Open Stack

On 2018-03-15 10:05, Thomas Morin wrote:
> Hi Andreas, all,
> 
> Andreas Jaeger, 2018-03-14 20:46:
>> Note that thanks to the tox-siblings feature, we really continue to
>> install neutron and horizon from git - and not use the versions in the
>> global-requirements constraints file.
> 
> This addresses my main concern, which was that by removing
> tools/tox_install.sh we would end up not pulling master from git.
> 
> The fact that we do keep pulling from git wasn't explicit AFAIK in any
> of the commit messages of the changes I had to look at to understand
> what was being modified.

Sorry for not mentioning that.

> I concur with Akihiro's comment, and would go slightly beyond that:
> ideally the solution chosen would not only technical work, but would
> reduce the ahah-there-is-magic-behind-the-scene effect, which is a pain
> I believe for many: people new to the community face a steeper learning
> curve, people inside the community need to spend time adjust, and infra
> folks end up having to document or explain more. In this precise case,
> the magic behind the scene (ie. the tox-siblings role) may lead to
> confusion for packagers (why our CI tests as valid is not what appears
> in requirements.txt) and perhaps people working in external communities
> (e.g. [1]).

The old way - included some magic as well ;(

I agree with Doug - we need to architect our dependencies better to
avoid these problems and hacks,

Andreas

> Best,
> 
> -Thomas
> 
> [1]
> http://docs.opnfv.org/en/latest/submodules/releng-xci/docs/xci-overview.html#xci-overview


-- 
 Andreas Jaeger aj@{suse.com,opensuse.org} Twitter: jaegerandi
  SUSE LINUX GmbH, Maxfeldstr. 5, 90409 Nürnberg, Germany
   GF: Felix Imendörffer, Jane Smithard, Graham Norton,
       HRB 21284 (AG Nürnberg)
    GPG fingerprint = 93A3 365E CE47 B889 DF7F  FED1 389A 563C C272 A126