Pandora should leave Angstrom

[WizardStan]

I think a full OS for SD cards is worth having, but only if there is also one that fits on the NAND

Which is almost exactly what we have now. A few issues if you do a full upgrade, but as the OMAP necessary changes get committed upstream this will change.

 

[Binky]

Since it will be possible to fit functional Linux in 512MB for a long while yet, I think the official OS should not force people towards SD card boot.


Does anyone boot from NAND, but keep a recovery SD card to boot from when they mess up?

 

[ZXDunny]

Since it will be possible to fit functional Linux in 512MB for a long while yet, I think the official OS should not force people towards SD card boot.


Does anyone boot from NAND, but keep a recovery SD card to boot from when they mess up?

I boot exclusively from NAND, but I always keep the latest hotfix/firmware on SD1 in case I need to reflash.


D.

 

[laurencevde]

Let's see: some reasonably complete OS MUST sit on the NAND. It IS big enough to put a (cleaned-up and stripped down) mainstream Distro on there, so there is NO reason to force the use of an SD-card. Forcing the use of an SD-card means maintaining 2 OS's. Such an OS can be copied over to an SD-card to gain enough space to make full use of the distro, satisfying everyone's needs.


Once someone actually builds such a mainstream distro for the NAND, there'll probably be little reason to stick with Ångström.


So, the biggest question still remains: who volunteers?

 

[ibisum]

I'd volunteer to help with the Angstrom-based OS, but the organization of the project frustrates me every time. We need a build server, some place to check things into a repository, some way to reproduce, automatically, the public-released versions of the OS (so, no hand tweaking and so on), and someone to watch over it all. The current situation is untenable and doesn't really allow for productive contribution - one has to pull the OE sources, patch them for Pandora, add additional tweaks, keep up with Notaz and his brilliance, and so on - this all would be better handled through a common build server tied directly to a repo that can be easily pushed to by multiple sources, with someone overseeing it.


I'd like it if I could push my changes to the repo, come back 10 minutes later and get a binary that I can boot from on SD.


I *don't* like having my own complete OE build system locally that has the potential of going out of sync with the main contributors, all the time.

 

[slaeshjag]

I might be able to get a server up for the purpose of playing buildserver until there's an official one, provided somebody can configure and oversee everything. The box is a first gen i7 with 4 or 8 GB of RAM.

 

[ibisum]

I think Notaz has the best current tree for building the OS. I think, also, that OP do have a build server. I don't know how they're organizing things internally, I have a feeling things are pretty close-knit, but I'm just stating what I think should happen if we're going to get more volunteers involved in the process of pushing things forward.

 

[Caine]

A pre-configured virtual machine with a build-able version of the OS, version management tools (for checkout and update, I would not recommend letting everyone commit), SSH and samba and a wiki page with with build instructions would greatly reduce the threshold to entry.

 

[ibisum]

I agree, Caine.

 

[notaz]

The "build setup" is basically pandora-oe-environment.git, you pull that, run initial-setup.sh, source op-env-oe.sh, do 'bitbake pandora-xfce-image' and wait.


Surely it will sooner or later fail due to any of these issues:


1. source mirror of package being built is dead


2. implicit dependencies to host packages that OE failed to warn you about


3. non-public sources for SGX and DSP, that we are not allowed to redistribute, but can be retrieved if you register at TI and pass their software export screening thing.


1 and 2 is what OE is all about, never ending breakages that you have to manually fix up each time. And even after you fix all of them, a week later some mirror will die again and some host distro package update will make something not buildable again.


3 is something we can do nothing about, distributing VM with these would make it illegal.


Sure something still can be done, you can make a list of files people have to get manually from TI, set up another mirror for dead sources, but it is not one evening's job. Not even week's worth of evenings. At the same time there are pandoras being made with RAM nobody can use because suitable kernel is not stable enough, should I delay that even more and go make that vm? Maybe not. Why do I have do this at all, it's not like I'm the only one who can type on a keyboard around here? It's not like nobody done it before too (aTc).


The eternal problem with pandora project is that there are not enough people doing things. You can blame mismanagement, but that's on top of ED that has 10000 other things to do.

 

[ibisum]

Notaz, if you could legally share a VM, wouldn't it be worthwhile for you to do so?

 

[notaz]

I'm not really sure it will help, if you can't bother to try checking out that git and trying a build (haven't seen any report from you about that), will you bother to do any firmware work?

 

[McLovin]

Well, Notaz. Thanks again for all your efforts. You are the one who knows the system and it's benefits/problems the most. Do you have something like a roadmap? It seems you like to keep your current project a secret, but it might actually help to team up if we knew your priorities.

 

[notaz]

Currently only tweaking the kernel, finishing SDL update, everything else is not firmware related.

 

[Linux-SWAT]

Currently only tweaking the kernel.

Is a new release expected before the end of the month ?

 

[skeezix]

torpor (and a few other people) -- take with a grain of salt but you are typical here -- you offer help, but never follow through Sure, you have entirely valid reasons, but thats not the point .. why not help beat those reasons? Winners win, doers do. You want you submit patches, you gotta get in and start hacking


ie: Working on a firmware is not something that can be handed to you on a platter; doing so would take up a few peoples time keeping it in that ready-to serve fashion - and as witnessed over the years.. we cannot get people to submit patches or anything at all, so getting someone to sign up as an official repo keeper or release manager or the like.. not likely to happen. (Do you want to do it? I thought not You offered to help set up package server and such, and ED got it 90% of the way there (its still there, still works, we just didn't formalize it..) -- jump on in


We have, as always, a very limited number of hands. notaz is _amazing_, and he caretakes the loader, the testers and flash scripts, the kernel, the improtant bits; he also does SDL and his emus and ports and so forth.


Me, I do what I can, but thats libpnd, minimenu, some of the daemons and pnd system (pndnotifyd, pndefvmapperd, etc), documentation here or there, design and so on, but I can't commit to growing my scope much.. got 3 small kids now, tis rough It pains me, but day one I sort of said I had to limit my scope due to RL. (Were that this a few years earlier, but such is life


djhwillis has some issues with family that tie him up, but he's still our firmware designer and such.. just not able to do day to day stuff, though he come sin and does burst.


ED -- well, he does everything, from working on scripts, major testing, the boot artwork and website and 1000 horurs a week in the forum, to selling, to video and showing up at shows..


And a few other stragglers with some patches and help (thanks!), and some folks like porg and pmprog and so on who help with the bug tracker and such. (edit: I'm not trying to forget people .. WizStan with his battery mods, sebt3 as thw owner of pnd_run.sh, vimacs are the father of pnd_run.sh, Ivanovic..!)


Thats just not enough; we can't take on any more work, we're already 1000 feet below and trying to get air.


So please.. jump in. Don't complain about why you're not.. start trying to solve those things


You want a VM thats ready to build.. go ahead, make it; its not hard. 'make setup-oe' invokes git and pulsl latest changes down, and you include a base firmware pull from day -X. 'make' -> kicks it all off. Boom, easy. The TI and other binary blobs are a pita, but include instructions on how to obtain them.


..


Your criticisms are valid, but not helpful, is what I'm saying


You need us to do _more_ to get folks to help? (Maybe that even makes sense..) -- but ain't going to happen any time soon, we're all buried (esp ED with building new units etc.)


..


It _does suck_ to start building OE, though.


..


We do have a build server that spits out all the ipk's and such; thats where notaz and ED run builds


jeff

 

[WizardStan]

WizStan with his battery mods

The battery mod which ended up causing problems until Notaz rewrote it, you mean?

 

[skeezix]

Well, no one is perfect (except the 27 people who are all collectively known as Notaz^H^H^H^H^HThe Dread Pirate Roberts.)


jeff

 

[Askarus]

Pandora Os is good for me and I use my Pandora as a Netbook and a Gaming console.


I can run LibreOffice with PanDebian. I only run our of Ram when I load big presentations but the I can use Swap on a fast USB-Stick and at least the whole Presentation loads..


What's the Problem.


I don't see a reason why we need a new OS. It might be nice but do we need it?

 

[pmprog]

@skeezix & notaz: I think the problem is that it's application developers looking to assist in the firmware, but it's actually quite a different beast. Skeezix can attest to how much I struggled getting him all the info to add WiFi into MiniMenu (how's that coming btw? ), Linux in general is just as alien to me as firmware side. So I don't think the questions here are more "please just give us everything", it's more like, "what am I supposed to be looking at?". Granted, the first step is actually to build the firmware I think, but Notaz's first to points "source mirror of package being built is dead" and "implicit dependencies to host packages that OE failed to warn you about" are surely going to be difficult to work through when you're not even sure what you're looking at, but I suppose until you've even got to that point, it's superfluous.


Real Lifetm has totally consumed my life for the past couple of weeks, but things are changing in the near future which should give me more time. At that point, I might see if I can get a VM up that can compile the firmware (though I might need some assistance, but we'll see how it goes), and have text files and/or scripts on the desktop on getting the latest code, running the build, what to look for when it fails, etc. Also point to where the proprietary sources should be put so that I can provide the disk image without problem.