removing old Kernels

[Kostas Sfakiotakis]

Greetings Paul,

Paul Howarth wrote:
> There hasn't been a kernel-source RPM in Fedora for quite a while now;
> the kernel has a regular SRPM just like every other package.
>
>   
To be honest i was missing that part .

>>> I have not compiled a kernel in awhile. I seem to recall doing this on 
>>> the vanilla kernel source. After you compile the rpm, you could remove
>>> it with the usual rpm -e  <kernel-custom> command.
>>>
>>> You could try make rpm.
>>>       
>> No rule to make target rpm . Do you by any chance mean ,
>> rpmbuild ? rpmbuild will fail because the top level directory doesn't 
>> have the spec file needed to do the job .
>>
>> Anyway what do i have to loose if i just go ahead without using
>> an rpm and compile the kernel the old fashioned way . I don't understand 
>> the problem to the stability of the system if the rpmdatabase isn't 
>> aware of the kernel being running.
>>     
>
> If you manually remove files that are "owned" by an RPM then your system
> will be in an inconsistent state; if you use RPM (or yum) to install
> something then you should use RPM (or yum) to remove it.
>   

I guess you mean that even if  updatedb  finds that  an "owned file" 
isn't there any more the
broken dependency will remain , so the problem will remain , untill you 
uncover it and patch
the broken dependency , if things don't break apart in between .

> If you have built and installed your kernel manually then you should
> remove it manually (but don't forget that most of the kernel lives
> under /lib/modules rather than in /boot).
>   
Couldn't have said things better .


> If you are running a custom kernel that RPM does not know about then you
> may have some dependency issues with userland tools such as systemtap
> that have dependencies or conflicts with specific kernel versions.
>
>   
Well this part i was missing . I have been running custom kernels ( 
manually built ) for the last couple
of years and never came across such a problem ( or if i did , i didn't 
understood it ) . Still your note
remains entirely valid .
> Paul.
>
>