Weekly Edition
Return to the Distributions page
| ||||||||||||||||
|
| ||||||||||||||||
Bits from the Debian GNU/Hurd porters
Hello, it has been more than three years since the last "Bits from the Debian GNU/Hurd porters"[1], high time for an update on the port. * Snapshot releases Three new snapshot releases have been done by Philip Charles, K14, K15 (which was only done as an updated mini CD-ISO, not a full snapshot), and K16. K16 has been released[2] on December 18th, 2007 featuring four CDs or two DVDs. Additionally, it also features a ready-to-go qemu-image[3] for the first time. K16 was also the first snapshot which included TLS (Thread Local Storage), a requirement for modern glibcs. New ported packages include Qt3, Qt4, SDL and Emacs22. * Base and toolchain status Currently, most base packages are current, with the notable exception of util-linux, which has been a big problem over the last years. However, Samuel Thibault got all outstanding issues of util-linux applied upstream so the version in experimental is mostly working. The toolchain is in pretty good shape as well since TLS support got implemented; we are using the current glibc, binutils and gcc Debian packages unmodified. * Xen support Besides qemu, which can be very slow to run, a Xen DomU port for GNU Mach has been made available by Samuel Thibault. It requires a non-PAE hypervisor and some minor manual tweaking, but is otherwise quite functional and stable already, see its wiki page[4] for further information. This will make people running the Hurd less dependent on specific hardware, as a lot of newer computers do not work with the underlying GNU Mach kernel anymore. * Autobuilder availability and archive coverage improved The percentage of packages built for Debian GNU/Hurd has improved from 40% to now nearly 60%[5] since the last Bits from the porters. Further, the backlog of outdated packages has been greatly reduced. This is due to the addition of two[6][7] Xen autobuilders earlier this year, which made the hurd-i386 autobuilders far more robust and fault-tolerant as they not need local admin attention anymore in case of problems with the GNU/Hurd guests. The remaining 40% of packages are either waiting for other packages to become available (see [8] for a (big) graph of those relationships) or are failing for some reason[9]; a complete list of build failures can be found at [10]. * Developer machine We are currently working on getting a general DD-accessible porter box setup. In the meantime, interested people can contact hurd-shell-account@gnu.org to get an account on one of the publically accessible (Debian) GNU/Hurd developer machines. For further details, see [11]. * Summer of Code 2008 This year, the GNU Hurd participated as its own organization at Google's Summer of Code, thanks to the coordination done by Olaf Buddenhagen[12]. All of the 5 projects were carried out quite successfully. The most practically relevant project for Debian GNU/Hurd was the implementation of a procfs translator[13] by Madhusudan C.S., which provides a traditional Unix-style /proc file system and the subsequent porting of the procps package, so utilities like pgrep etc. will be available after lenny, and procps Build-Depends no longer need to be special-cased on hurd-i386. Other GSoC projects were lisp bindings by Flavio Cruz, better system debugging and tracing by Andrei Barbu, namespace-based translator selection by Sergiu Ivanov and network virtualization by Zheng Da. More information on the details and outcome of those projects can be found on the wiki[14]. * Still no debian-installer Unfortunately, the Debian GNU/Hurd port still lacks d-i support. On the other hand, debootstrap now mostly works, even to cross-debootstrap a hurd-i386 installation from GNU/Linux, if one works around bug #498731. A relatively easy solution could be to use the GNU/Linux d-i to cross-install and setup a Debian GNU/Hurd system. People who have experience in d-i and possibly Debian GNU/Hurd are more than welcome to contact us at debian-hurd@lists.debian.org. for the Debian GNU/Hurd porters, Michael Banck [1] http://lists.debian.org/debian-devel-announce/2005/07/msg... [2] http://kerneltrap.org/node/15770 [3] http://ftp.debian-ports.org/debian-cd/K16/debian-hurd-k16... [4] http://www.bddebian.com/~wiki/microkernel/mach/gnumach/po... [5] http://buildd.debian-ports.org/stats/ [6] http://buildd.net/cgi/hostpackages.cgi?unstable_arch=hurd... [7] http://buildd.net/cgi/hostpackages.cgi?unstable_arch=hurd... [8] http://dept-info.labri.fr/~thibault/tmp/graph-radial.pdf [9] http://www.bddebian.com/~wiki/unsorted/PortingIssues/ [10] http://unstable.buildd.net/buildd/hurd-i386_Failed.html [11] http://www.bddebian.com/~wiki/public_hurd_boxen/ [12] http://code.google.com/soc/2008/hurd/about.html [13] http://packages.qa.debian.org/h/hurd/news/20080903T160206... [14] http://www.bddebian.com/~wiki/community/gsoc/ (Log in to post comments)
KVM? Posted Sep 15, 2008 17:45 UTC (Mon) by salimma (subscriber, #34460) [Link]
Considering a lot of Linux distributions are shifting to KVM (Red Hat/Fedora and Ubuntu, for instance), it's quite surprising that it did not merit any mention. Presumably the normal boot image works fine under it -- KVM happily runs newer BSD variants such as Dragonfly, which stomps VMware and VirtualBox -- but it'd be nice to know.
KVM? Posted Sep 15, 2008 17:47 UTC (Mon) by mbanck (subscriber, #9035) [Link]
I believe it works fine in general - though somebody recently reported that latest Linux KVM do not boot Mach anymore. I don't think anybody looked into this yet.
Michael
Bits from the Debian GNU/Hurd porters Posted Sep 15, 2008 19:16 UTC (Mon) by vaib (guest, #48292) [Link]
I always wonder about Hurd. I can see pretty much nothing in its changelog, on the Hurd mailing list I saw hardly 1-2 mail once a month. Is there any development going on Hurd at all? Does it lack leadership or its just people are not interested developing it as its double effort along with linux kernel.
Bits from the Debian GNU/Hurd porters Posted Sep 15, 2008 19:38 UTC (Mon) by sbergman27 (subscriber, #10767) [Link]
I'm not sure why anyone *would* want to use or develop a dead OS that was going to be the next big thing in 1990, other than for personal educational purposes. Its one possibly attractive feature was being a Free microkernel posix-like OS. Or at least it was supposed to be the only serious one. But now there's Minix 3, which may not be the the talk of the town, but at least is breathing.
Can Minix3 run GNU? Posted Sep 16, 2008 7:46 UTC (Tue) by coriordan (guest, #7544) [Link]
Hurd never had simply the academic goal to develop a microkernel. It had the practical goal to develop a microkernel for the Unix-like GNU operating system.
Minix3 exists, and that's probably great for enthusiasts of the microkernel architecture, but AFAIK it can't be used as the kernel for a Unix-like OS such as the GNU system. The Linux kernel can do this, as can the kernels of FreeBSD, NetBSD, and Solaris (but they're not microkernels, of course). The Hurd can already be used as the kernel for a GNU system, but it's capabilities are limited.
So Minix and Hurd have different goals.
Can Minix3 run GNU? Posted Sep 16, 2008 16:19 UTC (Tue) by sbergman27 (subscriber, #10767) [Link]
"""
So Minix and Hurd have different goals. """
Perhaps. Insofar as Minix3 ( http://www.minix3.org/ ) actually and currently *has* goals. Minix3 is only loosely based upon the version 1 and 2 releases, which were, indeed, designed for educational use. The Minix3 project intends, and is making concrete progress towards, creating a usable, real world OS ala Linux designed around a microkernel. While I'm not a huge fan of Minix3 or of microkernels, Minix3 has a credibility which HURD currently lacks. I'm not sure what particular significance "running GNU" actually has in that context. It's a POSIX-like OS, runs X, and a growing list of officially supported applications ( http://www.minix3.org/software/ ).
Considering that most of the Minix 1 and 2 code is of little or no relevance in light of the new purpose of Minix3, what Minix3 has accomplished it has done in 3 years. What little HURD has accomplished has taken 18 years.
It looks to me like "stagnant vs vibrant" is a more significant difference between the two projects than are "different goals" or "running GNU". Who want's to work on something that is going nowhere, regardless of what its goals and aspirations might, at one time, have been?
Bits from the Debian GNU/Hurd porters Posted Sep 15, 2008 20:50 UTC (Mon) by mbanck (subscriber, #9035) [Link]
The Hurd project uses a one-ChangeLog-per-directory approach; the top-level ChangeLog file mostly has build system changes.
That said, not that much code has been committed to the HEAD branch lately; however the Summer of code branches did see quite some activity naturally. Also the GNU Mach Xen-branch has seen lots of changes.
Michael
Bits from the Debian GNU/Hurd porters Posted Sep 15, 2008 20:24 UTC (Mon) by flewellyn (subscriber, #5047) [Link]
I thought they were porting from Mach to L4? What happened to that?
Bits from the Debian GNU/Hurd porters Posted Sep 15, 2008 20:37 UTC (Mon) by drag (subscriber, #31333) [Link]
Last time I checked the development effort dwindled down to just a bunch of die-hard people that love nothing more then having endless discussions about architecture.
-------------------------
For the MACH-based kernel, which is the one that actually somewhat worts, they've stopped all development completely. This is the kernel being discussed by the article.
The L4 kernel is what they are focusing on and as far as I know it's not remotely close to being functional.
Bits from the Debian GNU/Hurd porters Posted Sep 15, 2008 20:57 UTC (Mon) by flewellyn (subscriber, #5047) [Link]
Huh. So, basically, Hurd continues as it always has: spinning its wheels while accomplishing nothing much?
Bits from the Debian GNU/Hurd porters Posted Sep 15, 2008 23:45 UTC (Mon) by bronson (subscriber, #4806) [Link]
Ya, but it can spin each wheel individually!
Bits from the Debian GNU/Hurd porters Posted Sep 16, 2008 4:49 UTC (Tue) by sbergman27 (subscriber, #10767) [Link]
Keeping an audience entertained for years on end can hardly be considered "nothing much" as any veteran stage comedian will attest. Carol Burnett decided to gracefully end her show after 11 years. HURD's been at it for 18 and will no doubt continue until the gong or the hook is employed.
Bits from the Debian GNU/Hurd porters Posted Oct 4, 2008 23:07 UTC (Sat) by dirtyepic (subscriber, #30178) [Link]
Yes, congrats, it's the Carrot Top of the kernel world.
Bits from the Debian GNU/Hurd porters Posted Sep 16, 2008 18:07 UTC (Tue) by azrael (guest, #53640) [Link]
Am I the only one that reading this immediately have thought about Duke Nukem Forever changing engines?
Bits from the Debian GNU/Hurd porters Posted Sep 16, 2008 18:18 UTC (Tue) by flewellyn (subscriber, #5047) [Link]
Probably not.
The sad thing is, there IS interesting development happening in microkernel designs in the open source world. Coyotos, for instance. I don't get why Hurd is unable to make any progress.
Bits from the Debian GNU/Hurd porters Posted Sep 16, 2008 20:18 UTC (Tue) by nix (subscriber, #2304) [Link]
Duke Nukem Forever is much younger than the Hurd. :/
|
|
|||||||||||||||