Re: [tauruslib-devel] Taurus Continuous Integration and Documentation (was Re: Taurus-CI)

[Vincent M. <vin...@ma...>]

> That's nice, yes... although it is more or less what you get with Docker
> (except for the admin permissions part)

Yes, though I would say conda is more about managing work environments, 
while docker focuses more on isolation:

                       Docker           conda
                       v                v
< Virtual Machine --- | --- chroot --- | --- Virtual Env >

So let's say I want to run a clean preconfigured tango database for 
testing purposes. That's where docker really shines:

$ docker run -it --publish 10000:10000 \
   tango-controls/tango-database:latest

Now I want an interactive environment where I can play with an ITango 
console, run some scripts, pop-up a taurus GUI and write a small device. 
Then conda is probably a better fit:

$ conda create --name tango --channel tango-controls \
   python=3 tango=9.2 taurus itango

> I haven't worked with caching...
> I guess you 've already looked here: https://docs.travis-ci.com/user/caching/
>
> If the above does not work,  maybe you can instruct travis to deploy your
> build objects as a tar on every successful build to some ad-hoc repo (or some
> ftp server or whatever) and then download and untar it in the  before_install
> phase. This would work both with the Travis VM directly with docker (via a
> mounted volume).

I should have been more precise :)

Caching works fine, the issue is actually related to the build itself. 
It uses the file timestamps to decide whether a build object is 
deprecated or not. So I guess I would need a way to cache the sources as 
well and compare them somehow, but it sounds hackish...

Thanks for the insight,
/Vincent