Profile

Cover photo
Thomas Petazzoni
Works at Free Electrons
Attended Université de Technologie de Belfort Montbéliard
804 followers|363,738 views
AboutPostsPhotosVideos

Stream

Thomas Petazzoni

Shared publicly  - 
 
Just booked flight tickets and accomodation for the Embedded Linux Conference Europe! Arriving October 2nd in Dublin, to attend the +Buildroot developers meeting on October 3/4 before the conference itself on October 5/6/7. Since conference hotels are prohibitively expensive, we booked a big and fancy AirBnB flat in Dublin for the team of 6 +Free Electrons engineers! :-)

For ELCE itself, my Buildroot tutorial proposal got accepted, and I've also been asked to give a keynote, so I'll be talking about "Linux Kernel SoC Mainlining Tips (By a Bunch of Other French People)". I must say I'm a bit scared by the idea of giving a keynote, but we'll see...
23
1
Cedric BAIL's profile photoKarim Yaghmour's profile photoKevin Hilman's profile photoAnthony Harivel's profile photo
5 comments
 
You'll do great on the keynote, your talks are always well done.
Add a comment...

Thomas Petazzoni

Shared publicly  - 
 
So, taxis are on strike in France protesting against Uber, and physically attacking Uber drivers.

Do the taxi drivers realize why Uber is so popular ? Not because of the price, but also because the service offered by Uber to the consumers is so much better than what taxis are offering. I'm relatively regularly using taxis in France for work related trips, and I've used Uber once. So one could say that my Uber experience is limited, but here is anyway a quick comparison:

  - Taxis almost never accept credit cards. We're in 2015, and they still haven't discovered credit cards. With Uber, you don't even need you credit card with you, you're directly charged on your credit card, which is connected to your Uber account.

 - Taxi drivers are in the majority of the cases horribly impolite and unpleasant persons. Uber drivers, because they are rated by users, have no other but being minimally nice. The one Uber driver that drove me was way nicer than most taxi drivers you can catch from any french airport or train station.

 - Taxi drivers drive like crazy, sometimes to the point where you feel in danger. Like driving at 130 km/h when it's limited to 70 km/h. Uber drivers are rated, so if they drive like crazy, they'll get bad rates, and they don't want that. The Uber driver that drove me complied with all speed limits, and had a very safe driving attitude. I felt actually much more in security with this "amateur" UberPOP driver than with many supposedly "professional" taxi drivers.

 - With the Uber application, you know where your driver is. You know when he will be arriving. And you know how your driver is rated before choosing. Where are these features for regular taxis?

 - With Uber, you know before ordering the car what will be the cost of your trip. And the Uber application is telling the driver what directions they should take. Not like regular taxis which will regularly use crazy directions, and where the same trip will cost you highly different rates from one time to the other.

So, regular taxi drivers, you feel in danger?

Yes you are.

But not because the competition of Uber is unfair. Simply because the service you offer is completely crappy and outdated compared to what Uber is offering. Instead of complaining, just take a step back, and think about how you should modernize your profession and behavior, and learn from the good things that Uber is bringing to consumers. If you don't take this step back, you will loose, no matter how much you yell after Uber.
35
6
david chollez (dcz switcher)'s profile photoMélodie JM's profile photoJoyce Markoll's profile photoDavid Siorpaes's profile photo
15 comments
 
Hopefully even more people will use UberPop, more drivers will join UberPop and Taxis will die. What's the point in asking 200'000Euros for a Taxi licence from any driver? Supposedly Taxis are safer, yet the Taxi drivers are behaving like hooligans in France?
Add a comment...

Thomas Petazzoni

Shared publicly  - 
 
Release early, release often. So I've just published a few minutes ago on the Buildroot mailing list at http://lists.busybox.net/pipermail/buildroot/2015-June/130833.html, my prototype for a Buildroot runtime test infrastructure.

I have no idea if I've made the right choices and if it's going to be the ultimate solution, but at least it's an attempt.

Basically, we in the +Buildroot community have been doing automated build testing of random package configurations at http://autobuild.buildroot.org since a long time now. However, this only tests the build (never the runtime execution), and also only tests the build of "normal" packages, not things like bootloader packages, the Linux kernel package, the filesystem images support, and other Buildroot features.

So, this Buildroot runtime test infrastructure allows to describe test cases that associate a Buildroot configuration with a Python function doing some tests on the build results and/or by running the generated system in Qemu.

The code is at https://github.com/tpetazzoni/buildroot-runtime-test. Anyone reading this will discover how bad I am when it comes to Python programming, but it does the job, and patches are welcome.

Let's see what the feedback on the Buildroot mailing list will be.
[Buildroot] Buildroot runtime test infrastructure prototype. Thomas Petazzoni thomas.petazzoni at free-electrons.com. Thu Jun 25 18:02:35 UTC 2015. Previous message: [Buildroot] [PATCH] ipset: bump to version 6.25.1; Messages sorted by: [ date ] [ thread ] [ subject ] [ author ] ...
8
Add a comment...

Thomas Petazzoni

Shared publicly  - 
 
 
Buildroot 2015.05 release, Free Electrons contributions inside

Details about the 2015.05 release of Buildroot: wf111 WiFi drivers, uClibc-ng, imx-usb-loader, infrastructure improvements and build fixes...

http://free-electrons.com/blog/buildroot-2015-05/
View original post
2
2
Frédéric COIFFIER's profile photoClaudio Omar Biale's profile photo
Add a comment...

Thomas Petazzoni

Shared publicly  - 
 
Sent a few patches to +Buildroot over the week-end, with some toolchain related updates:

 - Add support for gcc 5.1. I've done some quick testing, and a minimal system works fine on ARM/uClibc, ARM/musl, ARM/glibc, x86/glibc and x86/musl. However x86/uClibc doesn't work so far, ld.so segfaults when running init.

 - Make +Buildroot use uClibc-ng by default instead of the 3+ years old uClibc 0.9.33.2 version + gazillions of patches. Waldemar Brodkorb is very active in the maintenance of uClibc-ng, which is great compared to the inactivity of uClibc.

 - Switch to uClibc-ng 1.0.3, which was released just today.

Also, did a lot of updates to http://autobuild.buildroot.org, with many updated toolchain configurations, and more of them to cover addition architectures / C library / options combinations.
Date, Status, Commit ID, Submitter, Arch, Failure reason, Data. 2015-06-14 23:02:59, OK · d6d237c8, Peter Korsgaard (gcc10), mipsel, none, dir, end log, config, defconfig. 2015-06-14 23:02:13, OK · d6d237c8, Peter Korsgaard (gcc20), xtensa, none, dir, end log, config, defconfig ...
12
1
S. Lockwood-Childs's profile photopicoFlamingo Project's profile photo
 
It is uclibc-ng which has the segfault with gcc 5.1 on x86?
Add a comment...

Thomas Petazzoni

Shared publicly  - 
 
Want to work on embedded Linux projects? Submitting patches to the Linux kernel, writing Linux Board Support Packages for custom platforms, working on build systems, bootloaders, and more? Want to be part of a very open-source driven company, where contributing patches, writing open documentation, and attending/giving conferences are the key aspects?

We're hiring at +Free Electrons ! We are looking either for experienced or beginner engineers who are willing to come work with us in our offices in Toulouse or Orange (France).

If you were to join our Toulouse office for example, you would be working on a daily basis with +Maxime Ripard, +Boris BREZILLON, +Antoine Ténart and myself.
16
9
David Anders (prpplague)'s profile photoRémi Rérolle's profile photoDrew Fustini (pdp7)'s profile photo〈Moved〉 VIA Technologies 〈 See info 〉's profile photo
7 comments
 
+Lionel Orry The current opening is indeed in one of our offices (Toulouse and Orange), but in the future we will most likely have openings for remote work.
Add a comment...
Have him in circles
804 people
Marc Souviron's profile photo
Marcin Wojtas's profile photo
Matthew Locke's profile photo
Steven Nguyen's profile photo
vipin tiwari's profile photo
Patrick Vergain's profile photo
Josiane Bats's profile photo
Shawn Goff's profile photo
Graham Whaley's profile photo

Thomas Petazzoni

Shared publicly  - 
 
So, this is going to disclose my complete stupidity when it comes to HW stuff, but my G+ readers might be able to help :-)

Back a few years ago, I got an ARM Energy Probe at some ELC/ELCE event, which looks to be a nice device to measure power consumption. I so far never had the opportunity to use it, but it seems like the day has come for me to start using this device, which raises some questions.

It came with a cable that has a male DC jack on one side, a 0.02 Ohm shunt resistor in the middle, and a female DC jack on the other side. This is, as I understand it, meant to easily measure the full power consumption of a board, by simply connecting it as a man-in-the-middle between the board power supply and the board itself. Looks great for a HW dummy like me.

My boards are Marvell boards, they are powered by a 12V power supply rated at 5A. I have no idea if the board consumes the full 5A, but I'm pretty sure that the boards are pretty power-hungry when running fully loaded.

The question is: is the ARM Energy Probe and its 0.02 Ohm shunt resistor cable capable of handling such a power of up to 60W ? The ARM Energy Probe is rated to up to 30V, so this seems good, but I'm a bit more worried about the shunt resistor.

See the pictures for more details, and http://ds.arm.com/ds-5/optimize/arm-energy-probe/ for details about the ARM Energy Probe.
5
Thomas Petazzoni's profile photoAndrew Bradford (bradfa)'s profile photoGustavo Zacarias's profile photo
18 comments
 
As +Andrew Bradford said think of the circuit as two series resistors, one is the shunt itself, the other one is the circuit/load you're measuring, which gives three connections. Ground, a "middle" point between the two and VCC/Vsupply. You can find which one is which by looking at the little PCB, or alternatively with everything plugged in with a tester/multimeter measuring voltage: if you're set right you'll see polarity right and power supply voltage+power supply voltage minus a little difference (measurment) while moving the positive lead among the remaining two pins.
Add a comment...

Thomas Petazzoni

Shared publicly  - 
 
I'm really happy that we are working with NextThing Co, as that they are fully committed to having upstream Linux kernel support for their platform. Quite a difference compared to some other popular low-cost development boards!

Thanks to this collaboration, you can expect to soon see some kernel patches from +Maxime Ripard+Boris BREZILLON and a bunch of other french people, to bring some significant improvements to the Allwinner SoC support in the mainline kernel. And it is worth mentioning that the effort will not be limited to just supporting the board with only the existing mainline kernel features: it is also about adding support for SoC features not yet supported in mainline!

Stay tuned, and watch for the patches on LAKML !
 
Free Electrons working on the $9 C.H.I.P computer

Free Electrons has been contracted by NextThing Co to work on supporting the $9 C.H.I.P. board in the mainline Linux kernel.

http://free-electrons.com/blog/free-electrons-chip-nextthing/
Free Electrons has been contracted by NextThing Co to work on supporting the $9 C.H.I.P. board in the mainline Linux kernel.
6 comments on original post
32
1
Vladimir Pantelic's profile photoDrew Fustini (pdp7)'s profile photoJan Lübbe's profile photoJeshwanth Kumar N K's profile photo
3 comments
 
Are you going to use barebox? :-) +Markus Pargmann​ backed at kernel hacker for us as well. 
Add a comment...

Thomas Petazzoni

Shared publicly  - 
 
 
Alexandre Belloni now co-maintainer for Atmel SoCs in Linux

Free Electrons Alexandre Belloni has recently become co-maintainer of Atmel SOC support in the Linux kernel.

This makes Free Electrons in an even better situation to help you with your projects using the Atmel chips!

http://free-electrons.com/blog/alexandre-belloni-atmel-co-maintainer-linux-kernel/
1 comment on original post
15
1
Marco Cavallini's profile photoNicolas Ferre's profile photoDavid Anders (prpplague)'s profile photo
2 comments
 
It has very little to do with Atmel's decision +Marco Cavallini​. It's lot more of a community process and +Alexandre Belloni​ + his +Free Electrons​ colleagues involvement.
Add a comment...

Thomas Petazzoni

Shared publicly  - 
 
Just looked at the registration cost for the Embedded Linux Conference Europe, at http://events.linuxfoundation.org/events/embedded-linux-conference-europe/attend/register. The early bird price is $725.

Is it me or they are increasing the rate by $100 every year? It has become an insanely expensive conference. Shouldn't we get rid of some of the luxury and fluff, and get the price down to make it accessible to more people? I'm pretty sure a lot of companies just decide to not send people: the conference ticket is horribly expensive, and the hotels around usually as well.
22
1
Geert Uytterhoeven's profile photoThomas Petazzoni's profile photoTim Bird's profile photoDrew Fustini (pdp7)'s profile photo
31 comments
 
I'm sad to see the prices so high.  We used to use fairly low end venues (our first ELCE was at a University), and it was a very un-fancy event.  It was also heavily subsidized by the CE Linux Forum, and it was smaller.  With more attendees comes the need for bigger venues.  And working with the LF we've gotten very professional (IMHO).  We now pay for video production, instead of relying on volunteers. These things add up.

As for talks, I think Thomas is one of the few people we'd readily accept more talks from.  We try to limit him to only 1 just to avoid overloading him. :-)
Add a comment...

Thomas Petazzoni

Shared publicly  - 
 
I am looking for a tool to automate the execution of test scenarios for Buildroot. Basically, I want to create a large set of test scenarios, each of which will be building a Buildroot configuration, maybe running the result in Qemu, or doing some additional commands to verify that a particular feature is working. So it's mainly executing some shell commands, checking their return value and that's about it.

But I'll have a fairly large set of test scenarios potentially, and I would need the test to run in parallel from each other to get a reasonable overall execution time.

For now, I've started playing with Python unittest. It's doing the job, but it's not clear how well it supports parallel execution of tests, and I don't know if it's the right tool for the job. Also, I'm not sure about the reporting capabilities of the Python unittest stuff (like generating a nice HTML output with the overall result of the different tests).

Dear readers, do you have any suggestion? I've Google'd a bit, but there are so many testing tools available that it's hard to grasp which one would be appropriate.
4
Lionel Orry's profile photoThomas Petazzoni's profile photo
19 comments
 
+Lionel Orry Yeah, I've seen that one, just didn't had time to experiment.
Add a comment...
People
Have him in circles
804 people
Marc Souviron's profile photo
Marcin Wojtas's profile photo
Matthew Locke's profile photo
Steven Nguyen's profile photo
vipin tiwari's profile photo
Patrick Vergain's profile photo
Josiane Bats's profile photo
Shawn Goff's profile photo
Graham Whaley's profile photo
Work
Occupation
Embedded Linux engineer
Skills
Linux kernel, Embedded Linux, ARM, Buildroot, Cross-compilation
Employment
  • Free Electrons
    Embedded Linux Engineer, 2008 - present
  • Seanodes
    Linux R&D engineer, 2005 - 2008
Basic Information
Gender
Male
Links
Other profiles
Story
Tagline
Chief Technical Officer and Embedded Linux Engineer at Free Electrons
Education
  • Université de Technologie de Belfort Montbéliard
    Computer Science, 2000 - 2005
Contact Information
Work
Email
Address
Toulouse, France