Ramdisk generation tool migration and 2.6.14-rc4 status report (Was: Bringing the 2.6.13 (and beyond) kernel to sid, solving ramdisk generations issues and stuff.)
Hello,
Here is a quick sunday evening update on the status of things concerning
the ramdisk generation tools migration :
- a fixed yaird and initrd-tools has been uploaded, but probably missed
dinstall, will be in unstable by monday's run.
- i (temporarily while Manoj is unavailable) imported kernel-package 9.008
into svn, and Thanks to Mattia Dongili was able to implement a tentative
solution. There is still a bit of work, see below.
- i created a kernel/releases/trunk/linux-2.6-2.6.14-rc4 branch, and
compiled it for powerpc with the above kernel-package. Not sure if the
kernel-img.conf parsing worked fine, but the default choice was ok.
- i tested all three patched ramdisk generation tools against a suitable set
of scenarios, and they worked just fine :
2.6 host, 2.4 kernel -> initrd-tools OK initramfs-tools NOK yaird NOK.
2.6 host, 2.6.8 kernel -> initrd-tools OK initramfs-tools NOK yaird OK.
2.6 host, 2.6.12 kernel -> initrd-tools OK initramfs-tools OK yaird OK.
2.6 host, 2.6.14-rc4 kernel -> initrd-tools NOK initramfs-tools OK yaird OK.
2.4 host, 2.4 kernel -> initrd-tools OK initramfs-tools NOK yaird NOK.
2.4 host, 2.6.8 kernel -> initrd-tools OK initramfs-tools NOK yaird NOK.
2.4 host, 2.6.12 kernel -> initrd-tools OK initramfs-tools OK yaird NOK.
2.4 host, 2.6.14-rc4 kernel -> initrd-tools NOK initramfs-tools OK yaird NOK.
- i added a conflict with non-fixed versions of the ramdisk generation tools
to linux-2.6 2.6.13 and 2.6.14-rc4 :
initramfs-tools (<< 0.31), yaird (<< 0.0.11-5), initrd-tools (<< 0.1.83)
Things still needing work :
- initramfs-tools needs to be uploaded, it seems to wait for mklibs in NEW,
but i was able to build it just fine, no idea what is going on here.
- initrd-tools and initramfs-tools may need a cleaning up of their patches
comparable to what was done with yaird. The current solution works though,
but a second set of (shell expert) eyes going over it with regard to the
yaird solution would be welcome.
- kernel-package needs to be uploaded, but before that :
o manpage and other documentation need to be changed to reflect these
modifications.
o kernel-package should add a conflict line to the generated kernel
images, this is not needed for official images, since we override
that, but it needed for self-built images. The line is :
Conflicts: initramfs-tools (<< 0.31), yaird (<< 0.0.11-5), initrd-tools (<< 0.1.83)
- in general the messages implemented in all these fixes i wrote need to be
checked for style and english :)
- we need extensive testing of this new setup in any number of cases. I did
so on powerpc, to some extent, but only with a 2.6.12 host and a
2.6.14-rc4 target. We need to see also what happens with 2.4.7 and 2.6.8
kernels, both as host and target.
Once that is done, the ramdisk migration should be complete, and we can upload
either 2.6.13-2 and/or 2.6.13+2.6.14-rc4-1 to experimental, and let people
test it some, and then, after a suitable time, preferably at around the 2.6.14
release, we can upload it to unsatble.
Like said, i created the 2.6.14-rc4 branch, to test some patches, and it went
surprisingly well for powerpc, so it may well be possible to forget about
2.6.13 and go with this one directly, especially if we want to try to upload
2.6.14 soonishly after its upstream release. it seems amd64 and powerpc did
ok, not sure about the other arches, so porters should work on it.
Also, i will not have much time for this during at least the first half of
next week, so i would appreciate if someone else would take over from this
point for the remaining issues.
Friendly,
Sven Luther
Reply to: