[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

[f-cpu] Re: help with linux from scratch

hi !

I switch to the main F-CPU mailing list because it might
trigger some useful answers from others.

Michael Riepe wrote:
> Hi!
> > > I did not install LFS, I built my system (and a bunch of others) from
> > > scratch. That's a difference... and it means that our systems differ
> > > substantially :(
> >
> > apart from the documentation, i don't see a difference ;-)
> - different software packages
> - different versions
> - different file placement / directory layout

well, the principle remains the same, i guess.

except that LFS is currently very poorly automated.
a few scripts would help to divide the installation
time by 2 at least. A few tricks would even reduce
the overhead from running ./configure multiple times
with the same configuration (making a link to a central
cache directory for example).

If you have ideas, i will be happy.
Of course, the manual installation is very instructive,
despite the length, and it was not a pain as i expected.
however it is very long and having to check all the scripts
manually, decompressing by hand, running the make distclean
by hand etc on all packages is sometimes boring.

It successsfuly booted, giving some error messages
and receiving unhandled signals, but it booted anyway.
Last time i compiled a kernel, i was forced to
reinstall everything :-)

With a bit of sleep (i just woke up after 11 hours,
the installation was more than 18 hours long)
and calm, i presume that the kernel unstability
might come from the use of hdparm. I have read that
there are corruption risks for certain options,
but not coming from kernel unstability (receiving
signals it doesn't know how to handle ?).

if you have a "stable" .configuration file for
a 2.4.17 kernel, i'll happily use it ;-)

> > > If you use the latest LFS, Simili will probably run like a charm.
> > i guess.
> >
> > > For Vanilla, you can use the same workaround as you do now (use the
> > > statically linked vv87 binary).
> > wouldn't a simple copy of the correct file (a library) do the trick ?
> Almost (you may need one or two configuration files as well).

there's one in /etc (ld.so.something) for the directories,
but what else ? I have identified the compatibility library to use
and where to put it, but i don't remember the rest...

> > > For the kernel, choose `Pentium-III'.
> >
> > it should do the trick :-) but the documentation
> > warns about too much optimisations which could make
> > the system unstable : not more than -O2 for the libraries...
> I've been using `-O2 -march=xyz' for my do-it-yourself systems.

most of the compilation were done with "686" by the makefiles.
i saw one or two tools with 486 ... but it's not a problem,
as long as it works.

I probably should not have followed the advice to run strip
for removing the debug symbols. Of course it removes megabytes
of useless stuffs but it maybe touched files it shouldn't touch...

> > > > Additionally, when the F-CPU simulator and the GCC patches
> > > > are ready, do you think that LFS is suitable for testing
> > > > and booting a simulated F-CPU ? (cross-compiled binaries
> > > > and simulated by a streamlined ncsim)
> > >
> > > I'm plan to use Linux for {si,e}mulation myself.
> > > I guess it's as suitable as any other 32-bit Unix ;)
> >
> > yes but i meant : LFS as a Linux flavour that could be first compiled
> > to F-CPU executables. Since LFS exposes all the sources
> > from a minimal working system, it is a good target
> > for attempting ports to F-CPU, no ? LFS as a first
> > distro for F-CPU would be rather easy, no ?
> If LFS is 64-bit clean, yes (has it been tested on Alpha?).

it's not a problem because F-CPU also handles
32-bit and 16-bits as well. The compiler has to do a clean
work, however.

> The really hard part is porting the kernel...

Anyway, LFS was a very interesting experience and
i also thank Etienne for his help. Without his
guidance it wouldn't have worked :-)

>  Michael "Tired" Riepe <Michael.Riepe@stud.uni-hannover.de>
To unsubscribe, send an e-mail to majordomo@seul.org with
unsubscribe f-cpu       in the body. http://f-cpu.seul.org/