something like this (excuse the timeline code, but it is the easiest way to demonstrate this bug).

button.addEventListener(MouseEvent.MOUSE_OVER, rollover);
button.addEventListener(MouseEvent.MOUSE_OUT, rollout);

function rollover(e:MouseEvent):void {
     gotoAndStop('over');
     trace('the instance name of the overclip is: ' + this.overClip.name);
}

function rollout(e:MouseEvent):void {
     gotoAndStop('up');
}

In the rollover function you will notice I added a trace that traces out the name property of the overClip that appears on the out frame (or frame 5 for those of you who like numbers). If you were to run this code you will notice that the rollover works, the playhead moves to frame 5 and visually you see the change, however, the trace statement throws the following error “TypeError: Error #1009: Cannot access a property or method of a null object reference.” Null?? How can that be, the playhead is at frame 5, and the overClip is on frame 5, why is overClip null? Good question! The other odd thing is if you trace out numChildren right after the gotoAndStop(’over’) call it will return the correct number of child displayObjects. How can the player know how many children there is but still throw an error saying we’re trying to access a null object reference?

This was a very common practice in AS2, especially in highly creative sites with both Flash Designers and Flash Developers. And as more creative shops start embracing Flash 9 I think this issue will become more common.

Well, the first thing I thought to do was to call stage.invalidate() before I called gotoAndStop(’over’), in theory this should force a redraw and fire the render event. Well it did fire the render event but I still couldn’t trace out the clip on frame 5. Arghhh!

The one thing I did notice is the error is only thrown the first time rollover is called. Everytime after that the reference to overClip traces out fine. So this got me thinking. It looks like displayObjects do not register until the current frame cycle is complete. For those who do not know, a frame cycle is two parts, first the code executes, once the code on that particular frame executes any display updates are executed (adding overClip to the timeline is a display update).

The first approach to get around this bug I thought up was to kick off a EnterFrame event in the rollever function, after the EnterFrame event is fired, it still didn’t trace out overClip. Then I thought maybe I need a second frame cycle to register everything because technically the EnterFrame event fires at the exact moment the playhead hits the frame which is exactly the same problem we already have. Ok, so lets allow for two EnterFrame events to fire, kill the event, and trace out overClip. Guess what, that works. Here is the code for this approach.

var count = 0;

button.addEventListener(MouseEvent.MOUSE_OVER, rollover);
button.addEventListener(MouseEvent.MOUSE_OUT, rollout);

function rollover(e:MouseEvent):void {
     count = 0;
     gotoAndStop('over');
     addEventListener(Event.ENTER_FRAME, rolloverDelay);
}

function rolloverDelay(e:Event):void {
     count++
     if (count == 2) {
          removeEventListener(Event.ENTER_FRAME, rolloverDelay);
          trace('the instance name of the overclip is: ' + this.overClip.name);
     }
}

function rollout(e:MouseEvent):void {
     gotoAndStop('up');
}

Yes this works, but imagine how complex this mess of code would get if we had a complex 12 state button with animations on each state. Not very practical.

The only way I could come up with to get around this issue was to create an EnterFrame that fires right off the bat, moves the playhead to the last frame of the timeline, and then returns it to the first frame. This ensures that everything that exists on the last frame will be registered and you will be able to access it immediately following the gotoAndStop(’over’) call. The key to this approach is only items that are still in the display list on the last frame will register. If you have a blank keyframe after overClip it will not register if the last frame of the timeline is beyond that. Here is the code for this approach.

button.buttonMode = true;
button.addEventListener(MouseEvent.MOUSE_OVER, rollover);
this.addEventListener(Event.ENTER_FRAME, frameentered);
this.addEventListener(Event.RENDER, rendered);

function frameentered(e:Event) {
     this.removeEventListener(Event.ENTER_FRAME, frameentered);
     if (overClip == null) {
          stage.invalidate();
          this.gotoAndStop(totalFrames);
     }
}

function rendered(e:Event):void {
     if (this.currentFrame == this.totalFrames) {
          this.gotoAndStop(1);
     }
}

function rollover(e:MouseEvent):void {
     trace('the instance name of the overclip is: ' + this.overClip.
}

Excellent, we’re making progress. But we are working with a very simple example, a two state button (up and over). What if we had a more complex button, say 4 states (up, over, down, and disabled), and at each state different assets were added to the timeline like so.

I know, I said complex, and this timeline is far from complex, where are all the tweens, and hundreds of layers, guides, and masks. Well, lets just pretend all that fun stuff is there for now.

If we were to try the last approach with this more complex timeline we still wouldn’t have access to the clips on the over and down frames. This is because we moved the playhead from the first frame to the last frame and back to the first frame again. The playhead never touched anything in between so the clips on the over and down frames would not have been registered yet.

The first approach I came up with to register the clips on all the frames was quite innovative, so I thought. Basically I used the currentLabels array property of MovieClip. For those who don’t know, the MovieClip class now contains a currentLabel and a currentLabels property. The currentLabel returns a FrameLabel object if one exists for the current frame that the playhead is currently on. A FrameLabel Object contains two properties, name, and frame. The name property is the frame label name and the frame property is the frame number (int) of the current frame the playhead is on. The currentLabels property of MovieClip returns an array of FrameLabel Objects containing all the FrameLabels for the targeted MovieClip. Using the currentLabels array I thought I could loop through each FrameLabel in the array and move the playhead to each frame label and once it is complete return the playhead back to the first frame. However, you cannot use a traditional for loop because the frame cycle will not occur until the for loop is complete. Instead I set up a counter variable and use an EnterFrame to move through each FrameLabel Object for targeted MovieClip. This approach is very similar to our previous approach with more stops in between. Lets call this approach the milk run. Here is what my code looked like for this approach.

button.buttonMode = true;
var count:int;
if (count == 0) {
     this.visible = false;
     addEventListener(Event.ENTER_FRAME, proxyframes);
     addEventListener(Event.RENDER, rendered);
}

function proxyframes(e:Event) {
     stage.invalidate();
     gotoAndStop(this.currentLabels[count].name);
     trace(this.currentFrame);
}

function rendered(e:Event):void {
     count++
     trace('render called' + this.currentFrame)

     if (count == this.currentLabels.length) {
          init();
          gotoAndStop(1);
          removeEventListener(Event.ENTER_FRAME, proxyframes);
          removeEventListener(Event.RENDER, rendered);
     }
}

function init() {
     this.visible = true;
     button.addEventListener(MouseEvent.MOUSE_OVER, rollover);
     button.addEventListener(MouseEvent.MOUSE_OUT, rollout);
     button.addEventListener(MouseEvent.MOUSE_DOWN, press);
     button.addEventListener(MouseEvent.MOUSE_UP, release);
}

function rollover(e:MouseEvent):void {
     gotoAndStop('over');
     trace(overClip.name);
}

function rollout(e:MouseEvent):void {
     gotoAndStop('up');
}

function press(e:MouseEvent):void {
     gotoAndStop('down');
     trace(downClip.name);
}

function release(e:MouseEvent):void {
     gotoAndStop('disabled');
     trace(disabledClip.name);

     //disable buttons if there was a parent clip parent.mouseChildren = false would also work.
     button.buttonMode = false;
     button.removeEventListener(MouseEvent.MOUSE_OVER, rollover);
     button.removeEventListener(MouseEvent.MOUSE_OUT, rollout);
     button.removeEventListener(MouseEvent.MOUSE_DOWN, press);
     button.removeEventListener(MouseEvent.MOUSE_UP, release);
}

So why doesn’t this work. Well if you remember what I said earlier, moving the playhead only registers the last frame that it was on. If you run this code you will notice the rollover and rollout methods return null but the release method traces out the correct value, this is because that was the last stop on our playhead milk run.

In order for this approach to work I would have to layout my timeline appropriately. Each state clip should be on its own layer with nothing after it (including blank keyframes) on the timeline, like so:

Now you will notice that the above code works, we are now able to trace out the extra clips that appear on each frame with a label. However, now that everything exists on the last frame we really don’t need to loop through each FrameLabel to register the clips, we can just use our first trick and move the playhead to the last frame, once it renders move back to the first frame. Even though the looping through the FrameLabels code is overkill for this example I thought I would keep it in so people know that capability exists.

So we are able to access sibling displayObjects now but there still is one functional issue. Since the frames are cascading on the timeline when the playhead is on the down frame both downClip and overClip are visible. Likewise, when the playhead is on the disabled FrameLabel upClip, downClip, and disabledClip are visible. That little bug is not going to slip by the creative department. Normally we would get around this issue by adding blank keyframes on the frame following where the extra clip was displayed. If we were to do this in our example the clips wouldn’t register because they do not exist on the last frame and we would be right back where we started. There are two ways around this, one involves more code (the developer approach), the other involves more keyframes (the designer approach). Both are valid.

The timeline approach basically consists of adding keyframes (not blank keyframes) on the frame after the frame where the extra clip appears and setting the alpha of the clip on the second frame to 0. This ensures that the clip is still on the display list in the last frame and allows it to register. With this approach your timeline would now look like so:

Depending on what else is happening in your application this approach may not work. This is when you have to resort to code (yay). Basically with code we do the same thing, however, instead of simply setting the alpha to 0 we’ll set the visibility to false. This only requires three extra lines of code which are included below.

stop();

button.buttonMode = true;
var count:int;
if (count == 0) {
     this.visible = false;
     addEventListener(Event.ENTER_FRAME, proxyframes);
     addEventListener(Event.RENDER, rendered);
}

function proxyframes(e:Event) {
     stage.invalidate();
     gotoAndStop(this.currentLabels[count].name);
     //gotoAndStop(this.totalFrames);
     trace(this.currentFrame);
}

function rendered(e:Event):void {
     count++
     trace('render called' + this.currentFrame)

     if (count == this.currentLabels.length) {
          init();
          gotoAndStop(1);
          removeEventListener(Event.ENTER_FRAME, proxyframes);
          removeEventListener(Event.RENDER, rendered);
     }
}

function init() {
     this.visible = true;
     button.addEventListener(MouseEvent.MOUSE_OVER, rollover);
     button.addEventListener(MouseEvent.MOUSE_OUT, rollout);
     button.addEventListener(MouseEvent.MOUSE_DOWN, press);
     button.addEventListener(MouseEvent.MOUSE_UP, release);
}

function rollover(e:MouseEvent):void {
     gotoAndStop('over');
     trace(overClip.name);
}

function rollout(e:MouseEvent):void {
     gotoAndStop('up');
}

function press(e:MouseEvent):void {
     overClip.visible = false;
     gotoAndStop('down');
     trace(downClip.name);
}

function release(e:MouseEvent):void {
     gotoAndStop('disabled');
     trace(disabledClip.name);
     overClip.visible = false;
     downClip.visible = false;
     button.buttonMode = false;
     button.removeEventListener(MouseEvent.MOUSE_OVER, rollover);
     button.removeEventListener(MouseEvent.MOUSE_OUT, rollout);
     button.removeEventListener(MouseEvent.MOUSE_DOWN, press);
     button.removeEventListener(MouseEvent.MOUSE_UP, release);
}

That’s it. Adobe, if you are reading this, please, please, please fix this bug. It is very annoying. It is obvious that Flash Player 9 was built with Flex in mind (aka no timeline) since that is all that was available when Flash Player 9 was released. I remember reading someone from Adobe admitting that on their personal blog, if I can find the link again I will post it in the comments below. I covered a lot here, but this bug goes even further. Sometimes Event.RENDER doesn’t even fire. In another related bug stop events on the first frame of nested MovieClip do not always work. Supposidly Adobe fixed this in version 9.0.45 of the player but as you can see by the comments on Emmy’s blog posting about the 9.0.45 release it was never really fixed. In fact we ran into this issue at work recently in the latest 9.0.115 version of the player.

I actually met with the Adobe Flash Player team last week and we discussed a lot of these issues, hopefully they will take these into consideration for the next release. To a developer they may seem like minor issues but designers and animators are going to have a difficult time getting around a lot of these bugs. If anyone else has any workarounds please feel free to explain them in the comments below.

P.S. this is my longest post ever, yay me.

Tags: , ,

Oct 28

Yay Leopard arrived on store shelves on Friday!! The upgrade takes forever but in true apple style they have created a very nice update to OSX. However, if you visit a site that is using Flash and the FileReference.upload() in any browser running in Leopard you will not be happy. Not sure if Apple will need to push out an update or if Adobe will. Let’s hope it is Apple, we all know how much fun it is detecting plug-in minor versions and an OS version just to circumvent certain functionality. As it stands the multi-uploader on Flickr isn’t working any more (more here). I found a few others discussing this issue too. And then there is this thread discussing all the bugs found in Leopard, don’t these people have anything better to do with their weekends ;)

Tags: , , , ,

Mar 28

Today at work we ran into another flash player / browser inconsistency. When using MovieClipLoader to load an external asset, the onLoadError event doesn’t fire in Firefox when trying to load an asset from a non-existent domain. In fact, the onLoadStart event doesn’t even fire. The other MovieClipLoader events don’t fire but they shouldn’t, there isn’t any progress and the load never reaches completion or initialization.If you attempt to load a non-existent asset from a valid domain the onLoadError event fires fine in Firefox. In IE and Safari the onLoadError event fires in both scenarios. These are the only browsers I tested but I would be interested in seeing what happens in the others.

So let me break it down. This code doesn’t fire the onLoadError event in Firefox. It just fails silently:

var mLoader:MovieClipLoader = new MovieClipLoader();
var mObj:Object = new Object();
mObj.onLoadError = function(ev:Object):Void {
     getURL('javascript:alert("onLoadError")');
}
mLoader.addListener(mObj);
mLoader.loadClip('http://kaljdflkasjdf.com/image.jpg', this);

However, this code and domain do fire the onLoadError event.

var mLoader:MovieClipLoader = new MovieClipLoader();
var mObj:Object = new Object();
mObj.onLoadError = function(ev:Object):Void {
     getURL('javascript:alert("onLoadError")');
}
mLoader.addListener(mObj);
mLoader.loadClip('http://www.scottgmorgan.com/image.jpg', this);

Thought I would share this knowledge with the world. Last thing I want is others pulling out as much hair as I do :)

Tags: , ,

Feb 14

I recently ran into another issue with ExternalInterface, when sending multiple calls at once? What I am seeing is when multiple calls are made certain calls get dropped unless I use setTimeout and add a 50 millisecond delay? I know this was an issue with the getURL(”javascript: in previous versions but I thought ExternalInterface was suppose to fix the overloading and buffering issues found in some browsers. I have tested in both IE and Firefox to no avail. And what really upsets me is the fact that it works sometimes, it’s not consistent.Is this a known bug, has anyone seen any workarounds (other than my hack of course). It’s happening in both AS2 and AS3. I am using the ExternalInterface to communicate between AS2 and AS3 (AVM1 and AVM2). I know I can use LocalConnection, but that is not the answer I am looking for, I need to use ExternalInterface in this particular project.

P.S. Happy Vallentine’s Day, I got chocolate covered bugs from my secret admirer, the ExternalInterface.

Tags: , ,