CI: use circleci #7603
CI: use circleci #7603
Conversation
stsewd
force-pushed
the
circleci
branch
4 times, most recently
from
612deab to
24b545e
Compare
stsewd
force-pushed
the
circleci
branch
4 times, most recently
from
9194052 to
8a796c7
Compare
Travis cut to half the number of builds for open source projects, this is slower builds. We are already paying for circleci.
.circleci/config.yml
Outdated
| lint: | ||
| <<: *default | ||
| docker: | ||
| - image: 'cimg/python:3.6' | ||
| environment: | ||
| TOXENV: lint | ||
|
|
||
| migrations: | ||
| <<: *default | ||
| docker: | ||
| - image: 'cimg/python:3.6' | ||
| environment: | ||
| TOXENV: migrations | ||
|
|
||
| docs: | ||
| <<: *default | ||
| docker: | ||
| - image: 'cimg/python:3.6' | ||
| environment: | ||
| TOXENV: docs | ||
|
|
||
| docs-lint: | ||
| <<: *default | ||
| docker: | ||
| - image: 'cimg/python:3.6' | ||
| environment: | ||
| TOXENV: docs-lint | ||
|
|
||
| eslint: | ||
| <<: *default | ||
| docker: | ||
| - image: 'cimg/python:3.6' | ||
| environment: | ||
| TOXENV: eslint | ||
| NODE_VERSION: 10.17.0 |
There was a problem hiding this comment.
Looks like we can speed up things by running all these in one job, otherwise circle will create a new container per each job and we can only have 5 concurrent jobs
There was a problem hiding this comment.
I don't fully follow. Meaning just have 1 job that calls tox and it runs all 5 steps? I think we do that on the ad server: https://github.com/rtfd/ethical-ad-server/blob/ede6a0a0d2f9609f9e6426f59b4666219038a882/.circleci/config.yml#L27
There was a problem hiding this comment.
yeah, just call tox with all except for the py36 environment (since that requires ES running, but maybe isn't that bad having ES running during all tests?). Also, splitting will give us the linter failures more quickly :D
There was a problem hiding this comment.
I think 1 job per tox step is easier to parse while reading the errors. Right now we have 6 check notifications in this PR (1 per job) and you know immediately what (docs, lint, etc) failed without going to CircleCI.
Also, if we have 1 job per tox step, don't they run in parallel?
There was a problem hiding this comment.
Also, if we have 1 job per tox step, don't they run in parallel?
Yeah, but then circle will be creating one container per each task, and with our current plan we can only have 5 concurrent jobs. So other PRs will need to wait till there is space available.
There was a problem hiding this comment.
I think we can run tox -e <env> on separate steps so is more clear what failed
stsewd
removed
the
PR: work in progress
|
by using CircleCI, will the output of the tests be public? we need a way to configure that probably, because right know I can't access without being logged in in CircleCI. |
|
The project is already marked as open source. Looks like you need to be logged-in anyway...
|
|
I think 2 checks seems like a good compromise. 👍 Requiring login is annoying though. Do we know if all users who login can see it, or do they need repo access? |
|
According to the docs they should be able without having repo access. But I don't have another account to test with :/ |
What should we do with this check? Why do we need to allow this test to fail? |
I think it's a "nice to have in the future" set of checks that we haven't enabled yet. I think we can probably just not run them, until they pass? |
I think is fine running the check locally, I never looked at that job. It fails because the rate limit from github (the changelog has a lot of links to github). |
|
Sounds good! I created an issue (#7212) to re-enable this check in the future. Maybe it's possible to disable the check in some pages to avoid this rate limit. |
Travis cut to half the number of builds for open source projects,
this is slower builds. We are already paying for circleci.
Circleci doesn't support "allowing failures", so I didn't add the docs-linkcheck job here.