Re: Why from there? A rethink...

Riley Williams (
Fri, 15 May 1998 01:53:14 +0100 (BST)

Hi Horst.

On Thu, 14 May 1998, Horst von Brand wrote:

>> Having thought further about this problem, I see that it isn't
>> necessarily a problem in the first place since they are dealing
>> with two different scenarios that do not need to intersect:

>> 1. The full kernel source tarball DOES need to include the kernel
>> version number to prevent problems caused by old files that are
>> still lying round. This is the point that I was addressing in my
>> earlier plea.

>> 2. The kernel upgrade patch files DO NOT need to include the
>> kernel version number since they are by their nature designed
>> to be applied to an existing kernel source tree, and each
>> patch file needs to be installed in sequence, so any suchlike
>> excess files will be deleted by the correct installation of
>> the relevant patch file for the version where it was deleted.

> So, if you upgrade from 2.1.85 to 2.1.97 by patching, they are all
> called linux-2.1.85?

Nope, that's entirely up to the individual user, as clearly stated
above, but I will clarify it for you anyway: If you upgrade kernels by
patching, it is up to YOU as the user of the system to either rename
the existing directory or copy its contents to the relevant new
directory BEFORE patching it. The tools do NOT exist to make any other

> Won't do. If I have 2.1.85, and patch to 2.1.86, and then untar
> 2.1.85 again, it breaks just as before, just the other way around.
> Plus all the library includes won't find the right kernel includes
> anymore, you'll have to go around and fix every last one of them if
> you don't want random programs misteriously breaking all over the
> place because of kernel-include version mismatches.

That was part of my point in sending the original message, but I have
to concede that in the case of upgrading by patching, the facility
just isn't available to rename the directory as required...

> Sorry, there just isn't any "nice, fully automated way" of fixing
> this, operator intervention _is_ required in all possible cases.
> The one with a fixed kernel source root is the one that requieres
> least, and was selected for that reason.

Not according to Linux - as per the email he recently sent to my and
this list, his reason for selecting a fixed kernel source root is
because he can't be bothered about doing anything else, and he isn't
worried about problems caused to others by his illogical stance.

If you don't believe me, read his message for wasn't far
in front of yours in my inbox, and he included this mailing list in
his CC list...

Best wishes from Riley.

To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to