Open Stack

Hi,

I worked hard this week-end to build all tagged packages of Openstack.
The result is available from here:

deb http://archive.gplhost.com/debian grizzly main

Please use it, and test it. If possible, please use one of the mirror
closer to you (in order not to overload the main FTP):

Singapore: deb http://qala-sg.apt-proxy.gplhost.com/debian grizzly main
Seattle: deb http://seattle.apt-proxy.gplhost.com/debian grizzly main
Atlanta: deb http://ftparchive.gplhost.com/debian grizzly main
Atlanta (ipv6): deb http://ipv6-ftp.gplhost.com/debian grizzly main
London: deb http://ftp.gplhost.co.uk/debian grizzly main
Paris: deb http://33.apt-proxy.gplhost.com/debian grizzly main
Barcelona: deb http://34.apt-proxy.gplhost.com/debian grizzly main
Zurich: deb http://601.apt-proxy.gplhost.com/debian grizzly main
Haifa: deb http://972.apt-proxy.gplhost.com/debian grizzly main

This repository should be enough if you run a Debian wheezy server. I
tried apt-get install openstack-toaster (without really installing it
... yet), and all dependencies were satisfied. I'll do more tests when
I'm back at home, on Wednesday.

Many packages have failed unit tests. It would be nice if the involved
people could have a look and try to solve the issues.

Glance (one error):
http://archive.gplhost.com/debian/pool/grizzly/main/g/glance/glance_2013.1~g3-1_amd64.build

Keystone (19 errors):
http://archive.gplhost.com/debian/pool/grizzly/main/k/keystone/keystone_2013.1~g3-1_amd64.build

Nova (16 errors):
http://archive.gplhost.com/debian/pool/grizzly/main/n/nova/nova_2013.1~g3-1_amd64.build

I was very happy to see there was no unit tests error for Quantum and
for Heat, I believe unit tests for Ceilometer didn't run at all, this
will have to be investigated.

Now, I was quite disappointed to see that Horizon requires
python-selenium for its unit tests which is non-free in Debian. Could
something be done about it? It's unacceptable that the package depends
on it, because it would mean Horizon would fall into the "contrib"
repository, meaning it wouldn't be officially part of Debian (which
means, it wont be in the official CDs, etc.).

Also, I've seen that pep8 1.3.3 is now a requirement in all Openstack
projects. Frankly, that's quite annoying me. Sure, code quality is
important, but here, I don't really see the point of requiring the very
latest version of pep8. I don't control it (eg: I'm not the maintainer,
and the current maintainer isn't responsive), so I'm not sure what is
going to happen. Probably, I will disable all pep8 tests (using the -P
option of run_tests.sh).

There's also quite a few build-dependencies that I have to workout in
Debian experimental (hp3parclient for Cinder, wsme, websockify,
oslo-config and probably more that I forgot...), but that should be
doable. I've uploaded updates for many of them already (eventlet,
warlock, cliff-tab, json-patch, json-pointer, jsonschema, etc.).

All (build-)dependencies were checked against what's in
tools/pip-requires and tools/test-requires, and doing so, I've noticed
that many of them were missing there. It would be a good idea to check
what's in the debian/control, and add what's missing. Is it ok if I
prepare such commit for the different projects, and ask for review (this
would be my first commit in Openstack "upstream")?

We will (with eNovance) do some piuparts and cowbuilder tests to make
sure that no dependencies are missing, in the mean while, let's hope
none were forgotten.

Last thing, in order to solve this bug:
http://bugs.debian.org/cgi-bin/bugreport.cgi?bug=700620

I am searching for an easy way to parse a .ini file which is not really
an ini file (eg, an ini file which may contain ":" instead of "="). I
have available *ONLY* what's there after a minimalistic Debian is
installed. Does anyone has a clue on how to do it without too much
hacks? I've in the past written something with posix sh, but it was too
big, verbose, and unmaintainable. The goal would be to replace this:

#!/bin/sh

RET=$(python -c "import configobj
config=configobj.ConfigObj('${FILE}')
print config['${SECTION}']['${DIRECTIVE}']")

and fill $RET with the value of $DIRECTIVE, but with something that
doesn't lamely crashes with there's ":" instead of "=" in the config
file. Does anyone has a clue on how to do this in a smart way?

Thanks everyone for the hard work on Grizzly. This release is very
impressive! I was shocked when I see there was 60k+ insertion between
nova G2 and G3!!! Let's test it and test it again before it's out...

Cheers,

Thomas Goirand (zigo)