From rob at landley.net  Fri Jul  8 05:46:01 2016
From: rob at landley.net (Rob Landley)
Date: Fri, 8 Jul 2016 00:46:01 -0500
Subject: [Aboriginal] Hello world.
Message-ID: <577F3E19.4010608@landley.net>

So I'm looking at http://github.com/richfelker/musl-cross-make and Rich
has got it doing native compilers now, and it theoretically supports all
the architectures except sparc and m68k, and I still haven't fixed up my
old toolchains (which may now be broken on _4_ architectures), so I
should probably just switch over and focus on getting lfs-bootstrap.hdc
updated and working and get back to toybox testing towards the 1.0 release.

The downside is, this means no more precompiled binary toolchains (cross
or native). I will not distribute GPLv3 binaries, ever. But if Rich
wants to put them up somewhere, I'm happy to point at his server...

Working otu thelayering is fun. I'm speaking at Texas LinuxFest tomorrow
and saturday so won't get much done on it this weekend, but I think
that's the path forward for the project in the short term.

Long term I'd still like to get llvm working, but that's a bigger jump
and still hasn't got a @*%(&# binutils replacement (which is what's
BREAKING more often than gcc is).

Rob


From alessioigorbogani at gmail.com  Fri Jul  8 07:20:05 2016
From: alessioigorbogani at gmail.com (Alessio Igor Bogani)
Date: Fri, 8 Jul 2016 09:20:05 +0200
Subject: [Aboriginal] Hello world.
In-Reply-To: <577F3E19.4010608@landley.net>
References: <577F3E19.4010608@landley.net>
Message-ID: <CAPk1OjGO-xGM8ZqPPqfc8ycN8Kd-81UsDJ4ZyBN51ftBKz7_RQ@mail.gmail.com>

Rob,

On 8 July 2016 at 07:46, Rob Landley <rob at landley.net> wrote:
> So I'm looking at http://github.com/richfelker/musl-cross-make and Rich
[...]

How this cope with autotools based software?

Ciao,
Alessio


From rob at landley.net  Fri Jul  8 08:31:15 2016
From: rob at landley.net (Rob Landley)
Date: Fri, 8 Jul 2016 03:31:15 -0500
Subject: [Aboriginal] Hello world.
In-Reply-To: <CAPk1OjGO-xGM8ZqPPqfc8ycN8Kd-81UsDJ4ZyBN51ftBKz7_RQ@mail.gmail.com>
References: <577F3E19.4010608@landley.net>
	<CAPk1OjGO-xGM8ZqPPqfc8ycN8Kd-81UsDJ4ZyBN51ftBKz7_RQ@mail.gmail.com>
Message-ID: <577F64D3.3030200@landley.net>

On 07/08/2016 02:20 AM, Alessio Igor Bogani wrote:
> Rob,
> 
> On 8 July 2016 at 07:46, Rob Landley <rob at landley.net> wrote:
>> So I'm looking at http://github.com/richfelker/musl-cross-make and Rich
> [...]
> 
> How this cope with autotools based software?

About the same as the previous toolchain? It's based on current versions
of gcc and binutils (plus mpfr and so on needed to build the current
nonsense).

It's just a compiler, it still needs command line utilities. It replaces
simple-cross-compiler.sh, and most of cross-compiler.sh and
native-compiler.sh. (Except that it doesn't build make and bash, which
native-compiler.sh does, so that probably has to move into
root-filesystem.sh.)

Rob


From piotr.karbowski at gmail.com  Sat Jul  9 10:43:12 2016
From: piotr.karbowski at gmail.com (KARBOWSKI Piotr)
Date: Sat, 9 Jul 2016 12:43:12 +0200
Subject: [Aboriginal] Hello world.
In-Reply-To: <577F3E19.4010608@landley.net>
References: <577F3E19.4010608@landley.net>
Message-ID: <6c923e34-2ac1-1969-6d11-d954da648383@gmail.com>

Hi,

On 07/08/2016 07:46 AM, Rob Landley wrote:
> So I'm looking at http://github.com/richfelker/musl-cross-make and Rich
> has got it doing native compilers now, and it theoretically supports all
> the architectures except sparc and m68k, and I still haven't fixed up my
> old toolchains (which may now be broken on _4_ architectures), so I
> should probably just switch over and focus on getting lfs-bootstrap.hdc
> updated and working and get back to toybox testing towards the 1.0 release.
>
> The downside is, this means no more precompiled binary toolchains (cross
> or native). I will not distribute GPLv3 binaries, ever. But if Rich
> wants to put them up somewhere, I'm happy to point at his server...
>
> Working otu thelayering is fun. I'm speaking at Texas LinuxFest tomorrow
> and saturday so won't get much done on it this weekend, but I think
> that's the path forward for the project in the short term.
>
> Long term I'd still like to get llvm working, but that's a bigger jump
> and still hasn't got a @*%(&# binutils replacement (which is what's
> BREAKING more often than gcc is).

If you do not wish to distribute GPLv3 binaries anywhere and you want to 
switch to this cross env, it means it will be not possible to do any 
build inside of the booted Aboriginal image. I would not like that. 
Right now I can combine the native-compiler with root-filesystem and it 
is great to be able to.

-- Piotr.


From rob at landley.net  Sun Jul 10 17:59:16 2016
From: rob at landley.net (Rob Landley)
Date: Sun, 10 Jul 2016 12:59:16 -0500
Subject: [Aboriginal] Hello world.
In-Reply-To: <6c923e34-2ac1-1969-6d11-d954da648383@gmail.com>
References: <577F3E19.4010608@landley.net>
	<6c923e34-2ac1-1969-6d11-d954da648383@gmail.com>
Message-ID: <57828CF4.9080507@landley.net>

On 07/09/2016 05:43 AM, KARBOWSKI Piotr wrote:
>> Long term I'd still like to get llvm working, but that's a bigger jump
>> and still hasn't got a @*%(&# binutils replacement (which is what's
>> BREAKING more often than gcc is).
> 
> If you do not wish to distribute GPLv3 binaries anywhere and you want to
> switch to this cross env, it means it will be not possible to do any
> build inside of the booted Aboriginal image. I would not like that.
> Right now I can combine the native-compiler with root-filesystem and it
> is great to be able to.

Well, two things:

1) You can still build them yourself from source, I just wouldn't
distribute prebuilt binaries.

2) Remember that I put the old simple-root-filesystem stuff into
initramfs last year, and the toolchain now lives by itself in its own
squashfs attached to /dev/hda (which is added to the $PATH by the init
scripts).

So if Rich wanted to host the right format of squashfs files, you could
wget that and drop it into the system-image directory and it should
"just work". (By which I mean I'd test it each release and make sure it
worked. :)

Haven't asked Rich about this yet, but worst case scenario I could make
a shell script that repackages his toolchain binaries into the squashfs
layout my thing needs, so if he does distribute any toolchain binaries
you can just grab those and run the script on them.

I'm doing my best to work around the FSF's damage. That's pretty much
been the past decade for me...

> -- Piotr.

Rob