On 01/31/2015 06:26 AM, poma wrote:
On 30.01.2015 15:15, Tom H wrote:
> On Thu, Jan 29, 2015 at 6:41 PM, poma <pomidorabelisima(a)gmail.com> wrote:
>> On 30.01.2015 00:21, Gordon Messmer wrote:
>>> On 01/29/2015 01:55 PM, Stephen Morris wrote:
>>>> I had to delete a dozen of these directories this morning manually,
>>>> why are these not removed when the associated kernels are removed?
>>> Were the kernel modules in those directories built locally by dkms
>>> rather than owned by rpm packages? That'd do it.
>>
>> These modules do not belong
to the official kernel packages, as well
>> as your hand-built Wi-Fi module.
>
>> The point is, DKMS - Dynamic Kernel Module Support is
not a perfect
>> mechanism, therefore, it is "normal" for it to leave the old modules
>> behind.
>
>>
https://wiki.archlinux.org/index.php/Dynamic_Kernel_Module_Support#Remove...
> Shouldn't there be a prerm hook that runs "dkms remove ..."?!
And why do you ask me!? :)
The wifi driver has only just
been used since the 3.17 kernel, all he
other kernels did not have that module, the pci card I was using had
support built in to the kernel. I have always had the binary nvidia
modules installed via dnf/yum, although I have also had the akmod
packages installed as well just in case the binary modules don't match
the kernel version, and the virtualbox modules are all the binary
modules installed by dnf/yum, so I would have expected the
/lib/modules/* directories to be removed when dnf/yum did the package
removes. If dkms was used on the odd occasion for the nvidia modules, I
would have expected the dkms tree to be rebuilt to remove the entries
relevant to the kernel being removed and the modules deleted, as used to
happen in the distribution I was using prior to Fedora.