Re: GSoC 2011 - BuildWithEucalyptus
On Mar 30, 2011, at 09:55, Zoltan Devai wrote:
> Hi Steffen,
> Let's keep top-posting then :)
> Basically, I'm happy to work on any project that makes sense
> and is part of GSoC, so I'm eager to learn what Jeremiah has
> in mind.
To be honest, I think anything that reduces the patch set between Debian and Maemo is good. If Ubuntu's CTO is working hard to make sure that there is as small a difference between Debian and Ubuntu as possible, that would seem to serve as a good template for Maemo.
That is a big area I know so maybe we should narrow it down. :-)
I think if developers have projects along these lines that they are interested in then that is the best thing because developers work hardest on the stuff they want to work on. I'll be happy to mentor (and help) anything as long as it addresses in some way the gap between Maemo and Debian.
I'll work on a more formal proposal to the Maemo lists today.
> But in any case, this is still one of my favorite projects, so I'd be
> happy if you could also consider me as a regular candidate :)
> About the co-mentoring:
> The google terms of GSoC explicitly forbid to be a student and
> mentor at the same time, but I'm certainly available to consult
> as an unofficial co-mentor.
> Did Rudy tell you more about his intentions ? If you want, CC him
> as well, no problem with that.
> 2011/3/29 Steffen Möller <email@example.com>:
>> Dear Zoltan,
>> firstly, sorry for top-posting. To answer your question:
>> sufficent, yes, feasible - if it all works more or less out
>> of the box ... but documentation alone would most likely
>> take up all your time. Your summary was very concise,
>> speaking for you more to take the role of a mentor than
>> as a student, really :)
>> For the virtualisation I had a previous contact with a
>> student named Rudy. He also made a good impression on me.
>> His description is not up on Melange, yet, but I would
>> want to wait for him, if he is still interested, just as
>> a principle. Would there be a chance that I get you as a
>> co-mentor for him, even when you are doing something
>> I am CCing Jeremiah from Maemo who very much liked the idea
>> to do a GSoC project to bring the Maemo achievements closer
>> to Debian. This should also me in the longer run as an
>> owner of the OpenMoko Freephone :) Jeremiah, would you
>> not want to discuss with Zoltan what he could do for you?
>> What is coming to mind:
>> * Cross compilation (and reduce Rudy's work on this)
>> * Maemo emulator
>> * Packaging of various bits and ensuring cross-builds of those packages
>> * Strengthening links to the embedded Debian community
>> * Getting the Qt-embedded bits as a regular Debian package (cross-built, most likely, takes a while)
>> Zoltan, please give this thought a spin. The overall context
>> is the DEX initiative to bring developments from "downstream"
>> distros back to Debian and from there downstream to them all
>> again. If you allow, the I would add Rudy to this thread
>> and you could then organise yourselves about who does what.
>> On 03/29/2011 01:31 PM, Zoltan Devai wrote:
>>> Dear Steffen,
>>> I would like to apply as a student for the Debian cross-platform
>>> development project, and I'd be happy to get some feedback from you for
>>> my proposal.
>>> The ultimate goal is to make cross-platform package building and
>>> development more easy, fix existing problems, and provide a long-term
>>> solution based on cloud-building. The focus would be on ARM systems, as
>>> it is probably the most popular platform after x86.
>>> (And the one I have the most experience with: I've been bringing up at
>>> least 5 different ARM-Linux systems from bootloader to userspace, with
>>> a custom, LFS-like distro).
>>> As an outline, the project would consist of:
>>> Work with qemu. Main goal is to create a single self-contained image
>>> that users can just boot with qemu and have an emulated arm system
>>> immediately. I would think that it makes sense to make a standard
>>> base system version, and one with all development stuff for package
>>> building pre-installed.
>>> 1.1. Standard way
>>> 1.1.1 Do a standard arm netinst in qemu, save the image.
>>> 1.1.2. Add development packages. Save the image.
>>> 1.1.3. Document above processes for further reference.
>>> 1.1.4. Make a preseeded installer initrd for automatic installation.
>>> 1.1.5. Wrap up the whole thing into a shell script for a
>>> "one-click" image creation.
>>> 1.2. Single-image way. Use a bootloader (u-boot?) to create a single
>>> image that can boot without kernel and initrd being explicitly
>>> specified. This also more close on how a real arm machine boots
>>> from a flash memory.
>>> 1.2.1. Create image including bootloader, kernel, initrd and rootfs.
>>> 1.2.2. Make automatic image creator also for this case.
>>> 1.2.3. Document above
>>> 2. Cross-building
>>> 2.1. Select a dozen of packages for cross-building (also some which
>>> don't have an armel port). Do it natively, summarise experiences.
>>> 2.2. Build the same in qemu. Summarise.
>>> 2.3. Fix issues related to cross-building, bring them to developers
>>> 2.3. Benchmark and tune qemu build times depending on qemu settings
>>> (RAM, SMP)
>>> 2.4. Dig up all tutorials/howtos/wikis on Debian related sites and
>>> correct/extend them. If needed, write a new one.
>>> 3. Eucalyptus
>>> 3.1. Set up a home eucalyptus farm for testing
>>> 3.2. Based on above experiences, set up a build server on the farm
>>> 3.3. Benchmark again.
>>> 3.4. Set up automated build server based on Eucalyptus
>>> 3.5. Document above
>>> So, what is your opinion? Is this feasible and sufficient ?
>>> Thanks for your time,