Conversation
|
@xirdneh you can push improvement on this branch as use this as a base. I am checking how service containers going to work here for integration tests |
|
pypy tests were failing before this PR. |
Codecov Report
@@ Coverage Diff @@
## master #6649 +/- ##
==========================================
- Coverage 89.23% 89.08% -0.16%
==========================================
Files 138 138
Lines 16623 16623
Branches 2099 2099
==========================================
- Hits 14834 14808 -26
- Misses 1570 1597 +27
+ Partials 219 218 -1
Flags with carried forward coverage won't be shown. Click here to find out more.
Continue to review full report at Codecov.
|
|
redis connection is throwing errors. the service container might need some adjsutment |
|
@matusvalo please suggest modification too. |
|
Let me check CI in celery. It should be pretty straightforward to port kombu/py-amqp CI to celery also... |
|
Just freshened this on master. I think I'll spend a little time on it today to see if I can get the redis suite quite working as a first step and then the rabbitmq/rpc prior to merge |
33d5350 to
a080602
Compare
|
The suite starts running now, but we have a busted test which fails the run after a few retries. |
|
This pull request introduces 1 alert when merging 1dac185 into 025bad6 - view on LGTM.com new alerts:
|
| 'routing_key': 'celery', | ||
| 'priority': 0, | ||
| 'redelivered': False | ||
| 'redelivered': ANY, |
There was a problem hiding this comment.
I'm not sure why these tests stopped passing. AFAICT the redelivered seems to be set to None implying that we "don't know" (or don't bother to default to False) if a task has been redelivered or not. These tests /probably/ (??) don't care about that so changing this to an ANY seems like it is okay.
I'd like for someone a bit more familiar with redelivery logic to think about this if possible. I just don't know what we should be expecting here and why it's not False like it must have been in the past.
There was a problem hiding this comment.
I'll see if I can find some time later this week
|
This pull request introduces 1 alert and fixes 2 when merging bfab125 into 025bad6 - view on LGTM.com new alerts:
fixed alerts:
|
|
Not sure why that inspect test is failing. It passes on my box reliably. Leaving this diff for the moment, someone else can feel free to take a look at it in the meantime. |
|
This pull request introduces 1 alert when merging 8f8d402 into 025bad6 - view on LGTM.com new alerts:
|
| python-version: ['3.7', ] #'3.8', '3.9', 'pypy3'] | ||
| # XXX: Does each matrix variation get their own copy of the service | ||
| # containers? Test cross-talk may be an issue... | ||
| toxenv: ['redis', ] #'rabbitmq'] |
There was a problem hiding this comment.
This should probably actually be called backend since it's only the tail of the toxenv name.
|
Hi. I'm curious what's the motivation for this change. What problem is it solving? Automated tests, their speed and UX for developer is an area of my (informal) research so I'm genuinely interested. |
Our integration suits aren't being run as CI at the moment due to the slow migration we've had away from Travis which is where our CI jobs used to run. This change is mostly just intended to get them running again. There is an outstanding discussion around fast feedback and preferring to run the unit tests prior to the integration tests (as a prerequisite) which are intended to avoid running the more CI minute intensive suite(s) when the unit tests can show that something is clearly broken. I think you'd need to ask a more specific question for any more detail. |
|
That's good information! Do GitHub actions have some build minutes limits for OS projects? I wasn't able to google any mention of that. Another way to save CI minutes would be to use pytest -x (stop after first failure). Pro: saves probably even more minutes, Con: only indicates the first error/failure so the PR author might need to do more back and forth if not careful. |
|
Freshened this on current master, I assume that inspect test is probably still busted |
I'm not sure tbh, I do know that minutes on GitLab shared runners are limited per project/namespace so maybe GitHub is similar.
We do use |
|
This pull request fixes 1 alert when merging 6b18fdd into 5d72aee - view on LGTM.com fixed alerts:
|
|
Looks like the 3.7-redis suite passed last time I pushed. I've restored the unit tests (for sanity) and the integration matrix of python versions. I'm yet to re-enable the rabbitmq backend matrix element since I'm not sure if/how to avoid spinning up service containers which are for a backend not being tested in a particular job... |
|
This pull request introduces 2 alerts and fixes 1 when merging ada7d5a into ee9a251 - view on LGTM.com new alerts:
fixed alerts:
|
The non-eager tests needed the manager fixture so that there was a worker to actually run the async tasks.
These come back as `None` rather than the previously expected `False`.
|
This pull request fixes 2 alerts when merging 762668e into d3e5df3 - view on LGTM.com fixed alerts:
|
|
This pull request fixes 2 alerts when merging a1ae742 into d3e5df3 - view on LGTM.com fixed alerts:
|
This seems pretty bad 😬 |
was the rabbitmq tests passing? I cant remember, I can only remember redis was failing. I will be doing some experiment in another PR to see if rabbit services & tests are running. I goose step at a time, lets see |
No description provided.