linuxsh-dev

[linuxsh-dev] Status report on import of Linux 2.4.0-test1

[Mitch D. <mj...@au...>]

Hello,

This morning I tried to import the official Linux 2.4.0-test1
into our CVS repository.  Unfortunately there are a number
of source conflicts which will need to be solved before the
import is complete.

I will work privately with Niibe-san to resolve these conflicts.

Regards,

Mitch.
--
| mailto:mj...@au...       | Not the official view of: |
| mailto:mj...@al... | Australian Calculator Opn |
| Certified Linux Evangelist! | Hewlett Packard Australia |

[linuxsh-dev] Status report on import of Linux 2.4.0-test1

[NIIBE Y. <gn...@ch...>]

Mitch Davis wrote:
 > This morning I tried to import the official Linux 2.4.0-test1
 > into our CVS repository.  Unfortunately there are a number
 > of source conflicts which will need to be solved before the
 > import is complete.

I've found ide.h breakage.  The maintainer of IDE code changed
asm-sh/ide.h.  I have right version on my private tree.

Other than that, I don't see any problem.

I've sent the diff to Linus, and it is included the 2.4.0-test1.
The development went further on SuperH part.

It seems for me that "importing" can't handle such situation.
-- 

Re: [linuxsh-dev] Status report on import of Linux 2.4.0-test1

[Mitch D. <mj...@al...>]

NIIBE Yutaka wrote:
> 
> Mitch Davis wrote:
>  > This morning I tried to import the official Linux 2.4.0-test1
>  > into our CVS repository.  Unfortunately there are a number
>  > of source conflicts which will need to be solved before the
>  > import is complete.
> 
> I've found ide.h breakage.  The maintainer of IDE code changed
> asm-sh/ide.h.  I have right version on my private tree.

Yes I saw that.  Here is the listing of changed files.
The "C" ones are ones in conflict.

mjd@orbital [~/linux-src/kernel] cvs -q up
M Documentation/Configure.help
R arch/arm/kernel/iic.c
M arch/sh/kernel/entry.S
M arch/sh/kernel/irq.c
M arch/sh/kernel/ptrace.c
M arch/sh/kernel/signal.c
M arch/sh/kernel/time.c
M arch/sh/lib/checksum.S
M arch/sh/mm/cache.c
C arch/sh/mm/fault.c
C drivers/char/generic_serial.c
C drivers/char/sh-sci.c
C drivers/ide/ide.c
M fs/super.c
R fs/umsdos/check.c
R include/asm-arm/proc-armo/semaphore.h
M include/asm-sh/cache.h
M include/asm-sh/ide.h
R include/net/slhc.h
M mm/vmscan.c
M net/core/skbuff.c

> I've sent the diff to Linus, and it is included the 2.4.0-test1.
> The development went further on SuperH part.

I will see how I can fix the test0 import, and import test1
as well, tonight.  (And binutils as well! :-()

> It seems for me that "importing" can't handle such situation.

The merge mechanism is not perfect, and can't handle overlapping
changes very well.  CVS is fine when people fix different things,
even in the same file.  But if two people fix the same problem
in different ways, it can lead to very nasty conflict problems.

Regards,

Mitch.
--
| mailto:mj...@al... | Certified Linux Evangelist!
| Note: I am no longer employed by HP.  Update your aliases!

Re: [linuxsh-dev] Status report on import of Linux 2.4.0-test1

[NIIBE Y. <gn...@ch...>]

I have a question about CVS.

Now the files in CVS is somehow corrupts (conflicts or updated wrong
way).  On the other hand, I have sane files in my private tree.

Then, I'd like to commit my sane files to CVS, regardles to current
files on the repository.

How?  Just commit?
-- 

Re: [linuxsh-dev] Status report on import of Linux 2.4.0-test1

[NIIBE Y. <gn...@ch...>]

Other note.

I have to follow most resent kernel (including testing pre-releases)
to make patches to Linus (or Alan).

But I found that this will be the cause of import conflict, if
we don't have the testing pre-releases in the repository.

It seems for me that two condition are required to avoid conflict.

	(1) Have most resent version in the repository
	    when makeing patches to Linus (or Alan)

	(2) Don't touch files after sending the patch to Linus
	    until he releases new version and we import it

I'm not sure.  Is it the case?  If so, this locking scheme might be
so expensive, IMHO.
-- 

Re: [linuxsh-dev] Status report on import of Linux 2.4.0-test1

[NIIBE Y. <gn...@ch...>]

Mitch Davis wrote:
 > I will see how I can fix the test0 import, and import test1
 > as well, tonight.  (And binutils as well! :-()

There's no test0.  SuperH changes is included in -pre9.

 > The merge mechanism is not perfect, and can't handle overlapping
 > changes very well.  CVS is fine when people fix different things,
 > even in the same file.  But if two people fix the same problem
 > in different ways, it can lead to very nasty conflict problems.

Situation is like this:

Say, Linux released 2.3.99-pre8.  We worked on our part based on
2.3.99-pre8, resulted 2.3.99-pre8+sh1.  I sent the diff
2.3.99-pre8-2.3.99-pre8+sh1 to Linus.  While Linus didn't released
any, our development went further, resulted 2.3.99-pre8+sh2.  Then,
Linus released 2.3.99-pre9 which included +sh1 among other things.

What we need is to make 2.3.99-pre9+sh from 2.3.99-pre9 and 2.3.99-pre8+sh2.

	2.3.99-pre9+sh = 2.3.99-pre9 + 2.3.99-pre8+sh2 - 2.3.99-pre8+sh1

in theory.

Anyway, I have the result in my private tree.

But CVS would say "Up-to-date check failed", when I have tried to commit.
How are things solved in this situation?
-- 

Re: [linuxsh-dev] Status report on import of Linux 2.4.0-test1

[NIIBE Y. <gn...@ch...>]

NIIBE Yutaka wrote:
 > But CVS would say "Up-to-date check failed", when I have tried to commit.
 > How are things solved in this situation?

OK, I've found simple solution.  Get the source from CVS, and change it
manually with my private sane files, then commit it to CVS.

I'll do that if this has no problem.  I'll post the report, how I do it.
If there's problem, please let me know.
-- 

Re: [linuxsh-dev] Status report on import of Linux 2.4.0-test1

[NIIBE Y. <gn...@ch...>]

NIIBE Yutaka wrote:
 > I'll do that if this has no problem.  I'll post the report, how I do it.
 > If there's problem, please let me know.

I did that.  There's no problem.

I did:
	$ cp -pa kernel kernel.bak
	$ cd kernel
	$ cvs -n up
	[ ... see newer files ]

Remove newly created files in 2.4.0-test0:
	$ rm Documentation/kernel-doc-nano-HOWTO.txt    \
	     Documentation/DocBook/kernel-hacking.tmpl  \
	     Documentation/DocBook/kernel-locking.tmpl  \
	     arch/i386/kernel/cpuid.c                   \
	     arch/i386/kernel/msr.c                     \
	     arch/ppc/mbxboot/gzimage.c 		\
	     arch/ppc/mbxboot/rdimage.c 		\
	     arch/ppc/mbxboot/vmlinux.lds
	$ rm -rf drivers/char/tda7432.c 		\
		 drivers/char/tda9875.c 		\
		 drivers/char/tea6420.c 		\
		 drivers/video/sisfb.c 			\
		 include/asm-arm/arch-l7200		\
		 include/linux/netfilter_ipv6 		\
		 include/linux/sisfb.h			\
		 net/ipv6/netfilter
	$ cvs up -d -P
	$ cvs commit

It include changes for stnic.c and nfs issues too.

OK, now the CVS repository is good state.
--