[Date Prev][Date Next] [Thread Prev][Thread Next] [Date Index] [Thread Index]

Re: [x86-64] AMD Dev Center usage proposal - draft 2



If you get this more then once, sorry for the spam.

If you are not yet subscribed to <debian-x86-64@lists.debian.org>,
please visit http://lists.debian.org/debian-x86-64/
We will be phasing out the use of the alioth mailing list.

I wanted to address some issues from draft 2, I will do so below.
Comments are welcome.

* Bart Trojanowski <bart@jukie.net> [030523 18:04]:
> I will try to summarize our progress as we go along...
> 
> I hope I've captured your comments accordingly.
> 
> 1) starting point (what we will require to start)
>   - dual boot system running SuSE 64-bit and Debian/Sid 32-bit
>   - serial console access from another box/terminal server/etc.
>   - a remote-access power switch to the system
>   - ssh access to the box itself
> 
> 2) decide on a port name
>   - amd64 was suggested (I agree personally)
>   - decide on the .deb naming to distinguish amd64/i386/biarch
>   - who should have the final say?

    - x86_64 has an advantage of being consistent with other distributions
    - configure scripts and dpkg/apt tools could have amd64 as an alias

> 2) testing 32-bit user-space with 64-bit kernel
>   - on the Debian/Sid partition we will run the SuSE binary kernel (to start)
>   - we should be able to run simple programs
> 
> 3) running 64-bit tool-chain
>   - we will install Arnd's biarch compiler and libraries
>   - we should be able to compile and run simple C programs
> 
> 4) ability to build a 2.4.20+ 64-bit kernel natively
>   - Arnd has done some progress here, but last I heard the work was not
>     yet completely finished
>   - we should be able to build, install, and boot a 2.4.20+ kernel.
> 
> 5) port all base libraries
>   - /lib64 migration for 64 bit libraries
>   - we should be able to compile and run simple C programs against these
>   - what are they?  how many?

I had a look at the Debian Policy Manual section 2.2 which describes
package priorities [1].  I propose that we port all packages listed as
required [2], important [3], and standard [4].  In our case the order
does not matter and due to dependencies it is important to reverse the
order.  There are 158 packages in total [5].

[1] http://www.debian.org/doc/debian-policy/ch-archive.html#s2.2
[2] http://www.jukie.net/~bart/debian/amd64/proposal/packages.required
[3] http://www.jukie.net/~bart/debian/amd64/proposal/packages.important
[4] http://www.jukie.net/~bart/debian/amd64/proposal/packages.standard
[5] http://www.jukie.net/~bart/debian/amd64/proposal/packages.r+i+s

I propose that at the completion of this milestone we call ourselves
'functional'.

In this step -- step 5: 'port all base libraries' -- we aim to port
libraries since that is the most challenging part of the work and
because they are the necessary evil to get to step 6, 7 and 8.  There
are 41 libraries (here assumed to be part of packages containing lib)
in the above set of packages [6].

[6] http://www.jukie.net/~bart/debian/amd64/proposal/packages.libs

> 6) port packages required by autobuilder
>   - what are they? how many?

I will admit that I don't know where to get autobuilder, so I don't know
what dependencies it requires.  Can someone lend a hand here?

> 7) port/setup autobuilder
>   - once running certain packages may need to be tweaked 
> 
> 8) port remaining base packages
>   - what are they? how many?

I am not sure what packages will fall off in 6, but I can generate a
non-library list [7].

[7] http://www.jukie.net/~bart/debian/amd64/proposal/packages.rest

> [ and later ]
> 
> * Hack dpkg to understand the new architecture.
> 
> * Make the ftp people create the new architecture in the archive.
> 
> * ability to build a 2.5.x 64-bit kernel natively
>   - we should be able to build, install, and boot a recent 2.5.* kernel.
> 
> * testing?
> 
> * install CD...

-- 
				WebSig: http://www.jukie.net/~bart/sig/

Attachment: pgpN413Jhfux6.pgp
Description: PGP signature


Reply to: