Skip to contentSkip to conversation

kbuild [which is not only CML2]

DISCUSSION Giacomo A. Catenazzi 4 participants 5 messages Feb 16, 2002

lore
Share & tools
lore
#0 lore raw

Hello

I have some comment/explications about the thread about kbuild.

1-
The discussion was also on lkml, ESR asked to kbuild people to give some comments, using also the feed-back of lkml.
As you noticed, in lkml the discussion went into flames, fogetting the important points.
[As this flame is going forgetting kbuild-2.5..]]

2-
The comments in kbuild list are accessible to all. You should read them (hmm, really there are only one full comment [my comment] + discution/flame]

3-
The importants points are:
CML2 and kbuild-2.5 are two different projects, which don't depend each other.
The need of the two project is also demostrated by the other never finished similar project (mconfig, kernconfig, mcml2, cml2config, dancing-makefiles, ...)

4-
Nobody talked about kbuild-2.5. It really correct actual makefiles. And I hope that Linus correct this bahaviour in a time manner (and not as last big makefile correction in 2.4.0-testX).

5-
What wrong with kbuild-2.5?
Keith seems to comunicate better with other kernel developers, [The main point of kbuild-2.5 are:
no more required make mrproper -> faster,
no more 'touch include/*/*.h',
multiple trees, read-oly sources,...]

6-
Old configuration programs, IMHO, are not so broken, but there is a big problem of maintainability. Every 10 patches, few are incorrect.
Few developers (maybe no developers) will use xconfig, and also few developers have read the Documentation/kbuild/. This inevitably port people to write Config.in as bash shell (WRONG!), forgetting that unset configuration have variable unset OR set to 'n'.

These problems seems not so huge, but with actual kernel development there are. It is difficult to push corection patch into the main kernel. Who will help kbuild people? [CML2 will solve the problem at the source: same engine for all interface, with more checks, so original sender will have the correct config.in].

7-
Andrea'WAlan did you make some tools for CML1 and did you not publish? Why? (not published == not existance from the other people).

8-
Read who wrote menuconfig, and you see why ESR want to replace it.

9-
Read some user list about configuration problem and you see that CML1 cause user much problem that a kernel hacker see. (menuconfig didn't compile, where is such symbols?)

No flame please.
Tell us what is wrong in kbuild-2-5 and in CML2. Don't flame! If you flame about CML2, ESR will answer you and then he forget the initial proposal of improvement.

giacomo, who don't understan why kbuild-2.5 is not in kernel

#1 lore raw
Quoted 6 lines · Giacomo A. Catenazzi

The discussion was also on lkml, ESR asked to kbuild people to give some comments, using also the feed-back of lkml.
As you noticed, in lkml the discussion went into flames, fogetting the important points.
[As this flame is going forgetting kbuild-2.5..]]

No need for a huge long e-mail to make a basic point:
Don't lump CML2 and Keith's kbuild work together.

I agree, and this is a good point.

Quoted 1 line

Tell us what is wrong in kbuild-2-5 and in CML2. Don't flame!

We've been doing that for CML2. For months, actually.

kbuild (again, as you point out) is quite another matter. Perhaps we can point Keith to post a test patch against 2.5.5-pre1, for us to review and comment on?

I believe there are probably still some issues to resolve, but I would love to see a better makefile system for 2.5.

Regards,

Jeff

Signature · 4 lines
Jeff Garzik      | "Why is it that attractive girls like you
Building 1024    |  always seem to have a boyfriend?"
MandrakeSoft     | "Because I'm a nympho that owns a brewery?"
                 |             - BBC TV show "Coupling"
#2 lore raw

Re: kbuild [which is not only ...2]

In reply to Giacomo A. Catenazzi
Quoted 1 line · [email protected] said:

I have some comment/explications about the thread about kbuild.

The thread isn't about kbuild. It's about <omitted>. Please do not confuse the two or even mention them in the same mail. It only serves to promote the confusion.

kbuild is far more obviously the right thing to do. The main objection to it last time I saw a discussion was that it has a performance problem in certain cases -- which I believe Keith is working on.

It's not clear to me that you asking for it to be reconsidered before that's complete is useful.

--
dwmw2

#3 lore raw

Re: kbuild [which is not only ...2]

In reply to Giacomo A. Catenazzi
Quoted 11 lines · David Woodhouse
Quoted 2 lines · [email protected] said:

I have some comment/explications about the thread about kbuild.

The thread isn't about kbuild. It's about <omitted>. Please do not confuse the two or even mention them in the same mail. It only serves to promote the confusion.

kbuild is far more obviously the right thing to do. The main objection to it last time I saw a discussion was that it has a performance problem in certain cases -- which I believe Keith is working on.

You will confise users...

Notation:
kbuild = kernel build utilities, now = 'kbuild-2.4' + 'CML1' + Configure.help.

Thus kbuild is not only Makefiles.

Historically (and I think still in MAINTAINERS file), the kbuild mailing list is for configurations. There was no real maintainer of Makefiles.

giacomo

#4 lore raw
In reply to Giacomo A. Catenazzi
Quoted 2 lines · Giacomo A. Catenazzi

Read who wrote menuconfig, and you see why ESR want to replace it.

I would hope he's not so petty. I believe mec once said something along the lines of "his patches were the most stable part of the build process"...

john

Signature · 1 line
"I'd rather be rudely informed than politely left in the dark."