[novica]

Summary

The smoke test's "scheduled invocation ran" check only looked for a START line in the function's logs, so it passed even though every scheduled run failed: a scheduled run calls the handler with no arguments, and parity(number) has no default.

• The smoke test builds from a copy of parity.R that gives number a default of 2. inst/parity.R is unchanged, since the README uses it to show why scheduled handlers need defaults.
• After the first scheduled run's REPORT line, the test fails if the run logged an ERROR line (lambdr logs a failed run with logger::log_error()). A failed log query fails the check instead of passing it.
• Log queries follow nextToken through a shared log_events() helper, also used to save the logs.

Related GitHub Issues and PRs

• Ref: fix: retry test_lambda() requests on any curl error #55 (documented that scheduled handlers get no arguments)

Are these changes tested?

Not yet against AWS: this changes dev/smoke-test.R, which needs Rscript dev/smoke-test.R to run. The script parses and passes jarl.

Are there any user-facing changes?

No.

Was AI used for this PR?

In accordance with the AI generation guidelines, please disclose whether and how AI was used in this PR.

PR code and description written by:

• Human
• AI

Reviewed before submission by:

• Human
• AI
• Not reviewed

Checklist

• I understand and agree to the Code of Conduct.
• The PR title follows Conventional Commits (fix:, feat:, docs:, ...); it becomes the NEWS.md entry.
• I have run air format . and jarl check ..
• I have updated roxygen comments and run devtools::document() if documentation changed.
• I have added testthat unit tests to tests/testthat for any new functionality or fixed bug.

Assisted-by: Claude Code