Quality of live: Improved error messages #1731
Conversation
|
Not sure why GitHub Actions wasn't able to fetch the pace data. Works on my machine. Maybe just a network glitch. @tbennun would you mind re-running that workflow? /cc @FlorianDeconinck FYI |
|
This has happened quite a few times the last 2 days. An occasional re-run fixes the issue, but there may be an underlying issue. I've restarted the workflow |
Hmm weird. Let me propose two things 1. |
|
This seems reasonable, good idea. Thanks! Using GHA caches could make this more robust |
Thanks for bringing this to the weekly meeting. |
According to @phschaad, the pyFV3 workflow occasionally fails with connection refused errors when attempting to download the test data set for regression testing. This PR configures `wget` to retry even in case of fatal connection errors (like 404 (not found) or 503 (connection refused)). Let's see if this helps in the short term. In the long run, we should think about leveraging caches for that data because it rarely changes and it's kind of a wast to download it every time from scratch. Follow-up PR from #1731 (comment). /cc @FlorianDeconinck FYI Co-authored-by: Roman Cattaneo <>
Found two unused imports and an unused variable. From other code, I inferred that you follow the convention to prefix unused (return) variables with underscores. I noticed that running `yapf` (as suggested in the contribution guidelines) modified the files in places that I didn't touch. I thus separated the pure formatting changes in the first commit. As argued in PR #1731, I think it would be beneficial (for the project) to enforce formatting as part of the CI. @phschaad not sure if you got to discuss this in the weekly DaCe meeting. I started a discussion page #1804 and I'm happy to contribute a corresponding workflow early next year (assuming we agree). --------- Co-authored-by: Roman Cattaneo <>
Small quality of live PR that improves two error messages. Especially if the state was called
state(as per default), it was hard to parse at first sightFor some reason,
yapfwas also changing the formatting of other code parts. I thus made two commitsThis allows to separate real changes from purely stylistic ones.
Would it be an idea to enforce linting with a GitHub Actions workflow? I'm happy to write an issue about this (and contribute early next year).