[Clfs-dev] why there isn't any standard CLFS ARM Book?

[Ken Moffat]

On Thu, Nov 22, 2018 at 01:41:26AM +0100, Michele Bucca wrote:
> Il giorno gio 22 nov 2018 alle ore 01:10 Michele Bucca
> <michele.bucca at gmail.com> ha scritto:
> >
> > Hello There
> >

Hi Michele,

I don't have any interest in RPi, but a couple of comments.

> > I was wondering why there wasn't a regular CLFS book for the ARM
> > architecture. Yes, we have clfs-embedded but we don't have a regular
> > CLFS Book. The only architectures supported are x86, x86_64, sparc,
> > mips, but not ARM. Why?
> >

For this, I believe that CLFS is moribund.  X86 and little-endian
won.  What is probably needed is (as in LFS) developers.

But whether a non-embedded Pi build makes general sense (rather
than "because you can") I have no idea.  In particular, a system
which lives on SD cards appears to have limited lifetime - based
on comments I was reading elsewhere about what people would like
in the next Pi.

[...]
> > Nowadays there are a lot of ARM based Single Board Computers like the
> > Raspberry Pi so I think that maybe the ARM community would benefit
> > from a book like this.
> >
> > I know that there's a PiLFS project out there, but it's not a book,
> > rather a set of scripts that build LFS for the RPI. Still it's a
> > pretty useful resource.
> >
> > I'm having trouble building a Cross-compiler for ARMv7 (soft float)

A quick gurgle found a comment on a forum that Raspbian targets
hard-float.

But like all replies on forums, it might be mistaken.  If so, I have
no idea, except that another result said that changing to armv7a
apparently solved a similar problem (it also said that armv7 is the
*intersection* between armv7a and armv7m, so not particularly
useful).  But hey, hacking on toolchains is always fun (for painful
values of fun).

> 
> Then I have another Toolchain for ARMv5, also with soft floating
> point. Glibc there builds only if I tell it to do not abort the builds
> when there are warnings (../configure  --disable-werror [..]). This
> toolchain was used to create a chroot: it builds all sort of things
> such as binutils, gcc, make (only if I patch a file with a sed
> command), busybox, xz but it fails to build tar with this error
> 

Using --disable-werror is often a good idea with a current
toolchain.

>  CC       areadlink-with-size.o
> In file included from
> /home/michele/x-tools/arm-linux-gnueabi/arm-linux-gnueabi/include/stdlib.h:1014,
>                  from ./stdlib.h:36,
>                  from areadlink-with-size.c:28:
> /home/michele/x-tools/arm-linux-gnueabi/arm-linux-gnueabi/include/bits/stdlib.h:
> In function ‘wctomb’:
> /home/michele/x-tools/arm-linux-gnueabi/arm-linux-gnueabi/include/bits/stdlib.h:90:3:
> error: #error "Assumed value of MB_LEN_MAX wrong"
>  # error "Assumed value of MB_LEN_MAX wrong"
>    ^~~~~
Again, google finds a number of matches for error messages like this.
Some were fixed by patching packages to add a header - I suspect that
your headers are wrong for v5, but no idea.  Maybe looking at that
line of stdlib.h might give a clue where to look.

ĸen
-- 
If a man stands before a mirror and sees in it his reflection, what
he sees is not a true reproduction, but a picture of himself when he
was a younger man.        -- de Selby