Re: Newbie question about kernels
From: Nix (nix-razor-pit_at_esperi.org.uk)
Date: 09/21/05
- Next message: DazMe: "how slow is ioctl?"
- Previous message: Nix: "Re: Newbie question about kernels"
- In reply to: Peter T. Breuer: "Re: Newbie question about kernels"
- Next in thread: Peter T. Breuer: "Re: Newbie question about kernels"
- Reply: Peter T. Breuer: "Re: Newbie question about kernels"
- Messages sorted by: [ date ] [ thread ] [ subject ] [ author ]
Date: Wed, 21 Sep 2005 11:44:23 +0100
On Tue, 20 Sep 2005, Peter T. Breuer said:
> Nix <nix-razor-pit@esperi.org.uk> wrote:
>> On Wed, 31 Aug 2005, Peter T. Breuer wrote:
>>> Marten Kemp <martendespamkemp@thisplanet-link.net> wrote:
>>>> Hmm. I'm a newbie at this (as mentioned in the message title)
>>>> so I don't have much invested in the current project. I'm using
>>>> Debian 3.1 which is IIRC based on 2.4.27. Would it be of any
>>>
>>> It is not "based on it". Kernels and systems are always INDEPENDENT.
>
>> This is incorrect: glibc can be told to refuse to run below specific
>
> Nonsense. I am running on 2.0 kernels still! (THIS kernel on THIS machine
> is a much loved 2.2.15). I haven't had any difficulties!
Sheesh. Learn to read already. `Can be told to refuse' does not equal
`must be told to refuse'.
(But if you have an NPTL-only glibc, the system *will* fail to boot when
running any kernel below 2.5.something. This includes small unimportant
systems like FC4.)
>> kernel versions, new tools are needed with newer kernels (and sometimes,
>
> Sure, fine, so what?
So they are not independent; they are related, although the relationship
is not always tight.
>> older tools with older kernels), and some programs may require features
>> present only in newer kernels.
>
> Tough. Teach their authors how to program properly then.
Really? I'll let you teach the iproute2 authors how to `program' ip(8)
so that it works on 2.0 kernels; this will be interesting to watch, as
the entire program's raison d'etre is to talk to the kernel via a mechanism
that didn't exist in 2.0 in order to manipulate networking functionality
that *also* didn't exist in 2.0.
Likewise the LVM hackers should have learned how to program properly,
and those idiots who made the mistake of using POSIX features like
clock_*() or anonymous mmap should have learned to program properly
before daring to use features in POSIX.
How far back should they restrict themselves before you restrain your
contempt? 2.0 kernels? 1.0 kernels? Is features available on the
bleeding edge in 2000 enough? How about 1987? (Restrict yourself to
shell features available in 1987 and your shell scripts will break
with Solaris /bin/sh.)
I think you're arguing a case which is too inconsistent to even be
wrong.
>>> You can run any kernel with any system - it is a question of your
>>> choice.
>
>> This is also incorrect. Try sticking a 2.6 kernel on an old libc5 system
>> originally built around 2.0, say, without changing anything; not
>
> That sounds like my system! I run libc5.
... with a 2.2 kernel, you said. That will work, and doesn't contradict
the statement I just made.
Of *course* libc5 and 2.2 will work reasonably well (although 2.2 kernel
features will be partially unavailable).
With 2.6 kernels, the disparity is getting *large*, and a *lot* will be
missing. Really an awful lot. (Although if you don't try to compile a
program that uses anything standardized (or added to the kernel) after
1998, you might be in luck. You'd have no chance compiling something
like KDE or GNOME or X.org or, for that matter, anything requiring a
recent GCC, though, so any programs using modern C++ are out of the
question. glibc1 support was removed from GCC over a year ago.)
> oboe:/usr/oboe/ptb% ll /lib/libc.*
> lrwxrwxrwx 1 root root 9 Sep 6 07:47 /lib/libc.so -> libc.so.5*
> lrwxrwxrwx 1 root root 13 Sep 6 07:47 /lib/libc.so.4 -> libc.so.4.7.2*
> -rwxr-xr-x 1 root root 634880 Apr 29 1995 /lib/libc.so.4.7.2*
> lrwxrwxrwx 1 root root 14 Sep 6 07:47 /lib/libc.so.5 -> libc.so.5.4.38*
> -rwxr-xr-x 1 root root 579164 Sep 9 1997 /lib/libc.so.5.4.38*
If you're still using those as your primary libcs (and it seems like you
are), then that system is so obsolete it's not even funny.
>> everything will work. (Among other things, modules will break, and
>
> If one sticks in a new kernel WITH modules, one adds the new module
> tools.
If they compile with libc5, it is pure luck. Since libc5 lacks any
definitions for new kernel features since about 1998, it'll be chancy.
(I can't test it: I don't run libcs with no security support since
1998. Neither does anyone sane: not on a network-connected system,
anyway, firewall-protected or no.)
> Please stop the nonsensical wibbling NOW. It's annoying.
I'd be happy to if you'd stop asserting things that are easily proved
wrong.
-- `One cannot, after all, be expected to read every single word of a book whose author one wishes to insult.' --- Richard Dawkins
- Next message: DazMe: "how slow is ioctl?"
- Previous message: Nix: "Re: Newbie question about kernels"
- In reply to: Peter T. Breuer: "Re: Newbie question about kernels"
- Next in thread: Peter T. Breuer: "Re: Newbie question about kernels"
- Reply: Peter T. Breuer: "Re: Newbie question about kernels"
- Messages sorted by: [ date ] [ thread ] [ subject ] [ author ]
Relevant Pages
|
