continuous integration about gpfit HOT 22 CLOSED

your call. should only take a couple of minutes to get things setup with our jenkins, or you can use one that github provides. We have complete control over ours, no idea what they provide.

from gpfit.

Another thing to consider is any dependencies you have. I don't know what machines they are providing for the CI testing, or what means you have to install your dependencies.

Let me know what you want me to do. I just need to know how to run your test (which I assume is the same as gpkit), and what dependencies you have for gpfit.

from gpfit.

Well, I started looking into TravisCI just so we could check out another CI setup, but quickly learned that "Python builds are not available on the OSX environment." That's enough to make me uncomfortable with Travis as a CI solution for our Python project.

Therefore, @galbramc if you could help us get set up on jenkins we'd really appreciate it. Dependencies for gpfit are:

• numpy
• gpkit

from gpfit.

I presume you want both pull request and push configurations just like you have with gpkit. I also presume you want me to use the master branch of gpkit as the dependency?

from gpfit.

correct, and correct.

probably don't need to test for multiple solvers -- the part of gpkit that gpfit uses doesn't involve solvers.

from gpfit.

ok, so do I need to build gpkit, or can I just add the gpkit to the python path?

from gpfit.

I'm not sure --

We definitely want to be testing gpfit against a current version of gpkit, such as that installed by pip install gpkit. Would the "add the gpkit to the python path" solution automatically keep gpkit up to date at the resolution of our PyPI releases?

from gpfit.

Ok, how about I create a virtualenv were I install gpkit every time we run the tests. That way you always have what you get with pip. Sound good?

Also, could you add from .run_tests import run to init.py in gpfit/tests just for consistency?

from gpfit.

sounds great.

Just did so in #32. Do you need that change on master or should I leave that branch open as the first test of jenkins/CI?

from gpfit.

Sure leave it open as the first test so I can get the script sorted out

from gpfit.

Do you want pylint? If so, you should probably add the pylint.cfg file from gpkit and grab the code from run() in run_tests.py

from gpfit.

You also need to give acdl-jenkins the same privlages to gpfit as it has for gpkit before I can test

from gpfit.

• ok, I gave acdl-jenkins access
• by grab the code in run_tests, I assume you mean the xmloutput stuff. Done -- added to this PR.
• I want pylint and added the config file to this PR, but I'm confused / have a question. I don't actually see pylint warnings over at gpkit, even though I know we have lots of them. Can you re-educate us on how to see how we're doing on pylint?

from gpfit.

hmmm, still getting this error:
Couldn't create web hook for repository hoburg/gpfit. Does the user (from global configuration) have admin rights to the repository?

Yes the xml stuff.

Let me check the pylint stuff

from gpfit.

nope, has write access, which is the same setting as gpkit. Does it need admin access?

from gpfit.

Yeah I think it needs admin access...

from gpfit.

Same goes for gpkit, I have seen some complaints about gpkit as well, and this might explain why

from gpfit.

Here is what is going on with pylint:

https://acdl.mit.edu/csi/job/gpkit_PullRequest/658/buildnode=reynolds,optimizer=mosek/violations/

You currently have 728 violation, however the jenkins default configuration was to allow 999 violations, so this is why you have not gotten any failed builds from pylint.

I can lower the number, but you probably need to start an initiative to become pylint compliant. Presumably it will be easier to get gpfit compliant as it appears to be smaller at the moment.

from gpfit.

We could gradually lower the threshold so you don't have to fix all problems at once

from gpfit.

I'm around all day today if you want to get this wrapped up. We should also talk about the pylint configuration

from gpfit.

coming over now

from gpfit.

add .pylintrc and pylint.sh to the repo and then I think you can close this issue

(don't forget about test/pylint.cfg)

from gpfit.