I have been playing around with chroot build enviroments as i'm hoping to support Larry with the development of the MiniMyth Chroot enviroment. During my tinkering I have had a few thoughts that i thought I would share.
I started off looking at the lnx-bbc debian chroot, which seemed pretty cool as with apt-get you can get all the tools you need to build MiniMyth. Then I got to thinking, if we are going to have an NFS-root MiniMyth, the size of the root filesystem is pretty much irrelevant. So I made the chroot enviroment into a diskless NFS-root booting system on my EPIA.
This is just an experiment at the moment but it does open up another way of running and building the system. With setting things up this way you can either chroot into MiniMyth and build it on the server machine. Or you have the option to ssh into your front end system and build it from there.
So it does open up the question, are we going to have a separate chroot build enviroment, or do we just have one root filesystem that does it all. Larry any thoughts on this?
Anyway, I am still experimenting, but if anyone wants to see what i have done so far give me a shout.
I have ditiched the idea of debian as the base chroot as it seemed sub optimal to me because it compilies with a 386 architechture.
I have now moved to gentoo, I have built everything from the ground up optimsed for my Nehemia(of course running from an NFSRoot, not that its a fetish for me to NFSRoot stuff or anything honestly its not! ). This is looking more promising. I have been compiling things on my EPIA frontend and also tried building things from my backend PC. Both seem to work really well.
I'm not really sure which way Larry is intending to go with the MiniMyth chroot yet, the options seem to me to be
A) An optimised distribution such as the gentoo system I have built
B) A DIY system.
Option A is good because we can leverage the work done by developers of distributions like Gentoo. Its quick and can be produced in a matter of days (i'm already a long way towards doing this from my experiments).
Option B - would be good because we will know that everyone is using exactly the same, and it would probably be a more controlled enviroment. e.g. we are all using the same compiler. I'm just looking at the amount of work required to get this going and its quite scary.
I would like to know if the chroot is going to be separate from the final running root file system or just part of it. I'd personnally like to see it be part of it, as if all else fails you can simpliy compile stuff and install it right from the frontend.
Also i'm curious to know if we are going to GAR the chroot system or just have it as a download.
I still have a bad taste in my mouth from uclibc from the last time I played with it and if we're using an NFS mounted root I see no need for it.
I briefly played with the debian chroot build environment referenced by GAR and it seemed ok but the build of lnx-bbc failed.
I am a little reticent to use gentoo as it seems (from watching IRC) that gentoo users have no end of problems with other peoples' programs. I'm not trying to start a big debate, just my observation.
It is my intention that only folks that want to contribute components to this system should have to build it, the rest should just be able to download it. I have more of a knoppmyth philosophy than a gentoo one.
To be quite honest, I like having built this from scratch because I know what every file does and why it is there. No bloat. So whatever we pick I want to be able to keep it to a minimum.
I *do* like using GAR as it is easy (usually) to upgrade individual components so initially I want to stay with that approach.
And then there is the obvious question of why don't I just add EPIA M support to knoppmyth. Because that is not as much fun as making my own distribution.
But first, I want to finish minimyth-0.4. Stay tuned...