Back In 2006 I was active for nearly two years in the Dutch Ubuntu LoCoTeam as site/forum administrator and release party organiser. Due to time constraints (among many other factors) my participation as LoCoTeam member steadily declined. But yesterday I became moderator of the official Dutch Ubuntu forums again. So far the community response (by e-mail) has been wonderful. It feels kind of good to know that a lot of people thought you did a great job back then and that they are happy with your return

I hope I can make enough time to actively help out the LoCoTeam – because there’s a lot of other stuff to do as well – and I can only do my best. The reason for my renewed participation is that I want to make a difference instead of complaining about some of the problems I was seeing. Of course some problems are to be expected: Ubuntu-NL has grown considerably. When I think back of my first days as forum administrator (about 2 years ago) I could easily read all posts myself. There were about 600 registered forum users in total, but today there are over 17.000!

Oh well. I hope I can make a difference.

About function interposition

I was wondering how I could override dynamic library calls in Linux, and I came across this technique known as function interposition. It is a powerful technique that allows you to override dynamic library calls. It might sound dull, but it can be very, very useful. There are some memory trace tools that make use of this technique to work, but perhaps a cooler example is the OpenGL capture system which was created by nullkey: it can capture OpenGL frames by overriding certain OpenGL functions. Another example are cheat tools (wallhacks, aimbots) which also make use of this technique a lot.

Some background

While Googling (did I spell that right?) I came across this recent blog article which explains the background very well. I will quote it here:

First, some background. When a program that uses dynamic libraries is compiled, a list of undefined symbols is included in the binary, along with a list of libraries the program is linked with. There is no correspondence between the symbols and the libraries; the two lists just tell the loader which libraries to load and which symbols need to be resolved. At runtime, each symbol is resolved using the first library that provides it. This means that if we can get a library containing our wrapper functions to load before other libraries, the undefined symbols in the program will be resolved to our wrappers instead of the real functions.

So if we create a custom shared library which overrides some of the functions of the original library, our functions will be called instead of those of the original library.

How to do it

• Write new functions which override existing functions
• Compile the written code to a dynamic library that is linked to the dynamic linking interface library
• Use the LD_PRELOAD environment variable when running an application to preload your custom library before all other dynamic libraries

The article by Jay Conrod has an example which shows you the basic implementation of a simple memory allocation tracer.

I have also cooked up an example myself. Because I’ve been busy learning more about OpenGL, I thought that it shouldn’t be too hard to create a wallhack for one of my favourite games: Soldier of Fortune 2 running in Wine. Just for testing purposes of course! I am no cheater It turned out to be relatively simple, although my first tries weren’t so very successful:

• Epic fail – At least I know my library is used now.
• Partial success – Disabling depth testing completely was only a partial success.
• Success – A debugger can tell that players are drawn using glDrawElements(). By knowing the number of elements for each character, we can disable depth testing selectively.

Soldier of Fortune 2 wallhack example

For those of you who are interested in the code:

Interesting thought about multiplayer cheats and Wine

If anti-cheat tools would perform a sanity check of the OpenGL or DirectX DLL, they would only find the virtual DLL’s when a game is run in Wine, right? I’m wondering whether this sort of cheats can be made undetectable then. In a way I hope not, because cheaters are very annoying when you’re playing a game, but on the other hand, it would be an amazing technological achievement. Anyway, anti-cheat tools like PunkBuster don’t even work with Wine at the moment, so it might be a non-issue. What are your thoughts?

I started with a CLFS (Cross-compiled Linux From Scratch) installation yesterday. After major issues which I encountered when trying to boot the new kernel, I managed to create a very minimalistic kernel configuration which built a kernel that was able to boot up! I now have a very minimal Linux installation.

It’s not done yet, though. I will now have to install base system software, build a more advanced kernel, and then start building and integrating different software components of my personal choice. After this project I might either start my own Linux distribution (either from scratch, or based on Arch Linux) or just provide a repository for Arch Linux which contains some special software and help with Arch Linux development – because the goals of the Arch Linux distribution are very similar to mine. I also want to provide a repository with custom kernels and software which improve performance for all laptops that are sold by Hasselt University since 2008. I would however like to improve some things upstream as well, for example add a graphical installer and create an auto-system-configure tool which sets up Arch Linux for specific usage, and I’m not entirely sure yet whether these would be welcome changes upstream. If not, I might have to resort to an Arch Linux fork or my own distribution from scratch.

The last few days I’ve been looking at x86 Linux assembler programming. I think it’s important to know how the code which you write in higher-level languages gets translated to low-level instructions and in what way things like libc or the kernel are involved. With some knowledge of assembler you can not only directly optimize some slower operations by implementing them in low-level assembly and by utilizing specific processor features, but in general you will also learn in which ways you can optimize your software by just writing the high-level code a little bit different – by anticipating how your code will translate into low-level assembler instructions.

That was my motivation for trying to do a few things with the Netwide Assembler (NASM). The result is a simple demo which utilizes libc, GLUT and OpenGL to display a rotating pyramid and cube.

compiled ogl.asm application running in Ubuntu 8.04

The resulting code can be viewed here:

I have also written two other smaller examples:
- marketing.asm is a really simple statistics application which calculates the ideal sample size given a few parameters
- alsawav.asm is a simple audio library which utilizes ALSA to play an unbuffered raw WAV/PCM sample

Please note that all these examples require an x86 Linux distribution. These examples will not work on any other operating system.

Hello everyone! It’s been a very long time since my last blogpost, and I haven’t updated my blogsoftware in a long time. Apparently, some spammers noticed this as well and were able to add some hidden spam links in some of my pages. I should seriously update my WordPress one of these days…

Anyway. Let’s talk a bit about Akademy this year. It was my first Akademy and it was a great experience! I met a lot of interesting and fun people! Everyone seems to agree that this Akademy was the best one yet, so I’m very glad to have been a part of that, although honestly most work was done by Bart Cerneels and Wendy Van Craen.

The first day

The first day was the hardest day for the organization, or at least for me and Pieter Vande Wyngaerde. After helping out at the Akademy location (Campus De Nayer) we had to make sure everyone had a place to sleep at Roo?[sz]enda[ae]l (the name was spelled different on every sign!). The rooms list wasn’t all that clear, because the numbering was different from the room numbers, but after a while we were able to figure it out, and we were able to give everyone a place to sleep.

Resting out on the floor after all that hard work

Party time!

On the second day I was just walking around the campus, helping the team with small things. I wasn’t able to see many talks, but I was able to at least see the Nokia keynote from Sebastian Nyström. A very interesting talk, although I’m still not sure exactly what direction Qt Software will be evolving towards with Nokia, but up until now it’s been going well, so that makes me happy enough for now.

In the evening it was party time: there was this social event at Het Anker in Mechelen, which meant free food and free Belgian beer! The “Gouden Carolus” beer is something even I – as a Belgian – had never drunk before. Personally, I think there are much better tasting Belgian beers, but maybe that’s just my personal taste. The beer was pretty strong though. But I’m sure that most of you noticed that

After the social event I went for a small evening stroll with some of the guys who were staying at Zandpoortvest. In the end, we walked all the way to the Zandpoortvest hostel, and after I had a look around the hostel, I had to walk all the way back to Rozendaal (I’ll just stick with this way of spelling it..) Unfortunately for me, I got lost. Too bad I didn’t have the N810 with its GPS (yet), otherwise I could’ve gotten back at Rozendaal a lot quicker. My cellphone batteries were dead as well – doesn’t this sound like some horror story? – so I just figured I’d walk and follow the signs pointing to Sint-Katelijne-Waver.

After a while I saw some arrows pointing to the train station, and from there, I was able to find my way back to Rozendaal. By then, it was around 5 am, and I had a big blister on my left foot from walking all night.

…I tried to stay up, but around 11 AM I felt really tired and I wanted to sleep a bit. And because of that I missed all the great presentations that I wanted to see! It was in the late afternoon somewhere that Pieter Verledens woke me up. Then I realized how late it already was and that I missed Zack Rusin’s Gallium3D talk, which I was looking forward to. Luckily, the talk was recorded and is available on blip.tv now (thanks to Bart Cerneels).

Nokia converts KDE developers to GNOME

The Mobile and Embedded day was one of the best days of Akademy. Not only because Nokia gave away a lot of Nokia N810 devices to more than 100 KDE developers, but also because the talks were really interesting. The big problem however is that it was a pretty busy day and I wasn’t able to see most of the talks (I really wanted to see the maemo, QEdje and OpenMoko talks as well). I ended up seeing only the OpenGL ES for Embedded Linux talk by Tom Cooksey, which was very interesting though. At the end of the talk he told us about the OpenPandora handheld device that will be released soon. Apparently it is a great device for OpenGL ES development. I found their website here: www.openpandora.org

Now, the title above this paragraph is inspired by Vincent Untz‘ blogpost. It refers to the free N810s that Nokia gave away. The devices run the Maemo OS, which is based on GNOME. And to be honest, I like Maemo a lot. It works very well. But I’m very interested in the progress that Marijn Kruisselbrink is making with porting KDE 4 to the Nokia internet tablets. Me myself, I haven’t done much with the device yet, aside from playing some games, listening to some music, and doing video calls with other KDE people I did however manage to get some Qt applications ported to the device, but that was ridiculously easy. Hildon integration Just Works™ out of the box and there’s very little that has to be changed to make your application look great on the N810. Except for graphics operations which are a bit slow at the moment, as explained by Ariya Hidayat in his blog post, so I also hope that the performance problem gets fixed in Qt 4.5.

The Nokia N810 Internet Tablet.

Oh. And my cat was chewing on my N810 adapter (even cats seem to like this gift from Nokia), so the pin doesn’t fit my N810 perfectly anymore. But it still works if I put it in carefully and don’t touch it, so I’m still good…

Boat and Barbecue

Thursday was a really nice day. A nice boat trip, and a very tasteful barbecue. I enjoyed it a lot. Jonathan Riddell was busy interviewing a lot of people, you can download the interviews from this link. I didn’t volunteer for an interview, but you can see me sitting in the back during interview number 44. That’s enough for me

At the end of Akademy, Aaron and Chani treated us (the Akademy organisation team) to a nice dinner. Thanks for the good meal and the great stories.

Everyone had to leave back home on Friday or Saturday. Sad times. I’m planning to come to Akademy next year as well. It was a lot of fun this year.

To wrap it up, here are some random notes:

• I’m not sure who’s the noisiest: the Amarok Wolves or Team Humongous
• If you are in dire need of some music, maybe you can hire the A-Team!
• Video’s of a lot of talks are available at http://stecchino.blip.tv/
• The Emsys guys really like Mega Mindy …

Sinds vorige week ben ik begonnen met mee te helpen aan de ontwikkeling van het open source DJ programma mixxx. Mixxx is een stabiel programma waarmee men live muziek kan mixen. Het heeft een aantal zeer interessante features, zoals bijvoorbeeld automatische ritmedetectie en ondersteuning voor een heleboel hardware.

Waar staan we vandaag? 1.6.0

Ik werk mee aan de nieuwe versie waarvan twee weken geleden een eerste bètaversie werd gelanceerd. Voor de nieuwe 1.6.0 versie zijn een heleboel nieuwe features gepland. Ik som even de belangrijkste veranderingen op:

• Scratchen via timecoded vinyl-platen [ link naar flash video 1, video 2 ]
• Kleurenschema’s voor skins
• Nieuwe muziekbibliotheek (muziekbrowser)
• Verbeterde ritmedetectie
• Verbeterde ondersteuning voor MIDI-controllers (hardware)
• HQ-equalizer toegevoegd
• Audio core herschreven/vernieuwd
• Ondersteuning voor LADSPA geluidseffecten
• Live broadcasten over internet (Icecast, Shoutcast)
• Rechtstreeks opnemen naar MP3, Ogg Vorbis, Wav, Flac

Ik werk momenteel aan de laatste twee punten. Het is niet zeker of het helemaal af zal raken voor de Hardy freeze in februari (het moment waarop programma’s naar Ubuntu 8.04 Hardy Heron geupload worden en niet meer aangepast mogen worden) maar Ogg Vorbis Icecast/Shoutcast-ondersteuning is zo goed als af, dus dat zal er waarschijnlijk zeker inzitten.

Zelf kijk ik uit naar de ondersteuning voor LADSPA geluidseffecten, maar ik vermoed dat dat niet af zal raken voor de 1.6.0-versie. Dan maar wat langer wachten…

Evolutie van de broadcasting code

Zoals eerder gezegd ben ik momenteel bezig met het implementeren van Icecast/Shoutcast ondersteuning zodat we een mix rechtstreeks over internet kunnen broadcasten. Aanvankelijk dachten we dat het simpel zou zijn om dit systeem te implementeren (gewoon libshout gebruiken) maar al snel bleek dat we eerst nog een encoder moesten schrijven. Ik ben de laatste dagen dus vooral bezig geweest met het schrijven van een Ogg Vorbis-encoder met behulp van libvorbis, libogg en libvorbisenc.

Afgezien van het feit dat er voor libvorbis geen technische API-documentatie beschikbaar is, is het ons uiteindelijk toch gelukt om de encoder werkend te krijgen. Alles ging goed, maar de audio latency was nu wel verhoogd omdat de ‘audio callback thread’ voor een korte periode geblokkeerd werd wanneer de encoder zijn werk deed. Toen we ook nog beseften dat we om metadata te updaten een nieuwe stream moesten initialiseren was het onvermijdelijk om een nieuw systeem te ontwerpen om de encoder parallel in een aparte thread te laten draaien (multithreaded) met een eigen buffer.

Albert ging hiermee aan de slag en na twee of drie dagen knoeien presenteerde hij een nieuwe ‘engine’ (de SideChain-engine) aan ons die de audio buffert voor andere engines (zoals de broadcast engine) en deze engines in een aparte thread draait. Het resultaat is lage audio latency en geen enkel performanceprobleem meer. Ook werd het totale CPU-verbruik verlaagd omdat de encoder nu iets minder vaak aangeroepen wordt.

Om het schematisch voor te stellen:

voor: [ afbeelding: Oorspronkelijke Broadcast Implementatie ]
na: [ afbeelding: Threaded Broadcast Implementatie ]

Wat moet er nog gedaan worden? De SideChain-engine moet nog een klein beetje aangepast worden, er moet nog een MP3-encoder worden geschreven, het instellingenvenster moet afgemaakt worden, en tenslotte moet de code wat opgeschoond worden, en moet alles grondig getest worden!

Het laatste jaar zijn er twee projecten ontstaan die het mogelijk maken om OpenGL en Alsa op te nemen zonder al te veel verlies van performance. Voor de gamers die niet weten waarover ik het heb: dit is bij wijze van spreken ‘FRAPS for Linux’. Ik heb beide projecten uitgetest en ze werken allebei prachtig: yukon en glc.

Klik hier voor de volledige video-opname op YouTube

Dit is Aquaria Demo in Wine
opgenomen met glc op 1024×768
met Compiz ingeschakeld.

Je bent vrij om te gebruiken wat voor jou het beste werkt:

• yukon
• glc

Persoonlijk ben ik momenteel voorstander van glc omdat de ontwikkelaar zéér snel is met het reageren op bugreports en omdat glc is geoptimaliseerd voor multi-core systemen. Ook heeft glc enkele speciale mogelijkheden zoals on-the-fly color conversion van BGR naar YV12 (geeft grotere compressie) en je kan zowel QuickLZ als LZO gebruiken om de stream on-the-fly te comprimeren. Een andere speciale feature van glc is de mogelijkheid om meerdere OpenGL of Alsa streams tegelijkertijd op te nemen.

glc werkte perfect bij mij, maar ik had een klein probleem met het opnemen van sommige games in Wine (problemen met audio en af en toe een crash). Na een kort gesprek met de ontwikkelaar (nullkey) en na het opsturen van enkele gdb traces werd het probleem binnen 10 minuten volledig opgelost door de ontwikkelaar. Prachtig, toch? De nieuwste versie van glc is momenteel 0.4.4 en bevat de Wine patches.

Ik ben bezig aan een patch voor SMPlayer voor mezelf:

1. ik wil de huidige tijd kunnen zien zonder full-screen te verlaten (en zonder een klok te kopen) – zoals je kan zien op de bovenstaande screenshot heb ik dat reeds voor elkaar gekregen.
2. ik wil de afspeeltijd van de film als een klein balkje op het scherm zien, niet via het tekstuele OSD, en ik wil eveneens de film kunnen doorspoelen door op een willekeurige plaats op mijn mooie alpha-transparante balk te klikken… want dat is veel mooier dan zo’n uitschuivend balkje dat je video naar boven duwt

Technische details:

• mplayer gepatched en gecompileerd met vf_overlay patch
• QPainter (van Qt 4.3) tekent de klok en tekst
• getekende bitmap wordt vertaald naar het benodigde formaat en naar een gedeelde geheugenplek geschreven
• mplayer (vf_overlay) leest de bitmap uit en legt hem over de video heen
• mediaspeler waar de patch in wordt ontwikkeld is SMPlayer (mplayer frontend)

Ik denk niet dat ik deze patch upstream zal opsturen, tenzij er veel vraag naar is. De reden daarvoor is simpelweg dat je een onofficiële mplayer patch nodig hebt (vf_overlay patch) om van deze patch gebruik te maken, en (ik maak een gok) 95% van de smplayer-gebruikers zijn dus sowieso niets met deze patch.

Ik ben wel van plan om de patch op het forum van SMPlayer te plaatsen voor degenen die het eens willen bekijken of zelf willen proberen, maar de patch is momenteel nog niet in een “releasebare” status.