[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]
Continuous Hurd cross-builds on Hydra
From: |
Ludovic Courtès |
Subject: |
Continuous Hurd cross-builds on Hydra |
Date: |
Mon, 24 May 2010 22:57:35 +0200 |
User-agent: |
Gnus/5.13 (Gnus v5.13) Emacs/23.2 (gnu/linux) |
Hello!
The GNU Hurd is now continuously cross-built on Hydra from GNU/Linux:
http://hydra.nixos.org/jobset/gnu/hurd-master
There are 3 ‘jobs’ there:
- The ‘tarball’ job, which does “make dist”.
- The ‘xbuild’ job, which cross-builds off the tarball.
- The ‘xbootstrap’ job, which does the full cross-bootstrap procedure
(i.e., what the ‘cross-gnu’ script does) using the tarball produced
above instead of the Hurd snapshot currently in Nixpkgs.
Getting ‘make dist’ to work was surprisingly tedious, presumably because
the project has a long tradition of not publishing source releases. ;-)
There’s a series of patches at
<https://svn.nixos.org/repos/nix/hydra-config/gnu/trunk/hurd/>. Most of
them should be obvious. I can provide proper ChangeLog entries so that
they can go in (I’ve already assigned copyright), and I’m of course open
to discussions.
The ‘xbootstrap’ job is the most interesting: it reuses the whole cross
toolchain packaging from Nixpkgs and simply overrides the source of the
Hurd package. This can be seen from the build steps and run-time
dependency graph of a build:
http://hydra.nixos.org/build/416968#tabs-buildsteps
http://hydra.nixos.org/build/416968/runtime-deps [*]
Eventually, it’d be nice to have it override Mach, MIG, and libpthread
as well, so we get complete regression testing for the cross toolchain.
Similarly, glibcs (sourceware.org and sv.gnu.org repositories) are
(about to be) continuously built:
http://hydra.nixos.org/jobset/gnu/glibc-master (always fails to build)
http://hydra.nixos.org/jobset/gnu/glibc-hurd (‘make dist’ fails)
Hydra can send email notifications when the status of a build changes
(e.g., from ‘succeeded’ to ‘failed’.) I can add whoever feels like
receiving Hydra news.
Thanks,
Ludo’.
[*] If you look carefully, there’s a small glitch: two glibcs, instead
of one, are built; one is built with the old Hurd headers and is
used only to build ‘hurd-minimal’ (which contains only libihash).
I think it can’t hurt, but it’s unpleasant to the eye. ;-)
pgpJiYVpfJna8.pgp
Description: PGP signature
- Continuous Hurd cross-builds on Hydra,
Ludovic Courtès <=