opendal icon indicating copy to clipboard operation
opendal copied to clipboard
verified publisher icon apache

[core split] Layer: retry

Open bytelazy opened this issue 2 weeks ago • 2 comments

Closes #6947

This PR splits the layer retry out of core as part of the core-split effort and follows the tracking issue #6829.

Summary

  • Keep public API/behavior unchanged.
  • Move the retry layer source into the new crate layout described in the tracking issue.
  • Add/update feature flags and tests as required.

bytelazy avatar Dec 10 '25 16:12 bytelazy

Hi, just curious, are you using a code agent for this PR? If so, are you using the prompts I provided in the tracking issue? Which tool and LLM are you using?

Xuanwo avatar Dec 11 '25 06:12 Xuanwo

Hi, just curious, are you using a code agent for this PR? If so, are you using the prompts I provided in the tracking issue? Which tool and LLM are you using?

@Xuanwo Sorry for missing the fact that the tracking issue already provided a complete prompt.

I will revert the two current PRs (#6964 and #6961), and then rework the changes by following the prompt you provided and submit a new set of commits accordingly.

For transparency, the LLM tool I used was Codex.

bytelazy avatar Dec 13 '25 12:12 bytelazy

hi @bytelazy, thanks for your work, I found the tests across these three layers (logging/retry/timeout) are interdependent, so it's best to modify them together in a single PR (see #7053)

koushiro avatar Dec 18 '25 06:12 koushiro

Thank you @bytelazy for working on this. As explained in https://github.com/apache/opendal/pull/7053, these three layers are highly coupled in many of our tests. Therefore, we finally decided to migrate them all at once.

We’re sorry, but we have to close your PR for now. We hope to see your contributions again in the future.

Xuanwo avatar Dec 18 '25 14:12 Xuanwo

Thank you @bytelazy for working on this. As explained in #7053, these three layers are highly coupled in many of our tests. Therefore, we finally decided to migrate them all at once.

We’re sorry, but we have to close your PR for now. We hope to see your contributions again in the future.

Thanks for the clarification and explanation.

That makes sense — I see why migrating the logging, retry, and timeout layers together is preferable given the test coupling. Thanks for pointing this out and for the reference to #7053.

I appreciate the feedback and will align with this direction in future contributions. Looking forward to contributing again when there’s a good fit.

bytelazy avatar Dec 21 '25 04:12 bytelazy