Active Topics

 


Reply
Thread Tools
Posts: 838 | Thanked: 3,384 times | Joined on Mar 2009
#1
This thread is continuum of my "Rebasing Fremantle on top of Debian 6" [http://talk.maemo.org/showthread.php?t=89416]

I have reached one milestone and now I'm looking for testers and starting discussion how to utilize all of this.

Here are packages:
http://www.cs.helsinki.fi/u/rantalai.../wheezy/repo2/
http://www.cs.helsinki.fi/u/rantalai.../wheezy/waemo/

(there are also older repository: http://www.cs.helsinki.fi/u/rantalai...5/wheezy/repo/
and here is squeeze packages: http://cc.oulu.fi/~rantalai/maemo5/squeeze/repo/ )

--------------
Motivation:
*Claim1: new version of package is better than old (todo: proof for every package)
*Compiling new code will be easier if libraries and tools are newer.
*It would be possible to install packages straight from Wheezy-armel repository.
--------------
Goal:
(Term package refers to 'binary'-package, which means installable package, not source package, which produce one or more (even tens) binary-packages).

Just installed/flashed Maemo5 has 1745 packages (http://wiki.maemo.org/User:Aapo/Waemo/DefaultPackages), 355 of them are closed (http://wiki.maemo.org/Fremantle_closed_packages) and not scope of this.

Most open packages are made by Nokia and there are no newer Debian versions (but there could be newer upstream or Meego code, but it is not scope of this.)

There are 181 Fremantle packages which are also in Wheezy ( http://wiki.maemo.org/User:Aapo/Waemo/Upgradeable ).
(There are also 26 Fremantle packages in Squeeze but not in Wheezy. These might or might not be in scope of this).


First task for every package is to get it compiled (and its build dependencies). Then installed (and its dependencies) without bricking the device. After that starting to look if it breaks some functionality or if it is causing any side effects (CPU, memory, disk space).

---------
Goal2:
Recompile ~1200 Nokia packages using Wheezy toolchain. (This comes very close to the CSSU and thumb-recompiling.)

This step might be out of this scope, because many package works without recompiling (thanks to good backward compatibility). There are at least one counter example: hildon-application-manager must be compiled against newer libapt (which is now done and it is not bricking the device).

---------
Status on 1 Nov 2013: http://wiki.maemo.org/User:Aapo/Waemo/Status

These packages can be installed without immediate brick: http://www.cs.helsinki.fi/u/rantalai...mo_install.txt
(Note: I have installed top part of the list more often than end of the list (so they are 'more tested'))

You can download script to the /home/user before flashing. After flashing rootfs you need either ssh-server or rootsh installed via Application manager and then just run the script, it will take ~30minutes to run.

Next steps (maybe in parallel):
* Testers: upgrade your device and report what breaks (for sure something breaks, prepare for reflash!)
* Maemo developers: How to use this all (or part) to improve Fremantle?
* How to upgrade hard packages? (hal, udev, gtk, gstreamer)
 

The Following 45 Users Say Thank You to AapoRantalainen For This Useful Post:
Posts: 3,328 | Thanked: 4,476 times | Joined on May 2011 @ Poland
#2
Three suggestions:

Goal3: provide a sane building environment to be able for developers to build against wheezy

Goal4: (after goal3) create a maemo-reloaded autobuilder and extras - they would benefit from the newer compiler and newer libs (if statically linked) plus community components which need e.g. libc6 2.10

Goal5: (somewhere inbetween) integrate with CSSU (its qt 4.7.4 and maemo-specific stuff)
__________________
If you want to support my work, you can donate by PayPal or Flattr

Projects no longer actively developed: here
 

The Following 4 Users Say Thank You to marmistrz For This Useful Post:
Alecsandru's Avatar
Posts: 439 | Thanked: 282 times | Joined on Oct 2012
#3
dammm , you're good at what you're doing
__________________
N900 Beast featuring :
8GB OPTFS@ext4(*performance mount options)
overclock: cpu[125 - 1,150(*VDD1)+Dsp overclock
CSSU-T latest version
 
Posts: 95 | Thanked: 66 times | Joined on Mar 2012
#4
ahh, the Neo900 should be running on this foundation...
actually, even android devices should! lol.
 

The Following User Says Thank You to bozoid For This Useful Post:
norayr's Avatar
Posts: 148 | Thanked: 216 times | Joined on Jul 2010 @ Yerevan
#5
0. When is the step of building flashable image from these packages?
i. e. completely free maemo image, without proprietary nokia packages. And then we'll see how it looks and what can be done, what needs to be rewritten.

1. If I upgrade using your repository, and device get bricked I have no idea how to check what's wrong, which package causes that. If something breaks, but device boots, and it's possible to open a terminal, then yes, something could be done to investigate the problem.

2. Most important. Anyway, I think you are very close to creation of flashable image. And not only for n900, because n900 is a dying hardware, but for any modern device, which allows booting non vendor made operating system. Of course you probably have to start with n900. Ask here or in irc about how to create the image.
I believe you should follow mer strategy and separate the kernel (hardware adaptation) from the rest of the system. So that you can build all maemo packages for different architectures, and in the same time, when preparing an installable image, at this step you unite those packages with vendor specific kernel / hardware adaptation, which is usually non-free because it contains non-free software, drivers, and usually we can do nothing about it.

3. Even active n810 users will benefit from that. For instance, I use n810 every day, and nokia did not provide fremantle image for n810, I believe mostly for marketing purposes. Openmoko community will benefit from that - they could run fremantle like operating system on their devices. Nexus users could benefit, they could run canonical GNU/Linux system instead of their inferior, sorry, androids.
 

The Following User Says Thank You to norayr For This Useful Post:
Posts: 3,328 | Thanked: 4,476 times | Joined on May 2011 @ Poland
#6
Aapo: did you count the sdk/dev stuff in?
__________________
If you want to support my work, you can donate by PayPal or Flattr

Projects no longer actively developed: here
 
Posts: 21 | Thanked: 27 times | Joined on Jul 2013 @ Czechia
#7
How hard would it be to install this system on an μSD card so as not to have to reflash? :-)

I already have Arch Linux ARM there as a backup system.
 
Posts: 71 | Thanked: 177 times | Joined on Aug 2013
#8
Originally Posted by lenoch View Post
How hard would it be to install this system on an μSD card so as not to have to reflash? :-)

I already have Arch Linux ARM there as a backup system.
Look here and here
But don't follow it... (no crazy copy&paste please! )

PS.: Going to spend my today on that!

Last edited by lexik; 2013-11-03 at 19:31.
 

The Following 7 Users Say Thank You to lexik For This Useful Post:
Posts: 12 | Thanked: 47 times | Joined on Sep 2013
#9
Hi,
as far as I understood you start off with PR1.3.1 and then add Debian packages to the mix. Why do you do it that way round? In my understanding "Rebasing Fremantle on top of Debian 6" would mean that you start with Debian 6 and then put Fremantle parts on top of it. That way you start with something that is fully Debian (apart from the kernel and some modules) and diverge from it with fremantle step by step. Your method seems to be prone to never have a "real" Debian system installed because of all the leftover cruft from PR1.3.1 with which you started.

Back in the days I booted Debian on my N900 using instructions from here: http://elektranox.org/n900/index.html
 

The Following 6 Users Say Thank You to josch1337 For This Useful Post:
Posts: 3,328 | Thanked: 4,476 times | Joined on May 2011 @ Poland
#10
Aapo: are there any status updates on this? Development halted? Simply too few testers too continue? Or maybe no time / focusing on bringing home the bacon / whatever? Or simply is it as silent as sound wave with amplitude approaching 0 as 1/t for reasonably big t?
__________________
If you want to support my work, you can donate by PayPal or Flattr

Projects no longer actively developed: here

Last edited by marmistrz; 2014-02-23 at 19:22.
 
Reply

Thread Tools

 
Forum Jump


All times are GMT. The time now is 08:29.