The main problem is that a lot of processes are started before you get such mounts done. So those processes/apps will fail with weird problems. In a proper unix this is solved in such a way that all programs on /bin and /sbin must work even without /usr (or /opt) mounted. Fro maemo that's not the case, distro maintainers didn't bother to move programs from /bin to /usr/bin (or the other direction) accordingly to the needs of system startup process, so they kept /usr on rootfs and rather moved packages to /opt (a mount) with symlinks one by one, with this abomination called optification.

/j

I have list of all open and closed maemo packages, also list of all preinstalled packages. See: http://talk.maemo.org/showthread.php?t=75174

I'm aware of your scripts, and I'm one actively writer of http://wiki.maemo.org/Fremantle_closed_packages Next I need wiki-page listing all preinstalled packages.

I have found so many useless nokia patches (some are written for Diablo, some are merged to the upstream) that I will try good old "If it compiles it is good. If it boots, it is perfect"

So first compile ~whole debian and just install package by package till something breaks.

I didn't know that about udev. But maybe optification is not yet issue. If preinstalled packages fit before I hope they will fit after.

I tried it and I hit immediately hard circular dependency loop, which will be no issue if there are packages from squeeze (everything needs newer dpkg, but new dpkg needs recent debhelper, gettext and something more). But yes when things start rolling Wheezy is target then.


I'm not sure what is benefit for very this moment use gcc-4.7-thumb (of course in long term). During process phone might brick often, so flashing should be easy and fast. Maybe this is good time to create own Fiasco-image with thumb-support.

It's a very interesting project but also very ambitious. Be aware that other similar attempts have failed previously (remember the first Mer?).

Also keep in mind that Debian is too big and not optimized for mobile use. Maemo has already done a lot of the initial work in that direction.

My suggestion for you, would be to start from an already working base, like CSSU Thumb (so you can be synced with fixes and features from CSSU) and newer toolchain, then go as far as you can without breaking the device's functionality.

There are several upgraded packages floating around the forum, some from Harmattan, some form Debian or upstream that could be easily added to that base. We have working (and better) replacements for common applications such as Worldclock, BusyBox Power, Hamster Filler, Open Media Player, improved Nokia Maps...

With not so much effort, and probably with more chances of receiving help from others, you could build a sort-of "Maemo Reloaded", keeping compatibility with most of (if not all) the already existing applications.

My proposal in such a project would be to keep as much compatibility as possible with existing applications, but without being as cautious as CSSU must be. Top goals would be to upgrade to upstream where possible and substitute closed source applications with OSS replacements where available.

You would need a new repository, in which applications could depend on these newer packages.

A second sub project could be the creation of a new fiasco image with saner partitions (i.e. ext4 and 5Gb for applications) and investigate the possibility of unoptification and decompression of root filesystem. But this can be made in parallel with the first approach.

Be active in the merproject instead of reinvent the wheel.

Being active in mer also is good for for sailfishos on upcoming devices and nemomobile on N9/N900. I dont get WHY people is using so much time on a already dead OS instead of being active to poirt apps/improve nemo instead...

@mikecomputing: because maemo is working. all other attempts to use any other system failed or were slow, not usable or missing applications. Maemo has fixed all problems: is working, is faster then any other os (on n900) and has lot of applications in extras. I really do not understand why everybody pointing to mer, nemo and sailfish...

@ivgalvez: Instead breaking and thinking how to fix opt/rootfs problem, first use eMMC or SD card. When system will be bootable, then start thinking what to do with fiasco and opt...

I share your dream, in fact I've been busy all week getting an alternative SDK setup and trying to recompile the squeeze version of pulseaudio...

So you can count me in!

And it may be a babystep, but updating the SDK fully to Lenny (Debian 5) would already be a good leap forward in my opinion. At least that would get most of the bottlenecks above water without the added extra complications of the usual dependency hell.

@Zaerc: Closed nokia pulseaudio modules are compatible only with maemo pulseaudio version. So when you update pulseaudio you will break full maemo system (cellular gsm/3g calls, ohmd, libplayback, mediaplayer, ...)

Rather do not touch pulseaudio, instead try updating other system/generic packages which are not patched by Nokia.

If you want, I can generate wiki page of all preinstalled/sdk/open packages.

For example nokia patched HAL package, it provides more addons which are needed for more maemo applications (some omap modules, gpio keys, ...). So make sure that you drop only useless patches (e.g apt patch which disable using apt-get for ovi store :-))

First create system which will work from one SD card paritition and then start thinking how to fit it into NAND+eMMC...

Old gcc in scratchbox has some bugs and is old. Rather use new gcc which has fixed bugs and have thumb2 support for n900. Really forgot not about flashing via flasher-3.5/0xFFFF. Use sd card and untar your rootfs directly. If this will work, I can write hints how to generate ubifs for nand and how to pack it into fiasco image with 0xFFFF.