[tongwentao]

The custom before_sleep logging example checks attempt_number < 1, but attempts are numbered from 1. Its INFO branch is therefore unreachable, and every retry logs at WARNING.

Check attempt_number == 1 so the first retry logs at INFO and subsequent retries log at WARNING. The README links to the same documentation source, so both views receive the correction. Library behavior is unchanged. Include a Reno note.

Behavior and review notes

This callback runs before sleeping for a retry, so the example's first callback already has attempt_number == 1. The correction changes the example's choice of log level, not Tenacity's retry count, wait policy, or callback scheduling. Running the exact documentation block checks the advertised INFO-then-WARNING sequence rather than a separately maintained approximation.

Validation on Linux, Python 3.13.5:

• Executed the exact example from the documentation, intercepting Logger.log: before the correction both events were WARNING; afterward they are INFO then WARNING.
• poe check: 185 tests and 35 documentation tests passed; strict Sphinx HTML build passed.
• poe lint and poe mypy: passed.
• git diff --check: passed.

Prepared and validated with Codex assistance.

[mergify]

Tick the box to add this pull request to the merge queue (same as @mergifyio queue).

• Queue this pull request