25 Aug 13
00:11

Mac Dictionary Services API Tease

Basic lookup working

I am a big fan of Dictionary.app. It’s pretty handy for English, but what makes it really shine is that it has a zomg-amazing Japanese dict called Daijisen. What’s more, with 10.8 Apple threw in German, French, Spanish, and Chinese dictionaries as well. However, after getting used to the app, while still using it, I have decided that it sucks. I should clarify. The dictionary content that it has is great. But the app itself is lacking some features that would make it so much more useful. I want search history that goes beyond a single app launch, and an interface for seeing what words I looked up when. I want to be able to export this list so I can make flash cards. I also want to be able to search by words other than by a ‘starts-with’ scheme, such as ‘contains’ or ‘ends with’, like many online dictionaries have.

For a while I just assumed there was no easy way of using the content the mac dictionary app uses. Then I found out about the mac dictionary services API, which looks promising at a glance.

I created the basic lookup pretty quickly while on the BART commute to work and back. But it became apparent after some tinkering that the existing API (which has only two functions for word lookup) is entirely incomplete. You can look up a string, and get a definition back from the same dictionaries that Dictionary.app uses. However, you can’t specify which dictionary, or which entry within a dictionary (e.g. for a word that has multiple definitions). This means you only will the first entry of the first dictionary that gets hit. So I decided to spend a good part of today trying to see what could be done about this.

I came across one or two or three interesting posts that showed some private API off. Most of these were for simple CLI programs, or for building their own dictionary. And used private calls such as DCSRecordCopyData() and DCSCopyAvailableDictionaries(). DCSCopyAvailableDictionaries allowed me to access specific dictionaries, and used in conjunction with DCSGetTermRangeInString and DCSCopyRecordsForSearchString, I was able to generate a reasonable list of candidate DCSRecordRefs entries from a single input word. The only thing missing was the definitions for each DCSRecordRef.

I wanted to make this dictionary for my Japanese studies. In Japanese, there are tons of homonym and homophones depending on if you use write the word with kanji or not. I didn’t see a function from my googles that would show the word listing, but doing an NSLog(@”%@”, record) on an example search for ‘いる’ showed some info about the DCSRecord structure:

lldb output:
{key = いる, headword = いる【射る】, bodyID = 111081}
{key = いる, headword = いる【要る】, bodyID = 104584}
{key = いる, headword = いる【居る】, bodyID = 163639}

It looked like the ‘bodyID’ or ‘headword’ field of DCSRecordRef had the most specific information about the result. So I went into the framework and searched for symbols of functions that might do the trick:

cd /System/Library/Frameworks/CoreServices.framework/Versions/A/Frameworks/DictionaryServices.framework
nm -gU DictionaryServices|grep DCS
0000000000007b06 T _DCSActivateDictionaryPanel
000000000000914d T _DCSCopyActiveDictionaries
000000000000916f T _DCSCopyAvailableDictionaries
[...]
0000000000007e07 T _DCSRecordCopyData
0000000000007e1e T _DCSRecordCopyDataURL
0000000000007e63 T _DCSRecordGetAnchor
00000000000076ef T _DCSRecordGetAssociatedObj
0000000000007e4c T _DCSRecordGetDictionary
0000000000007ebd T _DCSRecordGetHeadword
0000000000007e91 T _DCSRecordGetRawHeadword
0000000000007f04 T _DCSRecordGetString
0000000000007e35 T _DCSRecordGetSubDictionary
0000000000007e7a T _DCSRecordGetTitle
0000000000009181 T _DCSRecordGetTypeID
00000000000076d9 T _DCSRecordSetAssociatedObj
0000000000007ea8 T _DCSRecordSetHeadword
000000000000878a T _DCSSearchSessionCreate
[...]

The second entry for 棺 should be ひつぎ

There were about a 100 or so symbols or so that Apple didn’t feel like sharing via docs. I tried a few combinations, but it looks like DCSRecordGetTitle or DCSRecordGetRawHeadword produced the best strings for use with DCSCopyTextDefinition. This solves the homonym problem for the most part. However, this did not work at all with heteronyms (words that are spelled the same, but pronounced differently), since the headword/display title would be the same. For example, for the input 棺 I need the definitions for both 棺 read as ‘kan’ and 棺 read as ‘hitsugi’, but this method would give me two definitions for ‘kan’ instead. Eventually I gave up. I hope someone else figures this out. I put the intermediate result up on github. Let me know if you make any progress.

On iOS it goes without saying that you should probably avoid using private APIs, since the main means of distribution is through apple. On the mac distributing an app yourself is still viable, and thus you can use private APIs to your heart’s content. However, because the non-documented but exported symbols do not provide function signatures, it’s pretty much just a tease unless you want to spend a lot of time to figure out what each method takes. After googling symbol names, it seems dictionary services are relatively unexplored. But I’m mostly interested in this for making my hobby dictionary that I can nerd out on and add lots of obscure features to while using the apple-provided dictionaries.

09 Jun 13
23:13

Surreal Warp Nightmare – BaconGameJam 05


Surreal Warp Nightmare


BaconGameJam is one of several game jam/rapid development challenges like Ludum Dare and One Game a Month. The last jam took place over the weekend. The theme of “Lights Out” was announced friday night, and teams had 48 hours to design and implement a complete game. I signed up as a team after finding team members via a reddit post I made. The finished game is available here for mac/linux/win/android, including source on my github.

This was my first game jam, so I wanted to shoot for a pretty low target in terms of design. I saw previous jam results where many games were unplayable or not submitted, and this got me worried. Before the jam began I was already leaning towards a just a simple click control and a game mechanic that would wrap around it, because I wanted to be sure to finish.

concept/idea art by William Mckee

We brainstormed around the ideas of energy conservation, curfew, die hard, puzzles/game of life, and ultimately ended up designing a runner with a warp between color / grayscale on click. Enemies and good things like bacon would be able to interact with the player on either the color or grayscale background, and as the game progresses, they become less visible from the other warp background.

I had several motivations for joining this game jam. For one, it’s community interaction and as a SOHO whatever it feels good to work with others. Also, working with others is something I want to get better at. Working with a team I am comfortable and productive with is something I dream about. To get there this means I need to work with lots of people.

I also was interested in the challenge aspect. Not the competitive aspect, since this is my first jam, but rather the technical ability to prototype and design rapidly. Sometimes I struggle with longer term projects because I didn’t do enough prototyping and quick development. A 48 hour jam effectively forces you to identify critical design elements and prune the things that aren’t working. This form of greedy design is something that I have too often passed by because it does feel ‘greedy’; but now want to look at it as a tool for overall design rather than a restrictive philosophy.

There was also the technical challenge. These jams tend towards web apps because they are cross platform. However I do desktop and mobile apps, which means compiling to many different platforms and all the headaches that come along with that. However, this is something I want to become more fluent in, since lately I am using cocos2d-x, and the desktop cross platform ability is a major benefit of it that is often unused. BGJ actually gives you an additional day to package your game for additional platforms, but I ended up doing it all before the deadline, compiling a cocos2d-x project to mac/win/linux/android. Fixing cross platform stuff (microsoft’s C compiler oddities or GLSL differences) probably took 30% of the time, which is not so ideal, but now I’ve learned how to make things work with cocos2d-x on each.

Also, I like the idea of being able to learn some new skill with a game jam. I have been wanting to play around with GLSL shaders for a while but despite working with fixed-pipeline OpenGL, I never got around to the programming pipeline. The game jam was a perfect time to dive in. So the game has 4 different shaders from simple greyness/glow shaders, to give a tv-like channel switch effect on the warp, to texture deformation to simulate animation on static sprites. It was pretty educational and I can’t wait to use this on my next iOS game.

the warp swirl effect is done via a GLSL shader

Overall, I’m happy with the result. The game is playable and provides some basic fun. Although it’s not going to win any awards, I finished it, and I have a solid point of reference for what I can do in 48 hours for my next jam or prototype. I’m glad I worked on it as a team because the art (by William Mckee) and music (by Grant Matthews) make the game experience much more pleasant, and it was great to toss around ideas and get excited with real people. This was a really good refresher that has given me some new energy and perspective on my current projects too. Looking forward to the next one. If I have time later maybe I’ll do a post moretem on the cocos2d-x/C++ source/technical details and pitfall avoidance strategies I used, but if you’re interested in cocos2d-x alone maybe take a look at the github.

07 Apr 13
03:45

XCode: Dealing with “Could not inspect the application package” error

After setting up a new cocos2d-x project and having run successfully, after some minor changes I hit this weird error that had only a few hits in google, with no clear solution. To fix it I did the following things. I don’t know which one is the right solution.

1. There was a mostly unhelpful SO question – I tried cleaning to no avail, but the question reminded me to look in the device console (XCode Menu->Window->Organizer->IPhoneDeviceName->Console) to see what the error was. It was odd that the warning was in there:
peruse_package: App info dict loaded from "whateverMyAppBuildNameIs.app" did not have bundle identifier
So I looked in the directory that contained whateverMyAppBuildNameIs.app and think I saw that there was no info.plist. I added the info.plist to the target by clicking on it in XCode and clicking the box in “Target Memebership” on the right (Note: this is probably wrong as I had to undo this later – but it fixed the error and led me to a new one).

2. This gave me a new error that was already in SO where the selected answer did not apply – for some reason I got a “Could not launch myAppName – no such file or directory”. No idea why – I checked that the app existed at the directory. I did some combination of cleaning, deleting the app from my device and restarting xcode, and I got to the next error.


3. I then got a “XCode cannot run using the selected device – Choose a destination with a supported architecture in order to run on this device”. Okay, at least this was logical. So I added armv7 and armv7s, since I noticed these were in the target architectures but not in the valid ones (see screenshot). And then I cleaned, deleted the build dir at Users/your_usr_name/Library/Developer/Xcode/DerivedData/myappbuilddir12312314 manually, and then at some point I removed the target from info.plist (undoing step 1) and then it worked.

There’s a bunch of steps there were probably not necessary, especially since I had to redo one. As usual I’m too lazy to go through the bug again to disect it, I’m just posting this as a quick log. Please comment if you have a better clue than I do about any of the steps. It’s quite possible just the very last steps (everything in 3.) will fix it – if I had to bet money I would go with that. Putting everything in front of you in case it helps.

31 Mar 13
21:39

Cocos2D-X: Creating an iOS/android dual-platform project

I develop iOS apps. However, the first smartphone I had was an android device. I like open-source projects and bought it over the iPhone for idealogical reasons more than anything else. As a user, the android experience was okay, although iOS clearly had the apps and the polish. But when it came down to actually needing to make a surviving wage I decided to switch to iOS.

Lately, in podcast and forums, as well as third-party developer support tools, I’ve seen a comeback in indie iOS game developer interest for android, for whatever of the many reasons there might be from iOS over-saturation to app store visibility to the rising android markets. Especially for games. Some people are porting their obj-c projects to java. Some people are developing in Unity for the ability to deploy to both platforms. Unity seems to be great in that regard, but although free for PC/Mac deployment, it can cost a bit to deploy on iOS/Android at $400 a platform. Then there are the open-source “write once, run everywhere” solutions, for games such as libgdx and cocos2d-x. This is ironic because the slogan is from sun, but due to the walled garden nature of the walled garden, the term applies better to cocos2dx client code then the android java sdk.

cocos2d-x client code is in c++, which is my native tongue. Sweet. I had also played around a bit with cocos2d in obj-c and it was okay. So I decided to give cocos2d-x a shot to give android a second chance. Also you can supposedly deploy to windows 8 mobile and get $100 (note to self:not sure if joking).

Until recently I was one of the active contributors of Audacity, which is built using wxWidgets. For my last few sonfication/noisemaker projects I used Open Frameworks. wxWidgets and Open Frameworks are great because they let you write c++ code once and deploy on mac/windows/linux. But setting up the first project to work with each platform is a pretty big barrier for pretty much every cross-platform framework. This is just the nature of the beast – namely the inherent complexity of cross platform development.

There are a few guides out there on setting up your build for cocos2d-x for iOS and android. There are great guides, but they are about a year old, and so they didn’t work without modification, although they do get you most of the way. The main place to look was the ever-present raywenderlich.com’s cocos2d-x tutorial and it’s prerequisite eclipse/android tutorial. Then there was gamit.ro’s cocos2d-x tutorial, which leaves out the eclipse details but goes into some important stuff about android makefiles that raywnderlich.com’s doesn’t. Lance Gray from the cocos2d-x forums also mentioned there’s a nice guide at paralaxer.com that shows how to integrate iOS/android plus win. Anyway, I’m writing this as a supplement to those guides, and as a thanks to their authors as well as the various forum posts that allowed me to get the solutions faster than I could on my own.

In all, I think it took me about a day and some change to get everything set up the way I wanted it, even with these guides. All of the waiting had to deal with the install and setup of cocos2d-x’s android project setup, OpenGL ES 2.0, and trying to use a shared project directory external to the cocos2d-x root for source control management reasons. That being said, if you just want to try it and already have XCode working with iOS, you can easily follow the guides to get your cocos2d-x project setup in minutes. I’ll try to describe what I stumbled on, and maybe it can save you some time.

update – Instead of doing all of the 10 steps listed below, you should check out this python script that creates a shared project directory for ios/android/mac/win.

The script came out recently so I didn’t hear about it, but it should save a lot of time. You’ll still need to install android ndk/sdk, set up the environment variables, as well as the emulator trick I mentioned, but it’ll save you steps 6-10, and the necessary merging of the projects that I don’t cover yet. The script is at tools/project-creator/create_project.py, in the cocos2d-x root. The script is very simple, and creates all the projects in one line – Although I had the NDK_ROOT missing error I describe below. You can still use this guide – steps 1, 3, 4, 5 are still what you need to do for android sdk installation, and the rest might be handy for debugging errors.

0. Assumptions

I started with XCode/iOS environment already set up, so I’m not covering that. You should be able to run an app in the iOS simulator before starting here.

1. Get Cocos2D-X

You can download the latest stable (2.1.2 as of march 31) from their website. I should also note that alternatively, they also have a github repo that has the stable releases tagged. I forked this on my own github account so I can make changes and submit them. After cloning your fork you can checkout the stable tag). If you don’t know what I’m talking about it’s best to just use the website zip download.

The zip extracts to a directory named cocos2d-2.1rc0-x-2.1.2 (if you used github it defaults to cocos2d-x). Put this somewhere. I put it in my source control management folder (e.g. /Users/mchinen/scm/cocos2d-x). The location of this folder will be used in the future many times, referred to as the cocos2d-x root or as the variable $COCOS2DX_ROOT.

2. Set up Cocos2D-X to work with XCode

Go to the terminal, and cd to the cocos2d-x root and run install-templates-xcode.sh. Example

cd ~/scm/cocos2d-x
sudo ./install-templates-xcode.sh

This will install some xcode templates.

2.1 Test the XCode template

Rigid-body dancers endure awkward positions to make pyramid


Open XCode and go to the File Menu’s create new project. I wanted to use the chipmunk physics engine so I picked cocos2dx_chipmunk. You can pick the standard cocos2dx or box2d if you like, but just remember when you create your android project you will need to create the same type of physics engine.

I called my project ‘pet’ and put it at ~/scm/pet.

I was able to build and run in the simulator without any problems. I will note that there about 7 deprecated function warnings. Since I picked chipmunk it creates a scene which spawn lovely bald dancer sprite rigid bodies.

So that’s it for iOS setup, really. You now can modify and add cpp files as you like.
As I mentioned, the iOS setup is fairly painless. 95% of my time was spent on integrating android over the next few steps.

3. Eclipse setup

The (non cocos2d-x) android/eclipse raywenderlich.com tutorial I mentioned is a bit dated, and you only need one step from it for eclipse set up now in a bundle. Download the android sdk, which now includes eclipse+ADT plugin. There are also ways to expand an existing eclipse install (basically this follows the raywenderlich tutorial), but you can also just have two installs of eclipse (I have three installs of XCode on my computer).
The download comes as a zip and I extracted it to a folder in my home directory I use for non source control frameworks called ‘localframeworks’. So for me, the path to the android sdk bundle is /Users/mchinen/localframeworks/adt-bundle-mac-x86_64-20130219. If you ls it you’ll see it has two directories, one for eclipse, which contains the actual eclipse executable + adt, and another directory called ‘sdk’. When referring to the android sdk root, it’s this sdk directory, and not the parent. I suspect if you download the non-bundled (minus eclipse) version the directory structure is one less, with the sdk sitting on the top level.

4. Install C++ for eclipse

Go ahead and open the Eclipse application in the eclipse directory of the android sdk bundle.
If you are new to eclipse it will ask you about a workspace location. You can pick any user location for this – it won’t really affect the tutorial and does not determine the root directory of your android project. For example, I selected ~/localcode/. The default in ~/Documents is fine too.

Now once eclipse loads go to the help menu, and select install new software. Select ‘Juno’ (or whatever the current build name of eclipse is), and then check C/C++ development tools. Click next and install it. FWIW my screen didn’t match the raywenderlich tutorial, and had one less item installed. Here’s a screenshot:

5. Install the android NDK

Download it (you want the 64 bit one). I extracted it and put it in the same localframeworks directory that I put the android SDK in. So the location of android NDK for me is /Users/mchinen/localframeworks/android-ndk-r8e. This will be referred to as $NDK_ROOT in the future.

6. Create the android project

Remember how the XCode project had templates that would configure and create the xcode cocos2d-x project for you automatically? It’s likely that some kind soul will make them some day but for now that doesn’t exist for eclipse yet. Fortunately for you, there is a (command-line) shell script that can do this. It’s called create-android-project.sh and it’s in $COCOS2DX_ROOT.

Remember how you could pick wherever you wanted to place the objective-C project? You don’t have that choice anymore either. To use the script, you need to cd to $COCOS2DX_ROOT and run the script there – calling it externally yields an error. This is not ideal if you want to use source control management like git, but we can fix this later at the cost of some minor pain.

First, you need to create some shell environment variables that the script uses. You can run these directly in the terminal and the variables will exist for the terminal session, but I like to put them (append them) in ~/.bash_profile.

export ANDROID_NDK_ROOT=~/localframeworks/android-ndk-r8e
export COCOS2DX_ROOT=~/scm/cocos2d-x/
export NDK_ROOT=$ANDROID_NDK_ROOT
export ANDROID_SDK_ROOT=~/localframeworks/adt-bundle-mac-x86_64-20130219/sdk
export PATH=$PATH:$ANDROID_NDK_ROOT

Now that you have the shell variables set, we’re ready to run the script. The usage of the shell script is

./create-android-project.sh [-b|--box2d] [-c|--chipmunk] [-l|--lua]

So to create an android project that links to the chipmunk framework, you should append the -c flag.
You’ll be prompted for a few things like your package path (you can use your intended iOS bundle. For example, ‘com.michaelchinen.pet’). Pick a target SDK which is (I used 2.2), and then lastly choose the name of your project (I used ‘pet’.

For completeness, this is my truncated output example from the script:

mcair:cocos2d-x mchinen$ ./create-android-project.sh -c
use global definition of NDK_ROOT: /Users/mchinen/localframeworks/android-ndk-r8e
use global definition of ANDROID_SDK_ROOT: /Users/mchinen/localframeworks/adt-bundle-mac-x86_64-20130219/sdk
using chipmunk
Input package path. For example: org.cocos2dx.example
com.roughsoft.pet
Now cocos2d-x supports Android 2.2 or upper version
Available Android targets:
----------
id: 1 or "android-7"
     Name: Android 2.1
     Type: Platform
     API level: 7
     Revision: 3
     Skins: HVGA, QVGA, WQVGA400, WQVGA432, WVGA800 (default), WVGA854
     ABIs : armeabi
----------
id: 2 or "android-8"
     Name: Android 2.2
     Type: Platform
     API level: 8
     Revision: 3
     Skins: HVGA, QVGA, WQVGA400, WQVGA432, WVGA800 (default), WVGA854
     ABIs : armeabi
----------
id: 3 or "android-10"
     Name: Android 2.3.3
     Type: Platform
     API level: 10
     Revision: 2
     Skins: HVGA, QVGA, WQVGA400, WQVGA432, WVGA800 (default), WVGA854
     ABIs : armeabi, x86
----------
[...]
id: 13 or "android-17"
     Name: Android 4.2.2
     Type: Platform
     API level: 17
     Revision: 2
     Skins: HVGA, QVGA, WQVGA400, WQVGA432, WSVGA, WVGA800 (default), WVGA854, WXGA720, WXGA800, WXGA800-7in
     ABIs : armeabi-v7a
input target id:
2
input your project name:
pet       
Created project directory: /Users/mchinen/scm/cocos2d-x/pet/proj.android
Created directory /Users/mchinen/scm/cocos2d-x/pet/proj.android/src/com/roughsoft/pet
Added file /Users/mchinen/scm/cocos2d-x/pet/proj.android/src/com/roughsoft/pet/pet.java
Created directory /Users/mchinen/scm/cocos2d-x/pet/proj.android/res
Created directory /Users/mchinen/scm/cocos2d-x/pet/proj.android/bin
Created directory /Users/mchinen/scm/cocos2d-x/pet/proj.android/libs
Created directory /Users/mchinen/scm/cocos2d-x/pet/proj.android/res/values
Added file /Users/mchinen/scm/cocos2d-x/pet/proj.android/res/values/strings.xml
Created directory /Users/mchinen/scm/cocos2d-x/pet/proj.android/res/layout
Added file /Users/mchinen/scm/cocos2d-x/pet/proj.android/res/layout/main.xml
Added file /Users/mchinen/scm/cocos2d-x/pet/proj.android/AndroidManifest.xml
Added file /Users/mchinen/scm/cocos2d-x/pet/proj.android/build.xml
Added file /Users/mchinen/scm/cocos2d-x/pet/proj.android/proguard-project.txt
Resolved location of library project to: /Users/mchinen/scm/cocos2d-x/cocos2dx/platform/android/java
Updated project.properties
Updated local.properties
Updated file /Users/mchinen/scm/cocos2d-x/pet/proj.android/proguard-project.txt

This then creates a directory called pet within the cocos2d-x directory. If you open it you should see another directory called pet, as well as a very important directory called proj.android. This directory is kind of like your xcodeproj file, except it contains a lot more things including build output.

7. Build the android project.

If you cd to the proj.android directory, you can then build the project by running the build_native.sh script:
Here’s what I got (middle truncated, again)

cd ~/scm/cocos2d-x/pet/proj.android
./build_native.sh

NDK_ROOT = /Users/mchinen/localframeworks/android-ndk-r8e
COCOS2DX_ROOT = /Users/mchinen/scm/cocos2d-x/pet/proj.android/../..
APP_ROOT = /Users/mchinen/scm/cocos2d-x/pet/proj.android/..
APP_ANDROID_ROOT = /Users/mchinen/scm/cocos2d-x/pet/proj.android
Using prebuilt externals
make: Entering directory `/Users/mchinen/scm/cocos2d-x/pet/proj.android'
Compile++ thumb  : game_shared <= main.cpp
Compile++ thumb  : game_shared <= AppDelegate.cpp
Compile++ thumb  : game_shared <= HelloWorldScene.cpp
[...]
StaticLibrary  : libchipmunk.a
Compile thumb  : cpufeatures <= cpu-features.c
StaticLibrary  : libcpufeatures.a
SharedLibrary  : libgame.so
Install        : libgame.so => libs/armeabi/libgame.so
make: Leaving directory `/Users/mchinen/scm/cocos2d-x/pet/proj.android'

So at the end you can verify that chipmunk is being included in the build.
Great.
We’re ready to use this project in eclipse.

8. Configuring the eclipse project

Go to Eclipse’s File Menu->New->Project.
Then select Android Project from Existing Source in the next dialog.
After that you will be prompted to point to the existing source.
You want to point to the proj.android directory as in this screenshot.

After this pick an android target in the next dialog. I chose 2.2.

This creates a project in eclipse. Unfortunately, mine had errors that said “Cocos2dxActivity could not resolve to a type”.
I’m not sure what the best solution is, but here’s what I did. We need to get the Cocos2dxActivity stuff into our workspace, but stay outside of the pet project. So go to File->Import… and pick “Android->Existing Project into workspace”. Then choose the directory at $COCOS2DX_ROOT/cocos2dx/platform/android/java.

Now you have two projects in your workspace, pet, and the cocos2dx java libs in a project called libcocos2dx. You need to add a reference so the pet project knows to look at the cocos2dx java files. To do this, right click on the pet folder in the package explorer in eclipse, and select Properties. Then go to Java Build Path->Projects->Add… and select the libcocos2dx. This should fix the errors in the pet project, meaning that you’re ready to build and run on a device.

Note – You can fix this error by using the ‘link source folder’ option instead, and it will work, but I think this incorrect – please comment if you know for sure.

So you’re ready to run. At this point using cocos2dx 2.1.2, you still can’t run in the emulator because there are issues with OpenGL ES 2.0. But if you have a device that supports it, you should be good to run after hitting the play button and selecting the device. I busted out my 4 year old HTC Hero, flashed to android 2.3, and was hoping for something, but it just doesn’t support ES 2.0. Also, as important as it is to test on the device, the emulator can be a good way to get an idea of how your game deals with whatever resolutions you don’t have devices for. So I had to get the simulator working.

This thread describes how to do it. So if you need emulator support, you need to edit one of the java files in the libcocos2dx project. The patch code on the forum is missing some includes, so here’s the diff from my github that shows how to modify the file.
The file you need to edit is cocos2dx/platform/android/java/src/org/cocos2dx/lib/Cocos2dxActivity.java

 
@@ -27,9 +27,11 @@
 import android.app.Activity;
 import android.content.Context;
+import android.os.Build;
 import android.os.Bundle;
 import android.os.Message;
 import android.view.ViewGroup;
+import android.util.Log;
 import android.widget.FrameLayout;
 
 public abstract class Cocos2dxActivity extends Activity implements Cocos2dxHelperListener {
@@ -139,6 +141,10 @@ public void init() {
         // ...add to FrameLayout
         framelayout.addView(this.mGLSurfaceView);
 
+        // Switch to supported OpenGL (ARGB888) mode on emulator
+        if (isAndroidEmulator())
+           this.mGLSurfaceView.setEGLConfigChooser(8 , 8, 8, 8, 16, 0);
+
         this.mGLSurfaceView.setCocos2dxRenderer(new Cocos2dxRenderer());
         this.mGLSurfaceView.setCocos2dxEditText(edittext);
 
@@ -150,6 +156,19 @@ public Cocos2dxGLSurfaceView onCreateView() {
       return new Cocos2dxGLSurfaceView(this);
     }
 
+   private final static boolean isAndroidEmulator() {
+      String model = Build.MODEL;
+      Log.d(TAG, "model=" + model);
+      String product = Build.PRODUCT;
+      Log.d(TAG, "product=" + product);
+      boolean isEmulator = false;
+      if (product != null) {
+         isEmulator = product.equals("sdk") || product.contains("_sdk") || product.contains("sdk_");
+      }
+      Log.d(TAG, "isEmulator=" + isEmulator);
+      return isEmulator;
+   }
+
   // ===========================================================
   // Inner and Anonymous Classes
   // ===========================================================

After that, you’re almost there. Open the android device manager in the window menu of eclipse.
Click New…, then create a device. I created a galaxy s3 on 4.0. The main thing is you click the use host GPU button. Without this, the app would just crash. Here’s what my device config looks like:

After setting this, run, and you should see a cocos2d logo and fps. Note that it isn’t the same as the iOS one, but at least something is on the screen and not crashing.

9. Convert to C++ Project

What remains to be done is to convert the android project to C++ for convenience, and then to merge it into the iOS project. You don’t really need to convert it, since you already have a running application, but this will make it so you can stay in eclipse and don’t need to switch to the terminal to build the C++ each time you make a change.
Ok, honesty time. It’s been a few hours since I started writing and I’m getting lazy now, so you can follow the raywenderlich.com guide to convert to C++, but I will note I hit an error after creating the C++ project where building would fail with a “Please define NDK_ROOT”
Here’s how to do that:
right click on the project->Properties->C/C++ Build->New variable.
For the name use NDK_ROOT. For the value use ${HOME}/localframeworks/android-ndk-r8e



And all should build and run again.

10. There is no 10

After that you can follow the other tutorials on how to merge the projects. I’ll note that the raywenderlich one doesn’t actually fix the issue of having two unrelated directories, with the android one inside of the cocos2d-x directory.
I fixed this using a different method so that I could have both projects in one directory. Originally I planned to include that as well, but my time spent on blogposts is more than up, but maybe if there’s interest I’ll post another update on this. For now I need to get back to actually using the framework!

One other note that

21 Dec 12
18:46

XCode 4 with a networked drive: fixing ‘null character(s) ignored’ and ‘missing @end’ warnings

RIP command-option-up header/source switching, format indent

XCode 4 had a lot of changes over 3.x. Some seemed kinda okay, others seemed horrible. When it came out I did… absolutely nothing, for as long as possible and held on to the 3.x version as long as it was possible to do so.
Since I had to switch to the iTunes-ized Xcode 4, I’ve been using it only as a compiler/debugger, doing all my text editing in emacs on my linux box, using synergy to swap to my mac to compile. I want my editor to be on the same box as as the source code files in case something weird with the network or filesharing service happens. So the project I compile on my mac is served by smb on my linux box. This may sound needlessly complex, but I like having two boxes even after trying dual-screen mac. Also, I had already set up smb to my liking before trying it out, and I already was playing with this even while xcode 3 was current.

With XCode 4, I noticed something weird happening when I made multiple saves in a short duration, e.g. two quick saves in under a minute. The saved changes wouldn’t go into the build for whatever reason. I remember a similar bug with XCode 3, but if you just waited a second and recompiled, it would be okay. But with XCode 4, sometimes the new changes would never get propogated across the network and compiled even after waiting for a long time. I eventually realized if I saved the file on my linux box again by making a small edit, XCode would pick it up and compile it, but oddly enough, only if the save happened a minute or so after the first.

However, after inspecting the fileshared source code on my mac with other editors, it was clear that this wasn’t a filesharing issue, but simply XCode caching the file. The “Touch command” that existed in XCode 3 would presumably fix this (although I would be afraid to use it if caching is the suspect), but it was inexplicably removed with XCode 4.
When the caching is an issue, you get a warning that looks like ‘null character(s) ignored’ or an error that says ‘missing @end’ and sometimes the file will be cut in half in XCode.

After dealing with this for over a year I just found out you can prevent these by making sure you have no source files open in XCode. It seems the editor uses it’s own cache, and the compiler goes directly to compiling the cache instead of the disk. Note that if you click on an error in the left side panel, this opens up the editor. So now I just create a new tab with a .png file and close all other xcode tabs. When googling this issue over the last year I saw no mentions of this bug, so maybe no one else uses XCode networked. Noting here just in case.