[mkroot] To the people who keeping pinging me off list...

[Rob Landley]

On 03/08/2018 02:03 PM, Alain Toussaint wrote:
> 
>> getting the "and now we build linux from scratch and beyond linux from scratch
>> natively under the result" infrastructure back in place
> 
> I'd be the one in favor of that.

Not the only one. :)

> I have been accepted as editor of BLFS and even though I'm knee
> deep into build replicability territory (build LFS --> boot LFS --> build second LFS under first LFS
> --> see if it barf)

Back in 2010 I had a 6 month contract to help port Linux to the hexagon with
qualcomm, and I wrote plumbing that built rather a lot of BLFS after the
aboriginal linux LFS build finished. (Basically it was another build control
image, but the base it ran in was the LFS build control image output.)

For a demo I got x11 running (well, the client layer, there was no graphics
hardware in the comet boards so we connected through the network to an x11
display on another machine) and got xeyes and xchess and so on running. I
remember the xterm I built wasn't the stock one but a simpler one with fewer
dependencies (might have been rxvt? These days I'd probably try
https://git.musl-libc.org/cgit/uuterm/ ) And yes, you can genericize a build
invocation for the 37 metastatic components of modern x11 and just run through
them with a for loop. I know because I did it.

Alas since I did it behind closed doors in a corporation, I didn't write about
it in my blog and didn't check it in to the open source project. I kept meaning
to redo it but never got around to it...

> so that I haven't been able to contribute to the 8.2 release, we do have request
> for 686 build and I do know that mkroot even support 486 and build a kernel for it (I myself would
> be curious to find a 486 able to boot LFS 8,2).

Sigh, I thought there was an open 486 but
https://en.wikipedia.org/wiki/Zet_(hardware) is 16 bits only.

Alas ancient technology can be a minefield long after the IP claims _should_
have expired if a current company pumps money into a legal department continuing
to spray uncertainty over it. (See also
https://www.ncbi.nlm.nih.gov/pmc/articles/PMC3614593/ and
https://atp.orangenius.com/how-mickey-mouse-keeps-changing-copyright-law/ for
how IP law is generally evil and needs to go away.)

This is also why https://en.wikipedia.org/wiki/Amber_(processor_core) goes up to
armv2a but _NOT_ armv3 (which is the oldest one compatible with the modern arm
instructions, and basically what you need to run Linux). Notice how the Amber
verilog was written in 2001 and that's long enough ago that patents issued THEN
should have expired, and yet development's basically abandoned because if they
take one more step they'll get sued, regardless of patent expiration or prior art...

> Al

Rob