TV's are small, plus a little Clutter

So I was sitting watching my 40" HDTV the other day, and started thinking about why web browsing on a TV pretty much sucks, even with all the pixels you get at 1080i or what have you. And trust me, I tried it with our PS3 and other than watching YouTube videos, it's not a good experience.

But while I was sitting there on my couch (or sofa, depending on your English dialect) a pretty normal distance from the TV, I held my hands up to frame the size of the picture at arms length. Then lowering it to may lap, it struck me. The size of the picture isn't any bigger than a handheld gaming box, maybe slightly bigger than an iPod Touch (as I try to find one for my son's birthday on Thursday :( ).

The reality of the situation is that even with the higher pixel count, you still need to treat set top boxes as mobile, but not mobile, internet devices and entertainment units. And that especially goes for the UI. Don't try running GNOME on it, that's going to be brutal.

So I'm off taking another look at mobile devices and the user interfaces they present. The newest one is the 2.0 alpha release of Moblin, Intel's effort at a Linux distro for mobile internet devices and netbooks running their chips. The video in the LinuxDevices.com article is intriguing, and made me go look at what technology they were using to present their 3D animated GUI.

Well, it turns out to be another open source project called Clutter. They produce a library that abstracts away the grunge of OpenGL and OpenGL ES to build user interfaces. You create Actors that have images and such and declare their animation and event handling and then fire off into an event/display loop. You get pretty cool effects with not too much code.

Now, I have to pick at the choice of GTK as their paradigm mentor and, yes, if you're used to GTK programming, doing Clutter will be natural, but if you're like me and fell in love with the Qt and it's elegant use of C++, then you'll be a little put off. I did find a clutter-qt integration in their repo, so maybe you'll be able to do both in the future.

Someone once said, and I think he lived in Redmond, Washington, that there was no innovation in open source. This is a pretty significant counter to that. This project has been around for a while, sprouting out of the need to add GUIs on top of the new fancy 3D graphic chips appearing in handheld devices. They have a innovative and game changing solution. The just need people to discover them, and Intel, who also happens to be their new boss, is helping with that.

Can an LGPL Qt give C++ a lift?

Black Duck recently announced their top Rookie Open Source projects for 2008 which using a bit of a weird metric, revealed the top 10 open source projects that were created in 2008 that had the highest number of releases. More releases makes you good? O.K...

Anyway, the most interesting information from their news release was the stats they gathered on what programming languages these new projects were using. To the surprise of many, 47% of them were written in C (C Rules!). That was followed by 28% in Java and 20% in JavaScript. It's pretty interesting there was so much JavaScript usage. And again, these were projects that have just been created. But when you look at it, most open source projects target Linux, and by far the most popular language for Linux is still C.

One thing I noted, though, was that C++ wasn't even mentioned. It could be that they lumped C++ in with C, but I have my doubts. I rarely do see C++ in open source. The large open source game engines, like Ogre and Irrlicht as well as Firefox (of course), are in C++, and OpenOffice is written in every language imaginable including C++, but I see C way, way more.

[Watch out, bad segue ahead...]

I spent part of last weekend taking a deeper look at Qt, with its upcoming spanking new LGPL license. I have to admit I'm a GPL library bigot and kept away from Qt because of that, but boy do I regret that now that I've seen it. It's an incredibly complete C++ library for building apps of all kinds. And it has everything I've been looking at lately, WebKit, JavaScript (well ECMAScript but that's the same thing), and OpenGL, and incredibly smooth integrations between those and many more components.

So as Qt makes this transition, I have a feeling it's going to gain in popularity everywhere. And I think it'll show the power of C++ and pull a lot of the developers writing for the C-based GTK away. Heck even Ubuntu is thinking of switching to it for their mobile platform.

Kudos to Nokia for making this decision. I think it's going to pay dividends for them as developers take a fresh look at a great framework. Which, BTW, means there will be more developers working in the same environment that also happens to run on Nokia's phones ;).

J2ME? Why?

I was just reading the slides presented from the kick-off meeting of the Eclipse Mobile Industry Working Group. I believe this is the first working group at Eclipse and I think it's a great concept. Bring groups of companies together that are interested in the same or similar technologies and do some planning. Hopefully that will result in new investments in various Eclipse projects.

Anyway, one of the examples of work environments shown was for a J2ME developer. The first thing that jumped into my head, and of course I'm writing this entry without thinking more so I may come off a bit misinformed here but hey I'm just the dumb C++ guy, but who cares about J2ME any more? With the rich mobile development environments provided with Android, the iPhone, the new Palm Pre, and even Qt for mobile devices, why would you do J2ME development any more. Isn't there much more opportunity for greater riches writing apps for these new and wildly popular environments?

Anyway, feel free to comment and tell me the way it is. And I'm sure the J2ME people over in DSDP are the right people to do that ;), since they know that community. But I am curious about whether the J2ME community is still on the rise, or whether there is a migration happening to these new technologies.

Time to Get Qt?

Now this is interesting. I've mentioned a few times in my blog that the world would be a different place if Qt was given a free commercial friendly license, like LGPL. Of course when Trolltech was an independent company, that would have killed all their revenue. But now that they're owned by Nokia, I guess the time has come for them to make the change.

And I think this will open people's eyes to Qt. It's certainly a very rich framework giving pretty much everything you need to make a truly cross platform application, i.e. #ifdef free. And it's used in some very popular applications like Skype, Google Earth, and the VirtualBox manager. And, of course, it's the foundation of the Linux KDE desktop environment, which has it's devoted fans.

And again, being LGPL, I'd expect to see it used by more commercial applications. Heck, it will now pass my policy and I'll be able to include it in Wascana, and the SWT developers will also be allowed by their lawyers to write the port against it. Who knows...

But as I scout the horizon of desktop and mobile apps, I wonder if the apparent momentum away from C/C++ has become too great for this to make a significant splash, or will it just be a ripple. Maybe my head is too deep in WebKit these days and my view is getting tainted, but the web is slowly taking over. I guess the one thing Qt has going for it is a decent WebKit integration so maybe they can get the best of both worlds. Either way, it's definitely time for me to take a deeper look at Qt (and it's CDT integration, of course ;)).

Palm Pre and WebKit

I've been following the Palm Pre story a bit the last few days. While the technical details are still pretty sparse, it appears that one of my predictions for 2009 is already starting to happen.

My understanding, and I hope it isn't coming from sources who are also using the same technique to guess at the architecture of this thing, is that the UI for the Palm is rendered totally using WebKit. It appears that the applications for this device are written in JavaScript and use HTML and maybe WebKit's SVG support to render the graphics. Hell, maybe it's even using Dojo to make things look really sharp.

If this is true, then it's going to be a great test of how well this architecture works. I have my worries about how JavaScript scales and how easy it is to write traditional GUI apps, even handheld ones, using HTML as a rendering engine. But looking at the screenshots, it looks pretty awesome.

The other thing I notice is that there is a continuing trend of making it very difficult to build native apps that draw on the screen with these things. It started with JavaME in the "old days" and is continuing today with Google's Dalvik Java VM and now Palm's WebOS WebKit thing. They promise the power and openness of Linux and then shut the door. It's too bad, since a lot of these handhelds have 3D graphic acceleration in their SOCs, and you really need to go native to build a good 3D game or what have you.

I can't wait to see what the Pre SDK looks like, and whether developers buy into this architecture of GUIs based on web technology. And it'll be interesting to see how good the apps can be with it. But if they're as good as the prerelease demos, it's something to pay attention to.

Zune 30, Killed by Complexity

I first heard of it early New Years Eve, I guess. Hoards of Microsoft Zunes were committing mass suicide (a gruesome thought but the actual quote from the Slashdot article). Fears rose that some Y2K thing was happening, mind you things like that didn't happen in Y2K, at least not on this scale. Microsoft finally confirmed the issue as such though, a device driver hang on the 366'th day of a leap year. I'd love to see that code...

Well, thanks to the wonders of the internet, here it is! (I imagine this link will fall dead as soon as the Microsoft cronies make the rounds, as they should. It does have a Microsoft copyright). I actually found it through another blog where the guy put together a pretty good analysis of the problem.

The root cause? A brain fart. Either someone was in a hurry, or they couldn't handle the complexity of the algorithm once they started dealing with leap years. People blame testing for not testing all the paths. But, if you don't take the time to test all the paths, or don't have the skills to properly enumerate all the paths, testing isn't going to matter. At any rate, another great software engineering lesson learned for us all, just like the unhandled exception in the Ariane-5 rocket, except this one is recoverable and isn't as expensive (unless the Zune market share dives as a result, which could happen).

I spend a lot of my time these days working on software architectures and trying to come up with the most simple, extensible, and future proof. But none of that matters if you have code like this. And I've seen code like this all through my career. Hell, I've written some of it. But one thing I learned early from one of my great profs back at the U of S, was on code complexity. It is even measurable by counting the number of paths through your code. Complexity bad. Which implies that having more paths than you need is bad. As we see here, it becomes too difficult to test fully.

And that is certainly the case here. Too many 'if's. How do you convert days since 1980 into a time structure? Well in the Zune code (which is actually a common device driver in a number of Windows CE platforms), one of the paths leads to an infinite loop, when days is 366 and IsLeapYear is true. The author of the blog proposes a much simpler algorithm that works correctly but reduces the paths thus eliminating the bad one. I think you can make it even simpler.

One of my mantras is "I hate typing!". Of course it has nothing to do with typing (much). It's about producing simple elegant solutions to simple problems. Yes, Keep it Simple S(favorite ending here). Saves time. Saves money. Saves embarrassment. Saves your job. I'd hate to be the guy who wrote this code...

Predictions for 2009

I'm not usually one to make predictions. It's hard for me to tell the difference between a prediction and wishful thinking. But this article over at the Inquirer (still the best place to get an honest take on the industry along with /.) got me thinking about a couple of things I think are going to be important in 2009. So here we go...

2009: The Year of the GPGPU

This is more a continuation of a trend but the Inq article made some great points that I think will put some spotlight on general purpose programming with GPUs. The key one, is the recent standardization of a cross platform way of programming these things, OpenCL. ATI and nVidia have already signed up to provide OpenCL support for their chips and look for Intel's Larrabee platform to come with the same. I think there is still some software and hardware architectural things that need to be done to make GPGPU more efficient and easier to program. Look for LLVM (which needs an article on it's own) to play a role, as it already is with OpenGL, and look for one of the chip vendors to put a GPU on the memory bus shared with the CPU and make these things sing.

2009: The year of WebKit

Ok, yes, I'm playing it safe with these predictions. WebKit is already the base for Apple Safari, Google Chrome, and a host of Linux based browsers, so it already has a ton of momentum. The reason I think WebKit is going to the next level, is first of all the top of the class performance of it's new JavaScript VM (and I can't imagine why Google would continue with V8 in Chrome). But also, I am impressed with how easy it is to create your own WebKit based browser, and how easy it is to create a Linux based platform that uses WebKit as it's front end (launch X, launch a simplified WebKit shell in fullscreen, done). I expect to see a lot more mobile internet devices built this way. At the very least, it gives a reason for embedded developers to care about AJAX.

C++0x won't be C++09

I think that's a forgone conclusion but no one really wants to admit it yet. But look for the vote to finish this year at least. C++0x will be an exciting evolution of C++ into the next generation. No it doesn't have garbage collection, yet, but it does have smart pointers that do the job better if you use them right. C++0x makes it easier to do a lot of things, and the introduction of closures and lambda functions and expressions will breath some life into this stalwart of the software engineering community.

Well, that's it for now. If I think of more over the next couple of days I'll post them. There are a lot of things I hope will happen, but i'm not sure they will. But one thing is for sure, open source is here to stay and is becoming a core business model that companies still need to understand and learn to use effectively and I will continue with my work with Eclipse and Wind River to help figure that out and spread the word.

Have a safe and happy New Year! See you on the other side.

A look at WebKit

A few days ago, I was playing with Google's V8 JavaScript VM library and got it compiling with MinGW in Wascana. I submitted the patch to make it work but I haven't heard back. I guess it could be the Christmas break.

But one thing that struck me odd recently was an announcement that the next rev of Android would include WebKit's SquirrelFish Javascript VM. I guess that shouldn't be too surprising since SquirrelFish comes with Webkit. But then why is there ARM support (the CPU for Android) in V8? And if they are using SquirrelFish for Android, why don't they use the souped up SquirrelFish Extreme for Chrome? Especially since there are benchmarks showing it beating V8. I'm confused and can only chalk it up to Google being a big company and maybe the Android people don't hang out with the Chrome people.

Anyway, that got me looking into this whole WebKit business. I downloaded the latest nightly source build to my Debian Linux VM and after installing a boat load of packages needed to build it, I built it. I had heard the JavaScriptCore library which implements the VM was embeddable in C++ apps. The header files are there, but it looks like you actually have to embed the whole WebKit library to get at the VM.

That got me thinking back to an earlier idea I had. Use HTML with JavaScript as your main GUI framework. With Webkit, you can embed the whole browser into your application, and you can hook up new JavaScript classes to your C++ classes to provide scripting and to give access to them to the UI. Interesting to see how that would work in action.

I think I'm starting to figure out this whole JavaScript and C++ thing, with thanks partly to something a commenter said on a previous entry. Use scripting for quick turnaround, when you want to whip up a prototype or allow for easy extension of functionality. But use C++ for areas where you need to engineer functionality. Part of your architecture design is deciding what that means. And maybe something like WebKit might be the right platform to get you off the ground.

VirtualBox 2.1 and assorted Christmas Fun

Just some random thoughts on this Saturday after Christmas. My family and I had a good Christmas, despite a little "Fun with Autism" moment with my Autistic son, but it's all better now (patience is a key survival technique in our household). Yesterday was Boxing Day in Canada, which is a holiday here despite all the stores being open for your shopping pleasure. If you don't feel like going out, you are free to sit around, well, like boxes, which we did for the most part.

I'm spending a little time today while everyone is playing on the PS3 and various PCs around the house getting ready for my EclipseCon tutorial. I'm really looking forward to it. By the end of the tutorial, you'll walk away with Wascana which you use to build qemu, a little Debian Linux image running in that qemu, and a cross-compile toolchain and CDT integration that you also get to build to create apps for Debian from Windows (and maybe Linux). Lots of hands on and hopefully an appreciate of why the CDT is the first class cross-platform C/C++ development environment.

Before I get back into playing with qemu, it was cool to see a new version of the VirtualBox emulator come out, 2.1. It's a minor version increase but there are two significant features added. One, is 64-bit support on 32-bit platforms. This is critical for me and my installer work at Wind River, where I need to test and debug on 32-bit and 64-bit platforms. I don't trust 64-bit Linux enough yet to make it my main Linux environment, not to mention downright fear of 64-bit Windows.

The other cool thing is more on my personal interest front. They have an initial release of OpenGL support. If you read this blog regularly, you'll know I have a dream of an open Linux-based game console/multimedia set top box. I'd like to try some ideas out on a Linux platform with 3D hardware without actually buying any and this is the first emulator to have OpenGL support.

Unfortunately, they only have Windows guest drivers at the moment but have promised Linux/X drivers soon. I can't wait, but it does lead me to drop my plans for working on OpenGL support for qemu. Instead, I really need to spend what little hobby time I have learning how to write an X window manager, using a cross-compile environment with the CDT, of course ;)

I could have had a V8, oh wait, I do

I've always been intrigued by programming languages and what makes them tick, and what is the best one for what situation. That's why Dave Thomas's keynote at ESE still has me thinking about the mix of JavaScript and C++. So much so that I spent a few hours this weekend while waiting out the snow storm to get Google's V8 JavaScript VM building under MinGW for Wascana. I think it would be an intriguing addition to have the VM DLL available for developers using Wascana. With a few changes, I have it building and passing the unit tests and I have a patch into the V8 project. I'll make V8 available in the Wascana 1.0 alpha in the next couple of days.

Now that I have it, I have to ask myself - what the heck do you do with it? I've thought about building wrappers for the wxWidgets library to let you build thick client apps in JavaScript. wxWidgets also comes with Wascana, and thick client apps is kinda what Wascana is all about (aside from dreams of using it for game development, which could also benefit from a fast JavaScript engine).

But it's not clear where one would draw the line between JavaScript and C++. Given a C++ library like wxWidgets, or SDL, or what have you, is it enough to wrap it with JavaScript and have the developer do everything in JavaScript. Or should JavaScript just be this thing on the side that allows for extensibility of some larger application written in C++.

It makes me wonder if I'm following some crazy idea that some madman sold me in a bar in Germany. Or maybe this is challenging me to give it deeper thought, to think about how scripting and native languages are supposed to mix. Where in all this is the sweet spot of architectural balance. Or is there one? Either way, it'll be on my mind over the Christmas holiday season.

Fun with FEEDJIT

I'm not sure if you noticed, or are reading this blog from one of the syndication sites it gets copied too (like Planet Eclipse, or the Wind River Blog Network). But if you check back to the original site and scroll down a bit, you'll see a new panel called the FEEDJIT Live Traffic Feed. I know people express concerns about web things following them, and if I get enough negative response to it I'll pull it off. But in the meantime, I'm spellbound by this feature.

I'm learning quite a lot about the audience for this blog. The traffic feed gives me the city that where the person was, which is spread throughout the world, as well as a hint at how they got to my site. A few people come directly, I guess from an RSS reader where they've subscribed one way or another (Thank you!). More often, though, people end up here based on google searches, and I get the snippet that they were searching for! Creepy, but very useful.

So what are people searching for that pulls up my site? Well a lot of it lately has been the topics I'm most interested in lately, and that's CDT for Windows development, including Windows cross to Linux. It's good to see the interest from the community on that and I am continuing working on Wascana 1.0 as I write this (SDL is building in the background). I also often get a few queries on the Subversion Eclipse plug-in wars (I hate both right now, go git!). And you get the odd one looking for help, like today's "eclipse CDT autocomplete crap" (yeah, it has issues if you're environment isn't set up).

Anyway, it's pretty interesting to watch, and it humbles me immensely to see people from around the world reading what I write, especially when the google search reveals they searched for me by name. But I love to write and share my thoughts and I really appreciate it when people leave comments. Whether I agree with them or not, I always learn something from what they put there. It's a lot of fun and I encourage everyone to do the same. There will always be someone out there interested in what you have to say.

Fun with my little VIA console

At the Embedded Systems Conference in San Jose this year they handed out little VIA embedded EPIA systems to the attendees. I'm not sure everyone got one, but I was thrilled. It has a embedded VIA processor with a chipset that includes Unichrome 3D graphics, and also include a hard drive, ethernet, VGA, four USB ports, and audio in and out. It's a cool little unit.

I haven't done too much with it, but thinking about this Open Console concept (set top box with 3D graphics running Linux), I thought I'd try setting it up with some of the things I had in mind. I started by putting the Debian lenny installer onto a USB stick and installing from it. That was a little tricky until I reformated my USB stick and put syslinux on it properly. I installed enough packages to get X running with the openchrome driver for 3D graphics. glxgears ran pretty smoothly which gave me some hope I could actually use this thing to run games.

So I got adventurous and installed Nexuiz, an open source first person shooter. To my surprise, this and other open source 3D games are available from the Debian package repository. So a quick little 'apt-get' which brought down around 450MB of game, and I was off and running. We'll off anyway. I got about 20 seconds per frame, which makes it a little hard to even notice the thing was running.

Anyway, I tried a few other simpler games and they actually worked. I had to force myself to go to bed while hooked on billards-gl. It was fun. But I've slowly begun to realize that games built for the desktop aren't really ready to be played with only a joystick as you'd likely only have in a set top box scenario. So there would be work to be done.

I also started to understand first hand the commercial opportunity behind Linux, embedded Linux especially. Sure you can install a Linux distro and get a desktop environment up without too much effort. But try to do anything off that beaten path and you're in for a lot of work. If you can share in that work, fine. If you can pay someone to do it for you for cheaper than you could do, even better.

I also gave up on using this little VIA box for my play-totyping (hmm, new word). I need to start getting ready for my EclipseCon tutorial which will help me get back into the guts of qemu. Maybe I can do a little work there to bring GLX emulation to it, play time permitting, of course. Or maybe I'll shell out the $500 bucks to build a real system. Though playing in qemu would be funner...

Time for Distributed Source Control is Now

Imagine this scenario. You're part of a small team that's been following the CDT closely and have adopted it as the IDE for your commercial platform. You grab the CDT source at times convenient to your product deliver schedule and work on a local copy fixing bugs you find as you go through product testing. You're not a committer but you do submit patches from time to time and hope that the CDT team picks them up. But they're often busy with their own delivery schedules and the patches often grow stale and fall off everyone's radar.

So you live with your CDT fork and struggle every time you have to update to a new CDT version, so you don't do that very often. And since you're busy struggling in that environment, you really don't end up with time to get more involved with the CDT. You are a small team and you only have so much time in the day. You run into Doug once in a while at the Eclipse conferences and talk about what you do and promise you'll figure out some way to get more involved, but he knows your story too well and doesn't put much faith in it despite his appreciate for your intentions.

Sounds like I have experience with this, don't I. This scenario is too real and I'd bet is very common across all open source projects. Relying on CVS and Subversion at Eclipse with access controls limited to the select few committers makes it very difficult for those on the fringes to get more involved. It truly is a have/have not environment. The committers have it easy, checking in their changes whenever they want and those that aren't are struggling to keep up, or simply fork and go their own direction.

I've learned that the new Symbian Foundation as selected Mercurial as their source control system. Along with Linus's git, it's one of the new breed of distributed source control systems. These systems allow for multiple repositories and provide mechanism to pull and push changes between them. The introduction chapter of the Mercurial on-line book provides a great description of why this architecture works well for large globally distributed projects.

I invite everyone to read it, especially the Eclipse community. Because I think we need this kind of capability now. CDT needs an infusion of new blood and I know there are a lot of people who work with the CDT code base but have only a limited time to contribute back. If we had the infrastructure to better support them and make it easier to pull their changes into the CDT main line, and easier for them to keep up with everyone else's changes, it could be the formula we need to grow.

x86, the ultimate applet engine?

I need to watch out or people will start calling me a Google fan boy or something (well, too late). It seems everything they come up with lately grabs my attention. And I guess it makes sense, because they seem to be heading in a different direction than a lot of people, and more in a direction that appeals to me. First Android (open mobile handset), then Google Chrome (Webkit-based browser), then the V8 C++ friendly JavaScript VM, and now, Native Client.

If you haven't heard of it, it appears to be a Google research project into running secured native x86 code in a browser. Yes, we have tried that before with ActiveX and it was a security disaster. But the underlying need for high performance interactive web pages is pretty intriguing. If you could write browser applets in C++, why wouldn't you? I suppose...

I had to try it myself. The install instructions are for Firefox, but I dumped Firefox for Chrome a while ago. It's good that Chrome has some Firefox in it, because all I had to do was copy the plugins for Firefox into my Chrome Plugins directory (it's hidden in Local Settings, Application Data, Google, Chrome, Application, Plugins).

I was then able to go through their little demos and tests. They're cute and the Mandlebrot demo shows some of the power. There's also a demo of the open source SDL version of id's Quake. It's pretty complicated to build and I couldn't get it working on my Windows box (mainly because I'm Cygwin-free and it seems to need it). But it's an interesting idea, taking an SDL-based application and converting it to run in a browser (Native Client uses SDL to do audio and video). Maybe, they'll even expose OpenGL through SDL to the native code as well. That would be more interesting.

One thing though that burst my bubble with this whole experience were the results of the performance tests that they have. The C++ version of the tests were only marginally better than the JavaScript ones. I think that's thanks to the great job they've done with the V8 VM. If that's the case, I really wonder whether this stuff actually makes sense, other than porting old software rendered games to your browser, I guess. I need to stew on that one a little before buying into this idea.

A busy day for Khronos

My Khronos.org News feed filled up all of a sudden today. Looks like they've been busy and had a couple of announcements to make.

They released a new version of the 2D OpenVG spec. They added some APIs for text glyphing to make it easier to draw good looking text. I'm not sure anyone really uses OpenVG, especially when you are most likely to be drawing 2D in a web browser with Adobe Flash or SVG (and even then, most likely Flash). From the news release, this is probably most interesting to the mobile crowd.

The more interesting announcement for me was the release of the first OpenCL spec. OpenCL is a standard for running general algorithms on the newer GPUs in video cards. It'll also be ported to other multi-core systems like Cell and DSPs, but most likely you'll be using it with a video card. Of course AMD and nVidia were quick to announce their support for this spec, which gives it some immediate momentum.

OpenCL specifies a C-based language for parallel processing as well as APIs that drive them. Up until now, nVidia and AMD had proprietary solutions that didn't work cross platform. OpenCL opens the door to make parellel programming available to more and more programmers and I'm dieing to see what they'll do with it...

Wascana 1.0 in Alpha Testing

Well, that didn't take very long. I've spent a few hours building my special p2 artifact repository that manages installed files, including extracting them from an archive and deleting at uninstall time, along with it's associated p2 touchpoint that hooks it all up. It's not a lot of code and you can see it in CDT's CVS space (repo: /cvsroot/tools, module: org.eclipse.cdt/p2).

I've also created a generator that creates p2 repositories that use that touchpoint to install remote artifacts from various locations, mostly on SourceForge. Currently I only have support for the MinGW toolchain and the MSYS shell environment. I'll add libraries as I build them with the 4.2.3 compiler I'm using here. I'll start with SDL and also do wxWidgets and boost. We can always add more later.

It's working very well. Managed build picks up the mingw toolchain and uses it when you select the MinGW toolchain. MSYS doesn't work yet for Makefile projects but managed is usable now. And here's how:

1. Unzip the Eclipse IDE for C/C++ Developers anywhere you'd like on your machine. You can also start with any other Eclipse install as long as you have the CDT installed.


2. In Software Updates, expand out the tools/cdt/releases/ganymede site into CDT Optional Features and install the Eclipse CDT p2 Toolchain Installer feature. Allow Eclipse to restart to make sure things are initialized (I'm not sure if you really have to do this, I'm just paranoid).


3. Go back to Software Updates and add the Wascana repo site at http://wascana.sourceforge.net/repo. Install everything under the MinGW Toolchain category. This time you don't need to restart. You don't even need to apply changes.

Once you're done, you can go to the directory containing eclipse.exe and you'll see the mingw and msys directories there, ready to go. Well at least the mingw dir is, I still need to set up msys correctly to find the mingw compilers, but it is only an alpha :).

Feel free to give it a try and let me know what you think. I'm pretty excited with how this is going. While creating this, a new version of the win32 API component came out and I added it to the repo and the Update... feature found and installed it. Very cool!

It's a very interesting path where this is going. The ability to incrementally add in libraries and update new versions of the components will be a great showcase on how p2 can manage more than just bundles. Not to mention help me build one heck of a Windows development environment based on the CDT and open source tools and libraries.

Linux Kernel Debugging with CDT

Just ran into this awesome tutorial on how to use the CDT for debugging the Linux kernel using qemu's gdb remote debug service that makes it work much like a standard hardware/JTAG debugger.

This was something I played with a while ago when I looked at adding hardware debugging support to the CDT as an optional service. And I believe Elena from QNX has continued on with that work and we should hopefully see it completed for Galileo (if not before that).

But it further solidifies for me how important qemu is as a tool in the belt of the embedded software developer. We've seen it as a key enabler for Android without which I'm not sure it would have achieved the momentum it has. I think there are still issues with it, and of course one I'm looking at is ease at adding new hardware emulation and 3D graphics support. But I think there is plenty of opportunity there and being an open source project, the door is open to help make that happen.

The Future of Wascana

For those that don't know, I've been working on the side on a complete open source IDE distribution for Windows called Wascana Desktop Developer. It includes the CDT and the MinGW tool chain and a handful of libraries that enable cross platform development. I did the original "beta" release over a year ago and have over 12,000 downloads to date. But it's getting long in the tooth and I really need to respin with Ganymede Eclipse/CDT and gcc 4.x.

The question I'm dealing with now is what Wascana should look like going forward. My Wind River team and I are just wrapping up a p2-based installer for our Wind River products that are similar to Wascana but on a much bigger scale and targeting our Wind River platforms. We've learned a lot about how to extend p2 to manage the install, update, and removal of archived binary files into an install tree.

I want to bring that similar experience to Wascana and have started working on an open source version of these extensions. I'm starting doing it as part of the CDT since I need to support CDT 5.0.x with it and want to release around Christmas time. Once I check it in, the p2 team can look and see if the want something like this and give feedback on changes that would be needed to get it into an upcoming platform release.

In the end, Wascana will mainly be a p2 repository that ensures you have all the plug-ins installed to get a working CDT for MinGW, and that will allow you to download and install the MinGW tool chain and libraries, either from their home locations, or from the Wascana SourceForge download area if I need to rebuild for whatever reason. Updates and new components would be done by adding them to the repository.

So the question becomes, do I need an old time installer for this, or would the community be happy simply downloading the Eclipse C/C++ IDE package and working with the Software Updates tool to get everything they need. I have a feeling people will still be looking for that single setup.exe download to set everything up. Then I need to ask whether laying down the bits is sufficient, or whether I need to do a p2 director thing.

The good news is that I sense MinGW is maturing. Despite having an unmanaged release cycle (and I do have a second source for the mingw gcc tool chain thank goodness), it looks like it's ready for prime time, at least for my little distro. Enough so, I'm giving up on Windows debug support. My focus is cross platform, and my time is limited and building a pure Windows debugger is hard and without a significant contribution it won't happen, so I'm not counting on it. Wascana will do just fine without it.

Javascript and C++, eh?

I can't get my mind off of Dave Thomas's keynote at Eclipse Summit Europe. His words made so many things crystallize in my mind. I've stated many times before in this blog and in my day job, I hate Java. It's an incredible irony that I do my day to day coding in Java to support developers who focus so much on efficiency and performance and use C mainly to accomplish that with a sprinkling of C++ for good measure. And then to hear their constant complaints that Eclipse is too slow. My good friends in Java VM land tell me not to blame Java for that, but you know, it's so tempting.

Dave mentioned that applications should be written in C++ and JavaScript. I dunno. C++ has it's difficulties, there is no doubt. It's hard to write good C++ programs. That's why the mix with JavaScript made me think. Does it make sense to build an application where your hard core performance focused code and code that interfaces with the underlying system is written in C++, but all the code that manages interactions with the user is done in JavaScript?

I've started to take a look at Google's V8 JavaScript engine. As they say in their videos, they're built for embedding in C++ applications and they have implemented some interesting tricks to get JavaScript to run fast, such as a JIT compiler and some heuristics to make class property access faster. As well, they have an efficient memory management system which includes being able to persist snapshots of the heap, including the JITed code, out to the file system for faster startup.

That got me thinking of Eclipse, of course, or really IDE's in general. What if you take a cross platform GUI toolkit like wxWidgets, add in a component model to allow for dynamic extensions, plus rewrite the CDT parsers in C++ for speed, plus ..., and throw in a JavaScript engine like V8 to make it easy for users to program, wouldn't that make for an interesting architecture? But we already have Eclipse so why would we do that all again? Just a question...

An Interesting Ottawa Demo Camp

The Ottawa Eclipse Demo camp was tonight and I thought I'd write about it before I went to bed. The demos were quite interesting, a different mix than before which keeps it fresh. And the hospitality of the Foundation staff was awesome again.

I was especially intrigued by Nick Edgar's embedded web UI demo that he's working on as part of Jazz. This is something I thought of doing for my talk at ESE. Present information in a web page using Eclipse's embedded browser. And then have JavaScript on that page interact with the surrounding Eclipse environment. The workflow he showed was very clean and I think there are some pretty cool things we can do with this. The technique he used was quite a kludge and even he admits it (communicating through the status bar?) But the SWT guys are thinking of better ways and I can't wait to try this myself.

The other interesting demo was from the Zeligsoft gang. I worked with some of the fellows that started Zeligsoft. We were part of the Rose RealTime development team. It was interesting to see the product they've come up with and the simularities it has with the stuff we did back then. They're betting the farm on model driven development. I can't say whether they'll succeed or not, but they've done a few things better, but a lot is the same.

I also have to thank Boris and Eric for their demos on e4 and the model-based UI in particular. I have a better sense of what they are trying to accomplish. Whether it's better or not than what we have today, I'm not sold yet. But I'll have to give it some hands on before making a final judgement.

I also got some interesting feedback on my article on IBM and Eclipse. (BTW, it's not whether we can survive, it's that we better plan and make sure we can, which I think we're finally doing). There were a lot of IBMers at the Demo Camp which was good to see. And there were as many ex-IBMers there too. I think it's pretty healthy. The Eclipse expertise is spreading throughout our small and tight knit town and Ottawa has a great concentration of Eclipse expertise, which makes it a great place to be.