Re: `guix build hello' now succeeds on the Hurd

[Ludovic Courtès]

Hello!

Jan Nieuwenhuizen <address@hidden> skribis:

> The situation on the Hurd starts to look pretty good
>
>     janneke@debian:~/src/guix$ ./pre-inst-env guix build hello --no-offload
>     /gnu/store/a2sylb94rm1b6qxcp5mqvgiyx9szipz7-hello-2.10
>     janneke@debian:~/src/guix$ 
> /gnu/store/a2sylb94rm1b6qxcp5mqvgiyx9szipz7-hello-2.10/bin/hello
>     Hello, world!
>
> \o/

Woohoo!  Congrats!

How do you run guix-daemon?  (In the future it’d be great to perhaps
implement Linux namespaces on the Hurd in libc.)

> It has some 20 odd patches.  I chose to create workarounds to "get early
> success" rather than doing everything right.  While I worked on forward
> porting glibc patches, I only took the minimal set that I needed.  Also,
> I reverted to make 4.1 instead of debugging why make 4.3 fails (for
> now).

OK.

> Most controversial/problematic is the need to use fairly recent gnumach
> and hurd sources.

That’s OK IMO.

> What could be the next step?

Merging what you have—the earlier the better.  :-)

> Shall I push this to savannah as `wip-hurd' (possibly save wip-hurd->
> `wip-hurd-old?);

Yup, sounds like a plan.

> I could also rewrite wip-hurd-bootstrap?

Dunno!

To me, the difficult bit with porting and bootstrapping work is making
sure that bootstrap.scm/commencement.scm/base.scm/cross-base.scm remain
maintainable.  All this complexity adds up so we must spend time trying
to, for instance, minimize variation across platforms/OSes.  Every line
of code and above all every conditional avoided in these files is a win
in the not-so-long term.  That’d be my guideline as we merge it.  :-)

Anyhow, thumbs up!  I’m looking forward to merging it and having it
built on CI (we could offload to a Debian VM!)!

Ludo’.