Re: On lh-bootstrap and mkroot

From: Zach van Rijn <me_at_zv.io>
Date: Mon, 03 Sep 2018 14:34:41 -0400

On Tue, 2018-09-04 at 01:39 +0800, Casper Ti. Vector wrote:
> Although I have tried neither of lh-bootstrap or mkroot [1], I
> noticed that the toolchains of the former is nearly one
> magnitude smaller than those [2] of the latter, while both
> seem to be based on musl and statically linked.
>
> [1] <https://github.com/landley/mkroot>.
> [2] <https://b.zv.io/mcm/bin/>.
>
> Out of curiosity, may I ask what is the cause of this huge
> difference?

The ones in [3] are not stripped, and that accounts for most of
the size discrepancy. I will ask Rob to consider adding that to
his 'mcm-buildall.sh' script. I simply build/host binaries.

> Additionally, is there a possibility that the two projects
> join forces, so one becomes an add-on to the other, and
> minimising wheel reinvention?
>

As far as musl-cross-make goes, those are currently GCC 6.4.0
and there are concerns that 7.x and 8.x may have introduced
regressions. Several versions [1] [2] are of course available on
skarnet.org, so that's fine. Once musl 1.1.20 drops we'll have
m68k too. The mkroot/mcm toolchains follow directly from [4] and
at the moment we don't interfere with that process at all.

The other difference is that all of the cross toolchains in [3]
are i686 while Laurent's are x86_64 only.

So at the moment there does not appear to be any wheel problem.

I like the idea of combining efforts to offer a consistent, wide
selection of static toolchains and possibly additional tooling
such as static QEMU, MinGW, others (not just what's on my site).


ZV


[1]: https://skarnet.org/toolchains/native/

[2]: https://skarnet.org/toolchains/cross/

[3]: https://b.zv.io/mcm/bin/

[4]: https://github.com/richfelker/musl-cross-make
Received on Mon Sep 03 2018 - 18:34:41 UTC

This archive was generated by hypermail 2.3.0 : Sun May 09 2021 - 19:38:49 UTC