On Fri, Mar 18, 2016 at 08:01:45PM +0000, Ken Moffat wrote:
> On Fri, Mar 18, 2016 at 12:59:16PM -0500, Ronaldo Antonio Carballo wrote:
> > I'm trying to build my first LFS and I keep wondering if I'm doing this
> > right. This is what I'm trying to do with my LFS:
> > 
> > 
> >    - Create a Linux image that supports up to i586 processor to test
> >    software that can only run on the i586. We're trying to test an image 
> > that
> >    does not have MMX/SSE instructions.
> > 
> > I have gotten up to step 5.9 (Binutils Pass 2). That's when I see that
> > after I run the "make install" a "i686-lfs-linux-gnu" folder shows up under
> > "/tools/bin" of my LFS partition. So, I'm wondering if this step is missing
> > the "--target=i586-lfs-linux-gnu" in order to properly install the new
> > binaries in the "i586-lfs-linux-gnu" sub-folder.
> > 

[ further thoughts, now I am winding down, but first an elaboration
]
> That sounds a likely solution to your problem, but it means you will
> still be cross-compiling, so you probably need to watch every
> subsequent compile.  And the only way to judge success will be if
> the software you need to test works correctly for all tests - if it
> doesn't, identifying what is not i586 will be awkward.

At this point in LFS you are, of course, pseudo cross-compiling
(i.e. changing enough to make the toolchain think this is not a
native build).  But after pass 2 of gcc and binutils, everything is
native in a normal build.

For you, everything in the book needs to cross-compile for a lesser
CPU.

Perhaps you could be more specific about *why* you want to do this ?
At the moment we know you have a userspace application, presumably
without source, which only runs on i586 linux.

The "obvious" answer is to buy suitably old hardware, if you can
find any, and run the application for real (sometimes, running
binaries on much faster than expected processors has been known to
give problems - running on real hardware would avoid that).  Possible
problems include finding parts (particularly disks that such an old
motherboard will talk to) and the general problems of running an
application on obsolete hardware.  Ah, the days of 100 MHz single
processors and PC100 memory - I would NOT attempt to build a modern
linux toolchain on such slow hardware.

Is using a VM an option ?  If so, google finds e.g. i586 qemu rpms
from opensuse-13.2.  Not the absolute latest, but capable of running
far faster than a real 586 (and I'm sure you can detune the
processor memory, and perhaps the speed, in qemu - or maybe slow
down the CPU (a bit) using a cpufreq driver in the kernel).  Note
that I'm not certain those will definitely let the application think
it is on i586, but that seems a reasonable bet.

OTOH, trying to build current LFS for i586, but on i686, is an
interesting mental exercise.  But if this is for a commercial need,
whoever is signing off on the approach needs to understand what might
go wrong.  I would not want to put anybody's job on the line.

Of course, if it is a task that has been set to you as a training
exercise, give it a go and I hope you succeed.

ĸen
-- 
This email was written using 100% recycled letters.
-- 
http://lists.linuxfromscratch.org/listinfo/lfs-support
FAQ: http://www.linuxfromscratch.org/blfs/faq.html
Unsubscribe: See the above information page

Do not top post on this list.

A: Because it messes up the order in which people normally read text.
Q: Why is top-posting such a bad thing?
A: Top-posting.
Q: What is the most annoying thing in e-mail?

http://en.wikipedia.org/wiki/Posting_style

Reply via email to