[Fedora-livecd-list] kadischi+kickstart is broken how exactly?

[Jane Dogalt]

> Where Kadischi is now, I do not see any benefit in having nightly 
> tarballs, or RPMs
> generated. Maybe someone else does.. I doubt it.
> 
> J. Hartline

>From my experience with regression testing, here is the kind of system I
envision for my own project, and which I would recommend for the kadischi
maintainers-

a daily(/hourly/per-cvs-commit) script which-

a) starts a headless virtual machine which has a pristine image of a completely
default fc5 image.  I.e. vmware displaying to a vnc x server, or xen or qemu
which I haven't played with yet. 

b) script ssh's a command into the virtual system which does the following-

b2) cvs checks out kadischi tree

b3) invokes script within tree which rpmbuilds tree into src and binary rpms

b4) yum/rpm install fresh kad rpm (and reqs) into this virtual system

b5) using working kickstart sample included in rpm, build a new basic default
    livecd

c) perhaps still within this VM, boot new livecd in a VM.

d) ssh into livecd VM, and run a regression test.  first regression test is
just a 'hostname', proving that you could ssh into the livecd and run hostname.

Eventually of course you would probably want the original host system to be a
livecd generated from a stock kickstart file included in the kad rpm.

If you have the above process completely automated and headless, it can sit and
churn out tests after every cvs commit to the tree, and 

a) let you know when a cvs commit breaks the tree

b) when the tree is unbroken, provide automatically generated rpms which one
can confidently say have no dependencies on some developers build
system/environment.

The phrases 'self hosting' and 'eating our own dogmeat' should come to mind.

Hopefully this is one of the reasons why you guys are spending so much time
getting xen to work... :)

-jdog


__________________________________________________
Do You Yahoo!?
Tired of spam?  Yahoo! Mail has the best spam protection around 
http://mail.yahoo.com