[WBEL-devel] x86_64 images (ready, but not uploaded)

[Pasi Pirhonen]

Hi,


On Tue, Dec 16, 2003 at 09:54:11PM -0600, John Morris wrote:
> On Wed, 17 Dec 2003, Pasi Pirhonen wrote:
> 
> 
> > Anyway. I have built few images for x86_64 testing. Last night i
> > installed 'whitebox enterprise linux' to my dual-opteron box and just
> > recompiled the stuff once again. Today i've made ISO-images, but i have
> > only crappy 384kbit/s uplink, so i'd need some space (3.2GB to be more
> > exact).
> 
> Sounds like it is ready for a wider release.  If you would like I can drop 
> a copy on whiteboxlinux.org so the mirror network can grab it.

That would be ideal place for distribution.

> 
> Guess now is as good a time as any to work out the rules on ports.  Fair 
> warning though, AMD64 is probably a platform I'll eventually adopt myself 
> and start building inhouse.  But unless Intel or someone tosses me a 
> MB/CPU I probably will never get around to Itanic.  Same for PPC/s390/etc.
> 

Thos are just platforms that i have in my hands, so why not to use my
freetime to do ports for those archs.


> So how about this as a rough first cut on requirements to submit a port 
> and have it have equal status with i386:
> 
> 1.  Must be complete, same packages as the equivelent RHEL port.
> 

This in not problem. I was pretty strict that i used the SRPMS from the
i386-port and not those i grabbed from RedHat.

> 2.  Built from the same SRPMS as WBEL. (except for platform dependent
> packages.)  Any changes required should be submitted back and merged.
> 

That is pretty obvious.


> 3.  A commitment to get errata recompiled and submitted for inclusion into 
> the mirror network within 48-72 hours of the errata notice appearing.

That's gotcha for itanik and s390 for now. The AMD-stuff is very much
my own (paid myself), but the Itanium is my employees as is the
MainFrame-stuff. You're correct about the commitment, but are you all
by youself committed to get all this fullfilled for next 5 or so years?
There must be some kind of _community_ to get above fullfilled.

I am (now as the linux is kicking in) actually mostly doing
linux-stuff anyway. I keep anyway selfcompiled repositories available
for AS/ES 2.1 and RHEL3-stuff, so WhiteBox would just be the platform i
keep RHEL3-stuff up to date anyway. Frankly i don't much trust guys at
RedHat. By compiling you get one extra bit of (false) security - you
have the _sources_ that the binaries are built from. As .rpm you can't
say.


> 
> 4.  Because of #2 above, all of the SRPMS can be signed by me, but you 
> should generate your own signing key.  Name it GPG-RPM-KEY-whitebox-<port> 
> on the root dir of the CD and in the up2date package.  For now just 
> document it in the RELEASE-NOTES that it should be manually imported.  
> Need a better solution.

That would be obvious. Keys behind one person/organisation are problem
tho. If one is strict about that '48-72h', you'd not have vacation
ever. Not a problem for me as my vacation is computers anyway - work is
not disturbing my hobbies then :)

> 
> > I've done few NFS-instalations by burning the boot.iso on CD-RW-disc.
> > Burning now whole disc1 to see if that is really working ok.
> 
> Ok, one last minute gotcha that I haven't documented on the webpage yet
> that you might not have caught then.  Make sure you edit .discinfo in the
> main repository and snip the timestamp to two digits after you run
> buildinstall.  Because splittree truncates to two digits in the individual
> disc trees and if the longer stamp gets into the comps file
> redhat-config-packages will reject the CDs as not being the right set of
> discs.  That was the last showstopper bug that had me going till yesterday 
> evening.
> 


This didn't hit me at all? I was not aware about the above as i've
never seen it for 2.1-series. I even did media chech for first fisk and
that was OK too. I actually installed from the CD too.



> > PS. ia64 is building itself same way all the time. It's just that damn
> > slow box :) Last night whitebox install and now recompile, making
> > images, few test-installations and that is done too.
> 
> See why I care not for Itanic?  :)
> 

Judging hardware from it's compilation speed is kind of lame (i am
guilty too :). You should check what machines are at very TOP of
TPC-bech just now. The are about 700k from IBM and the there are two
over million TPCs - both are Itanium2. The better one is Linux (RHEL3) and 
the worse million is HP-UX (must hurt people at HP-UX team as they've
access to the itanic from very beginign and the hardware is from the
same house :)

If the above is messy, it's because i oversleps and am i horrible
hurry. I should really get some work to done - work that gives me money
for Opteron-systems and such :)


-- 
Pasi Pirhonen - upi@iki.fi - http://iki.fi/upi/