= Iceweasel 3.0 (firefox) and Openoffice.org 2.4 = Many people have asked for these in Debian/armel - now we have them:) = Kernels = Currently we provide kernels for 4 flavours: * iop32x - Our buildd machines * ixp4xx - The SLUG * orion5x - The popular orion-based nas devices * versatile - for qemu Now to expand our userbase, the easiest way would be to replace the versatile kernel with something else qemu supports. We want kernels that: 1) That the hardware is actually suitable for Debian - Enough storage space and RAM, reasonable CPU and a MMU. 2) Have a strong userbase - competitevly priced products available from general IT shops around the globe. And/Or that they will be available in future too. 3) Well maintained mainline kernel support. We don't have resources to maintain big out-of-free patchsets. I see atleast the following three options: * PXA2xx Mostly the zaurus. Seems there is a strong ARMEL/debian using community. The zaurii kernel support appears to be well mainlined. The main concerns are lack of new and upcoming PXA2xx hardware. PXA320 hardware on the other hand are hard to find.. * Omap Omap2 support (nokia n8x0 machine) in qemu svn. There is also work going on to make it possible to build multi-omap kernels (omap2+omap3), potentially allowing same kernel to support nokia n8x0, as well as the forthcoming pandora console and beagleboard. * Samsung s3c There is out-of-tree qemu machine support for OpenMoko. Also the OpenMoko kernel is out-of-tree, while s3c appears to be in good condition in general - 18 machines building out of s3c2410_defconfig is nice :) I believe many people would like to see openmoko supported by Debian. That's still a far cry from the 82 configurations ARM linux autobuilder[0] tries. However, we simply cannot scale with current Debian infrastructure to support each and every arm board supported by Linux. And we are not even talking about out-of-tree kernels. One option would be to start a semi-official "extra-kernels.debian.net" site that provides more kernels with less lax requirements that our official kernels have. I'd be interested in hearing feedback and ideas... = Armel installers = Reminder to everyone, the preferred way to install armel is using lenny debian-installer beta2: http://ftp.nl.debian.org/debian/dists/testing/main/installer-armel/current/images/ For NSLU2 users, use the image which includes the non-free firmware: http://www.slug-firmware.net/d-dls.php = Upgrading ARM machines to ARMEL = After Lenny, we plan to drop support for oldabi ARM port. Not to leave existing arm users in cold, there is a plan to make it as painless as possible to switch to armel. The idea is to launch a conversion script (arch-switch) from initramfs booted with a armel kernel. We use initramfs to do the heavy-lifting of finging where the / is located. Therefor the code is part. So far in my testing, upgrades work when launched from serial console, but networking up from nslu-2/thecus initramfs to allow upgrades on systems without soldered serial consoles has turned out problematic. Currently the code is available as part of rescue-initramfs: http://git.debian.org/?p=collab-maint/rescue-initramfs.git = Compiler status = We are still suffering occasionally due to http://gcc.gnu.org/PR35965 . In general gcc 4.3 seems less stable than gcc 4.2, There is a couple of ICE bugs I need to file.. = Porting help needed = Much of this stuff is from http://wiki.debian.org/ArmEabiProblems: Please feel to pick up :) dietlibc: Lots of small and compact utilities appear to choose dietlibc. I presume these would actually be usefull for many armel users. Recent mail to dietlibc mailing list[1] claims to have the patch. GNAT: It seems there is some interest in porting GNU ADA over to arm cpu's. OpenMPI: Needs porting libatomic-ops. ffcall/clisp: The alternative is to port clisp to use libffi. [0] http://armlinux.simtec.co.uk/kautobuild/2.6.26-rc7/index.html [1] http://article.gmane.org/gmane.linux.lib.dietlibc/1287 -- "rm -rf" only sounds scary if you don't have backups
Attachment:
signature.asc
Description: Digital signature