Re: fakeroot with chroot.

[Junichi Uekawa <dancer@netfort.gr.jp>]

> > I once tried to do something similar, but noticed that  
> > user-mode-linux does the same thing to a fuller extent. 
> >  
> > If you look at it this way, user-mode-linux is a fakeroot that traps 
> > all syscalls. 
>  
> user-mode-linux has too big overload and it is available only for i386 
> architecture. I've tested my modified fakeroot on alpha, sparc and ia64 and it 
> works pretty well. 

user-mode-linux can be ported to other arches, and although most ports 
are not complete, I have seen several attempts to port them.
  
> I read another thread about fakeroot and I think I'll start new project called 
> i.e. fakechroot which will be based on original fakeroot. The original project 
> should be as stable as it is possible. My library is too experimental. 
>  

Yes, fakechroot would be a good name, I considered using that name
myself when I tried writing my version of extended fakeroot.

However, with fakeroot you are only overriding function symbols,
and syscalls are not overloaded. I don't think it's really a recommended
practice to call system calls directly without using glibc, but 
it is possible, and fakeroot will not work in that case.

There are pretty ugly corner cases due to the way symbols are 
resolved, but since you seem to have gone quite far with fakechroot,
I would wish you good luck in completing it, it looks like a very
exciting project  :)


regards,
	junichi