On Sun, 2008-12-21 at 09:48 +0100, Tobias Powalowski wrote:
Am Dienstag 16 Dezember 2008 schrieb Jan de Groot:
On Mon, 2008-12-15 at 14:42 -0600, Aaron Griffin wrote:
On Mon, Dec 15, 2008 at 2:41 PM, RedShift <redshift@pandora.be> wrote:
Why not solve it the right way, by creating a correct package?
I think you're missing why this is impossible. Anyone that has m-i-t 3.5 on their system now has the .bin files that are owned by no package. pacman correctly does not overwrite these files.
What "correct package" are you suggesting?
As these files are generated by module-init-tools whenever it is run for your kernel, why not solve this using post_* functions? Whenever the kernel gets installed/upgraded, depmod is run from post_install/upgrade. Whenever the kernel is removed, the files created by depmod for this kernel get removed from post_remove. Shouldn't be hard to do, the result is the same and no file conflicts will exist. Well depmod is already called in .install file. Why to make a remove in PKGBUILD and create untraced files on the system? This bug affects only people who run testing and installed new external modules. This should be not a big part of the community, mostly noone runs testing ;)
BTW: Please remove the -v flag from depmod in kernel26.install, it's stupid to make depmod print verbose messages while redirecting output to /dev/null. fixed
Before releasing .10 kernel I would like to have a clear answer on how to handle this. My suggestion would be the --force option, the other thing is a workaround which is imho not needed here.
If someone runs testing, they better be able to figure out "pacman -Sf kernel26", especially if we make posts about it. I'm anxious for 2.6.27.10, it's supposed to fix my hibernate regression. Dale