Le fourre-tout à Geo

I was at FOSDEM 2009 in Brussels, and that was fun! I enjoyed seeing so much geeks, with their distrib t-shirts and trolling each other. That's nice to see that all these open source geeks are not only nicknames on IRC :)

I saw some interesting conferences, projects, people, and now I have enough ideas to play with for a few months. Here are some thoughts about the conferences I attended:

Reverse engineering of proprietary protocols, tools and techniques, by Rob Savoye

That was something people were waiting for! The room was full! The speaker was the typical GNU guy: a geek with a great beard. He talked about his work on Gnash, the GNU flash implementation, and how he reverse engineered the RTMP protocol, a proprietary streaming protocol from Adobe. He explained the requirements and the process of reverse engineering in a few points:

• First, talk to your lawyers: you may not know about some obscure cases and laws, and you can be sued over whatever little thing you might have forgotten. Using the "clean room" method is recommended: one group writes specifications while spying on the protocol, and the second writes the software using the specifications.
• Next, have a calm, non disturbing environment. Reverse engineering a protocol is a long and hard process, don't expect to do it in one month on your spare time. And be prepared to throw away a lot of code. You may go into a direction, and then realize that the parser you just wrote is completely wrong. Using test cases is a good way to see if your code is wrong. Raw dumps will be useful for the tests.
• Rob presented some tools he used to work with all the hexadecimal dumps he used, like od, ngrep, tcpdump, hexdump, wireshark, strings, vim, emacs, etc. You have to have a good hexadecimal editor: humans are really good at finding text, patterns, and you can save a lot of time just by looking at the packet dumped.

Marionett, networking for dummies, by Marco Stronati

This was really interesting, but too short. I would have liked to see some demos. The Marionnet project was started by a professor from université Paris 13, because teaching networks was hard: setting up a network for a course takes time, sometimes the hardware fails, there's not enough of it, students can't try at home... So, this professor started Marionnet, to provide students with a network testing tool, based on virtual machines. With this, they're not limited by the cost of hardware, the number of routers, and students can work at home. It's even easier with the live DVD they built. Marionnet supports the classic protocols: TCP, ICMP, DNS, BGP, etc. And it has some nice features: you can create "clouds", network items you can't determine, you just know some packets come through it. You can add lag and packet loss on a cable. You can make your virtual network interact with the real network your computer is plugged to, and even interact with another Marionnet virtual network over real networks. I hope we'll use it in Rézoléo to teach networking to new people.

Warning, troll inside: it's a french project, made by some academic people, so it's in OCaml (and they seem very proud of it). French CS professor like things that nobody uses, like Merise or ADA... *sifflote innocemment en regardant le ciel d'un air vague*

Wt, a C++ web toolkit, for reach web interfaces to embedded systems, by Pieter Libin

One of the most interesting sessions of FOSDEM. Wt is a C++ library providing web services. It can serve a website either using the builtin web server or through FastCGI (Apache, lighttpd,...). With its small footprint (2-3 Mb), it is aimed at embedded devices. Adding an ethernet port and a little flash disc cost less than adding a full interface with keyboard and screen. Managing your industrial devices can be done from your desk, in your navigator! The API looks a lot like Qt, so, easy to read. And it can use nice things like Ajax or SVG.

I'm pretty sure everybody in the room was thinking about a web interface to their coffee machine :)

Apport - Automatic application crash reporting for OpenSUSE, by Jan Blunck

opensuse: apport (crashreporting) what do you send back to bug reporting? parsing the crash report? symbols? having a stacktrace using /proc

I went to this lecture just for fun: I wrote a system like this in VLC for Windows just two weeks ago (see http://git.videolan.org/?p=vlc.git;a=commitdiff;h=f7c9acaebd4b620d8b4ed6aedbdae3a3ee52d38e and http://git.videolan.org/?p=vlc.git;a=commitdiff;h=82bc4a310683688843c8b361a1c767260f522ecc, more on this in an upcoming article). I was able to resist the urge to troll the speaker, but that was hard :)

While writing your crash reporting application, you encounter these classical and annoying problems:

• what do you send back? Can dumping memory harm the privacy of your users? Can you send the OS version? The good solution is asking the user: some won't like sending the content of their memory. And some users will be pleased to send you a message describing what they were doing (and insulting you on the way...).
• How do you get a stacktrace? How can you get symbols? Well, I've got a solution for the symbols (see my commits and the strp man page)
• how do you use the crash report? I was disappointed to see that they had no working (read "in production") system to parse crash reports. As I still have no system to do it (*sifflote*), I was expecting that the Novell guys would have something to keep me lazy... To parse a report, you typically have to find similarities with other reports. A good way is to assemble crashes with similar stacktraces.

Google Summer of Code: a behind the scenes look at large scale community management, by Leslie Hawthorn

Leslie, the Summer of Code's community manager, talked about all the hard work she did interacting with all the students, mentors, projects, Google people, and about what you can do to help a large scale project like this to go smoothly. Kudos to her!

And so on...

I saw many other nice things, like FreeIPA (an identity management system using LDAP, and Kerberos. I'll look more into this in the next few weeks), CMake and Qt (well, thanks to VLC, I knew a bit about these projetcs, but attending a small lecture helps understanding), syslinux (PXE with HTTP, yay!), cpan2port, and I came to the CACert assuring party. I now have 100 points, I can try to pass the assurer's test! I didn't do the GPG signing party: I was too late to register my key on the keyserver they built for the event, and, well, they did it outside, and it was cold :P

J'ai eu pendant quelques temps un petit problème avec mon PC : le lecteur de cartes SD intégré était reconnu et installé, mais pour les cartes SD, elles étaient reconnues, mais Windows ne trouvait pas le driver. Comme j'ai des partiels à réviser cette semaine, je me suis attelé à ce problème aujourd'hui :)

J'ai cherché un peu sur Internet, et sur pas mal de résultats, c'était soit "installe cette mise à jour Windows", soit "désinstalle le lecteur et redémarre". Sauf que les mises à jour ne s'installaient pas (parce qu'apparemment elles ne sont pas faites pour mon PC, je trouve que Windows Update fait de la méchante discrimination ;), et que désinstaller puis redémarrer a eu pour effet de transférer le problème au lecteur, qui ne trouvait plus son driver du coup...

En cherchant un peu plus, je suis tombé sur ce lien (bon, ok, le 4ème résultat quand on tape "vista sd" dans google.fr contient ce lien, c'est juste que j'ai eu la flemme de regarder avant un moment).

La solution est très simple :

• Allez dans C:\Windows\System32\DriverStore\FileRepository
• Cherchez les dossiers dont le nom commence par sdbus.inf
• Ouvrez le dossier sdbus.inf** le plus récent
• Copiez le fichier sdbus.inf dans le dossier C:\Windows\Inf
• Essayez de réinstaller votre lecteur de cartes, normalement, ça devrait marcher :)
• Si vous avez de la chance, les cartes SD seront reconnues et installées sans problème, mais comme moi j'ai pas de chance (Calimero, tout ça), j'ai du bidouiller un peu plus
• En gros, faites la même manip, mais cette fois en cherchant le dossier le plus récent commençant par sffdisk.inf, puis en copiant le fichier sffdisk.inf dans C:\Windows\Inf
• Essayez d'insérer votre carte SD, pour voir si elle est reconnue et installée
• Si ça marche et que vous êtes moi, retournez bosser votre mécanique des fluides (parce que les tuyères et les ondes de choc c'est très important, sisi!)