:: [DNG] Building from source [was Re:…
Top Pagina
Delete this message
Reply to this message
Auteur: Didier Kryn
Datum:  
Aan: dng@lists.dyne.org
Oude Onderwerpen: Re: [DNG] Runit crisis: was GNOME usability "improvements"
Onderwerp: [DNG] Building from source [was Re: Runit crisis: was GNOME usability "improvements"]
Le 12/08/2017 à 05:37, Adam Borowski a écrit :
>
> ... You can tinker on your home desktop, which is nice for
> development and for exploring new ways, but it's not something for an
> average user, nor anything that has a place anywhere nearby a production
> machine. If I deploy a server, I can run a git version of its main service,
> but the system's base is supposed to be well integrated already.
>

     I'm building a GNU+Busybox Linux OS with all applications 
statically linked against Musl libc. It is for fun and will never 
replace a distro; and it is very time-consuming actually.

> There's a reason LFS is a curiosity rather than something for everyday use,
> despite its undeniable educational value.
>

     Gnu/Linux is almost never assembled from scratch in reality. It is 
built from a full-featured and recent Gnu/Linux development platform. 
When you want everything recompiled from scratch, say when porting to a 
new arch or linking against another libc, there is a bootstrapping 
problem. I tried LFS several years ago. The manual and the patches had 
taken so long to the authors to work-out that all packages were very 
outdated. In addition the version based on Musl libc (which I was 
willing to build) was not working AFAIR.


     I now have, in a chroot, a functional development platform, meaning 
I have passed the bootstrap phase and everything is now much easier. All 
commands are statically linked except Python. It took me 4 years partial 
time, with a lot of periods in which I was totally discouraged. I 
haven't documented all the steps, but the system can recompile itself 
entirely from the official sources. I didn't start from LFS but from a 
mix of Aboriginal and Debian and it was tricky because I wanted that my 
gcc understand Ada; it has Ada, C, C++, Java, Objective-C and 
Objective-C++ (I only speak the first two).


     I plan to use this system to experiment how X.Org and Mdev can play 
together - still need to build X.org :-) . I prefer experimenting in 
this simple environment to not be encumbered with package-related 
things, eg Udev.


     Didier