Package management and creating the file system layout

lists at kitepilot.com lists at kitepilot.com
Tue Dec 16 23:24:13 PST 2008


Hello there: 

Maybe this is irrelevant and I hope I don't offend anyone... 

I've been working on a LFS package manager that works out of the tools 
available in the "tools" directory (plus rsync and file) and that I could 
use to manage my LFS servers. 

I used the union-fs tools and I believe I have achieved some useful results. 
For one thing, I begin building usable packages using an unprivileged user 
right from Chapter 6 and I separate my packages (roughly) into "bin" for 
production and "dbg" for, well, everything else...  :) 

I strip production binaries and verify that dependency libraries are 
installed. 

I am already instantiating real machines out of my tool and I have devised a 
bunch of scripts that facilitate the build.
Among other reasons, because you have to chroot 3 times for every package 
you build, but my set of scripts and aliases make the pain almost 
transparent. 

I also built a script that downloads the packages an instantiate a machine 
that is indistinguishable from a machine built by the book.  Which creates a 
LFS configuration sort of like the one is proposed here (sort of like!) 

If there is any interest, please take a look at a sort-of-working version 
at:
http://www.kitepilot.com/LFS/B.-MANAGE-MACHINE-README.txt 

This message may look a little bit incoherent because I am tired (and 
frustrated) trying to build an iniramfs generic kernel to boot my machines.
I'll have it done tomorrow or will give it up... 

If there is any interest, I'd like to write a hint about this.
Thanks
Enrique A. Troconis
LFS ID 13133 

 

 

 

 

 

 

DJ Lucas writes: 

> Guys, while I'm waiting on my automated build to complete, I'm running 
> through a manual build on a second system that I have, in order to test 
> the idea of using a DESTDIR installation.  I must say that it has been a 
> *long* time since I did a full manual build.  Anyway, can anyone provide 
> a technical reason not to create the entire 'base' while in chapter 4?  
> I think with current host system requirements, the syntactical 
> differences can probably be thrown out the window (and if not, then the 
> host requirements can be updated). 
> 
> The reason that I ask is that I'm looking at this from a packager's 
> point of view.  Everything in section 6.5 and the second, third, and 
> fourth set of commands in section 6.6 (creating /etc/passwd, /etc/group, 
> and /var/log/{lastlog,{b,u,w}tmp}files) belongs in back in Chaper 5 so 
> as to create an "LFS-Base" package.  Additionally, this gets rid of the 
> "I have no name" bit, and the extra invocation of bash.  The "I have no 
> name" prompt is the only educational piece that is lost (and it could 
> stay, although it'd have to be changed a bit), while logically creating 
> the entire base at one time sets the stage for proper packaging (IMO). 
> 
> Thoughts? 
> 
> -- DJ Lucas 
> 
> -- 
> This message has been scanned for viruses and
> dangerous content, and is believed to be clean. 
> 
> -- 
> http://linuxfromscratch.org/mailman/listinfo/lfs-dev
> FAQ: http://www.linuxfromscratch.org/faq/
> Unsubscribe: See the above information page



More information about the lfs-dev mailing list