William L. Thomson Jr. | 1 Aug 2005 01:14
Favicon

Re: Re: RHEL4 kernel compile -- why make "Mr. Proper" exists ...

On Sun, 2005-07-31 at 17:22 -0500, Bryan J. Smith wrote:
> 
> The "big problem" is that people used to use the same tree to build more
> than one kernel,

I ran into with a fresh install of the kernel src rpm. I had not built a
kernel using those sources before and it failed on the first attempt.

FYI my old method was just after installing kernel source. I would tar
it up. So I always had a copy of the original pristine source. Once I
made a kernel, if it was not for that machine, and I needed to make
another. I would wipe out the kernel tree I just compiled in, and untar
the backup of the original. make mrproper might have saved me from that,
but it was not a big deal.

> I was just saying that Make "Mr. Proper" ("Mr. Clean" in the US) was a
> Linus invention to combat various details such as "leftover" .config
> files and other differences before make config/dep.  It's not a Red
> Hat'ism at all.

True, but till I had issues with RH provide kernel sources. I never had
to use make mrproper before that. If I was making repeat builds for the
same machine. I would just use make clean. Otherwise I would wipe out
the entire tree and proceed as I described above.

> I don't mean to be "abrasive," but by your own admission, you claimed to
> have not used anything but Red Hat back then.  I know many things come
> off as Red Hat-centric, but they are typical in packages distros, or
> even some source distros.

(Continue reading)

Bryan J. Smith | 1 Aug 2005 01:38
Picon

Re: RHEL4 kernel compile -- why make "Mr. Proper" exists ...

On Sun, 2005-07-31 at 19:14 -0400, William L. Thomson Jr. wrote:
> I ran into with a fresh install of the kernel src rpm.

kernel-source.i386.rpm (/usr/src/linux) != kernel.src.rpm.
The two are different.

> True, but till I had issues with RH provide kernel sources.

The kernel-source.i386.rpm, yes.
I had to run "make mrproper" on those too.

Once I stopped using kernel-source.i386.rpm, and starting using the
kernel.src.rpm (circa 2001) per Red Hat recommendations, that solve the
problem.  I did so once I started maintaining formal kernel modules in
RPM format.  It's really the only way to do it.

> As far as a full on distro mainly RH. I did also use the LRP (and still
> do). I was making kernels for it via sources from the LRP, and from
> kernel.org.

But what were you building the LRP on?

> So I have extensive time with building a variety of kernels.

But only one main distro, that was my point.

I wasn't saying you didn't have experience building kernels, you just
had experience only on Red Hat (including Red Hat forks like Cobalt) so
that's why you attributed it to Red Hat.

(Continue reading)

Bryan J. Smith | 1 Aug 2005 01:50
Picon

Re: RHEL4 kernel compile -- original 2 William comments revisited ...

On Sun, 2005-07-31 at 11:49 -0400, William L. Thomson Jr. wrote:
> I did not see it mentioned in the thread, but typically to build a RH
> kernel there is one additional step.
> make mrproper

On Sun, 2005-07-31 at 15:30 -0400, William L. Thomson Jr. wrote:
> It can't really hurt, and without it can cause kernel compilation
> issues.

Do _not_ do this when you prep from SRPM.
It can have unforseen issues.
Always re-prep the source tree (rpmbuild -bp kernel-2.6.spec).

If you read up on "make mrproper," which comes from "Mr. Proper," the
common, international brand for "Mr. Clean," it's named so for a reason.
You want to avoid doing it if you can on _any_ distro.

"make mrproper," like "Mr. Clean," kills everything.
It is not like "make clean".

On Sun, 2005-07-31 at 11:49 -0400, William L. Thomson Jr. wrote:
> From memory I think I did that before configuring the kernel, but I do
> not think it matters. It does need to be done before you do

Of course it matters because "make mrproper" _removes_ the .config file
(and all localized config/Makefiles).  If you run it, you'll not only
have to re-configure _everything_, but possibly re-prep the source tree.

Because of varying issues, Red Hat started formalizing the kernel build
process in Red Hat Linux 6.x from SRPM.  In the Red Hat Linux 7 time-
(Continue reading)

William L. Thomson Jr. | 1 Aug 2005 02:16
Favicon

Re: Re: RHEL4 kernel compile -- why make "Mr. Proper" exists ...

On Sun, 2005-07-31 at 18:38 -0500, Bryan J. Smith wrote:
>
> kernel-source.i386.rpm (/usr/src/linux) != kernel.src.rpm.
> The two are different.

Yes, and kernel-source.i386.rpm != vanilla kernel either.

> Once I stopped using kernel-source.i386.rpm, and starting using the
> kernel.src.rpm (circa 2001) per Red Hat recommendations, that solve the
> problem.  I did so once I started maintaining formal kernel modules in
> RPM format.  It's really the only way to do it.

RHEL 3.0 docs. #2 mentions make mrproper
http://www.redhat.com/docs/manuals/enterprise/RHEL-3-Manual/sysadmin-guide/s1-custom-kernel-modularized.html

> > As far as a full on distro mainly RH. I did also use the LRP (and still
> > do). I was making kernels for it via sources from the LRP, and from
> > kernel.org.
> 
> But what were you building the LRP on?

A variety of things. Cobalts, RH, and now Gentoo. Not really system
dependent other than say gcc, and possible a few others.

> > So I have extensive time with building a variety of kernels.
> 
> But only one main distro, that was my point.

Debian as well, just not as long as RH. I tried migration to it when I
left RH, and had issues. Thus Gentoo. Also I had buddies that were into
(Continue reading)

Bryan J. Smith | 1 Aug 2005 02:52
Picon

Re: RHEL4 kernel compile -- why make "Mr. Proper" exists ...

On Sun, 2005-07-31 at 20:16 -0400, William L. Thomson Jr. wrote:
> Yes, and kernel-source.i386.rpm != vanilla kernel either.

But kernel-source.i386.rpm was used to provide a near-equivalent in
build approach/steps to kernel.org kernels, by plopping out a kernel
source tree in /usr/src/linux.

Red Hat's kernel.src.rpm does not, and has not since Red Hat Linux 6.1.
Debian's kernel building facilities are similar in approach.

Just because you _can_ get a kernel tree in /usr/src/linux and by-pass
the package manager does not mean it's the recommended way.  For those
of us using the formalized Red Hat and Debian kernel processes, there
are other details involved.

> RHEL 3.0 docs. #2 mentions make mrproper
> http://www.redhat.com/docs/manuals/enterprise/RHEL-3-Manual/sysadmin-guide/s1-custom-kernel-modularized.html

Yes, it's the "slam me" solution introduced by Linus that solves the
problem on a lot of distros that have varying codebases.  It has not
been recommended for a _long_time_ in Red Hat because of this.

However, Red Hat maintained it for legacy compatibility and its
inclusion in the documentation is more for legacy purposes.  People
still scream about the lack of kernel-source.i386.rpm to this day
because they assume they can just have "free will" with /usr/src/linux.

> A variety of things. Cobalts, RH, and now Gentoo.

So you _were_solely_ using Red Hat and _direct_ Red Hat forks.
(Continue reading)

Bryan J. Smith | 1 Aug 2005 03:10
Picon

Re: RHEL4 kernel compile -- where the SPEC file may call mrproper ...

On Sun, 2005-07-31 at 19:52 -0500, Bryan J. Smith wrote:
> Which is why your _other_ reference (in the same set):  
>   http://www.mjmwired.net/resources/mjm-kernel-fc4.shtml  
> Which is from MJM himself -- one of the _core_ kernel developers --
> _explicitly_ differentiates between the using the kernel.src.rpm and
> kernel.org and _only_ says to use "make mrproper" in the latter.

Looking at the latest kernel-2.6.12-*.src.rpm (2.6.12 w/some 2.6.13
patches) SPEC file on Fedora Development, it does use "make mrproper"
but it is important to know _when_.  It does it _after_ modifying the
EXTRAVERIONS in the Makefile (among other source tree changes) but
_before_ running the config.

In some other kernel.src.rpm (notably 2.4), I have seen it before the
EXTRAVERSIONS tag change in the Makefile (but after other
configuration).  And in other kernels still, I haven't seen it.

It's use is very "abrasive" so I avoid it unless I run into issues.
Especially since the kernel.srp.rpm prep really does build a near-
pristine kernel, although it all depends.

--

-- 
Bryan J. Smith   b.j.smith@...      http://thebs413.blogspot.com
--------------------------------------------------------------------- 
It is mathematically impossible for someone who makes more than you
to be anything but richer than you.  Any tax rate that penalizes them
will also penalize you similarly (to those below you, and then below
them).  This is why someone who makes more than you always gets at
least the same, if not a bigger, tax cut.  Otherwise is impossible.

(Continue reading)

William L. Thomson Jr. | 1 Aug 2005 03:22
Favicon

Re: Re: RHEL4 kernel compile -- why make "Mr. Proper" exists ...

On Sun, 2005-07-31 at 19:52 -0500, Bryan J. Smith wrote:
>
> Just because you _can_ get a kernel tree in /usr/src/linux and by-pass
> the package manager does not mean it's the recommended way. 

A RPM  put the sources there for me to use, and it was the documented
way. Still looking for official user docs on another way.
http://www.redhat.com/docs/manuals/linux/RHL-7.3-Manual/custom-guide/ch-custom-kernel.html
http://www.redhat.com/docs/manuals/linux/RHL-8.0-Manual/custom-guide/ch-custom-kernel.html
http://www.redhat.com/docs/manuals/linux/RHL-9-Manual/custom-guide/s1-custom-kernel-modularized.html
http://www.redhat.com/docs/manuals/enterprise/RHEL-3-Manual/sysadmin-guide/s1-custom-kernel-modularized.html

> So you _were_solely_ using Red Hat and _direct_ Red Hat forks.

And Debian, not so much me, but I made custom kernels for my buddies
running Debian. Not sure about where the sources came from. They were
already there. I just did the configuring and make stuff.

> > Debian as well, just not as long as RH. I tried migration to it when I
> > left RH, and had issues.
> 
> This does not bode well for your comment of "extensive experience" in
> building kernel distros outside of Red Hat (prior to Gentoo).

I have built probably 20-30 kernels for Debian systems, before my
attempt at going to Debian. I guess that's not good enough? Now RH
kernels. I could not even guestimate how many of those I have built.
Must less now on gentoo. Heck I made 3 today for my development network.
Still have to do one for my laptop. And I did like 20 last weekend. Of
course all Gentoo. But making custom kernels is a trivial thing for me
(Continue reading)

William L. Thomson Jr. | 1 Aug 2005 03:27
Favicon

Re: Re: RHEL4 kernel compile -- where the SPEC file may call mrproper ...

On Sun, 2005-07-31 at 20:10 -0500, Bryan J. Smith wrote:
>
> Looking at the latest kernel-2.6.12-*.src.rpm (2.6.12 w/some 2.6.13
> patches) SPEC file on Fedora Development, it does use "make mrproper"
> but it is important to know _when_.  It does it _after_ modifying the
> EXTRAVERIONS in the Makefile (among other source tree changes) but
> _before_ running the config.

So if it's still in use, was I so off base by mentioning it. Much less
all the subsequent posts and etc. It did not need to be blown out of
proportion.

> In some other kernel.src.rpm (notably 2.4), I have seen it before the
> EXTRAVERSIONS tag change in the Makefile (but after other
> configuration).  And in other kernels still, I haven't seen it.

Most times when making a custom kernel on RH, one will set their own
custom versions. I used to but as I got lazier I really did not care.
It's only a name, not anything that is important or used.

> It's use is very "abrasive" so I avoid it unless I run into issues.

make mrproper is not abrasive. It does not do major damage or any at
all. As I have proved in another post.

Please in the future can we keep these things small. There was no need
for all of this, nor has there ever been. I have posted links to
relevant documentation and etc. Not to mention gone through the entire
aspects of make mrproper. From the help, to the actual commands in the
make file, to actually running it, and showing the results of what make
(Continue reading)

Bryan J. Smith | 1 Aug 2005 04:17
Picon

Re: RHEL4 kernel compile -- all I can offer is "best practices"

On Sun, 2005-07-31 at 21:27 -0400, William L. Thomson Jr. wrote:
> So if it's still in use, was I so off base by mentioning it. Much less
> all the subsequent posts and etc. It did not need to be blown out of
> proportion.

What I blew out of proportion were your 2 original statements:
From:  http://mailman.jaxlug.org/pipermail/jaxlug-list/Week-of-Mon-20050725/015087.htm  

 "I did not see it mentioned in the thread, but typically to build a RH
  kernel there is one additional step.

  make mrproper

  From memory I think I did that before configuring the kernel, but I do
  not think it matters."

It is _not_ "typical" to run "make mrproper" arbitrarily and it _does_
matter when you run it.  Everything since then has been postering,
backtracking and, most problematic, a quest to "defend yourself" against
things I didn't say largely because you don't understand how
"kernel.src.rpm" differs from /usr/src/linux tree approaches.

> Most times when making a custom kernel on RH, one will set their own
> custom versions. I used to but as I got lazier I really did not care.
> It's only a name, not anything that is important or used.

Unless, of course, the name is already in use.  ;->

> make mrproper is not abrasive. It does not do major damage or any at
> all. As I have proved in another post.
(Continue reading)

Bryan J. Smith | 1 Aug 2005 04:30
Picon

Re: RHEL4 kernel compile -- the agenda exposes itself (and why I can't continue on this list)

On Sun, 2005-07-31 at 21:22 -0400, William L. Thomson Jr. wrote:
> Personally the more distros vary from how kernrel.org or how the core
> kernel developers are building kernels. The less I am going to even
> consider or like the other distros. Doing to much their way instead of a
> standard way.

And now we get to the "real, underlying theme" and ultimate "agenda"
behind your participation in this thread.  At this point, I'm going to
have to bow out of this list as a whole, because there is no point in me
helping FC/RHEL users here.  Every thread breaks into this, and when I
try to explain things with detailed, technical history to show why, it
is under0appreciated.

Several people have complained I do not respect the opinions of others.

When someone is a peer with _current_ knowledge on a subject, I always
do.  I honestly do.  And even if not, I try to respect and even educate.
I'm sure I'm not always the most tactful at times, and I could
definitely use some course in communicating via e-mail (which I don't
use at work for a reason).

But when I'm trying to offer "best practices" for a distribution I am
_current_ with, offering the best technical details I can, and I'm
constantly crossed, questioned and continuously responded to (it takes
at least 2 people for that), only to be "stuck with the check," I
honestly don't know what to say.

Especially when the original recommendations are not only _not_
applicable, but "made on a whim" without attention to technical
accuracy.  That should be a "wake up call" to the person who made them
(Continue reading)


Gmane