[Lordskid]

What

Reject transcript timestamps containing impossible calendar dates before parsing them. Valid leap days, fractional seconds, and timezone offsets remain supported.

Why

JavaScript's Date.parse() normalizes impossible dates, such as February 31, into dates in the following month. This can cause transcript timestamps to disagree with the parsed instants used by deadline checks.

Closes #188

Checks

• pnpm -r --include-workspace-root build
• pnpm -r --include-workspace-root test
• Golden vectors untouched
• CHANGELOG.md has an [Unreleased] entry

[maho0638]

Thanks for addressing #188. Heads-up that #190, opened earlier, already implements the same calendar-date rejection in src/transcript.ts, with regression cases for invalid February/April dates, valid leap days, fractional seconds and offsets. Its exact-head fork CI passed the full build and 177 tests: https://github.com/maho0638/tclk/actions/runs/37065379276. The implementations differ in shape/error wording, but appear to close the same issue; only one should merge. Flagging the overlap so reviewers can compare them and select a single implementation rather than spending time resolving avoidable conflicts.

[Lordskid]

Thanks for addressing #188. Heads-up that #190, opened earlier, already implements the same calendar-date rejection in src/transcript.ts, with regression cases for invalid February/April dates, valid leap days, fractional seconds and offsets. Its exact-head fork CI passed the full build and 177 tests: https://github.com/maho0638/tclk/actions/runs/37065379276. The implementations differ in shape/error wording, but appear to close the same issue; only one should merge. Flagging the overlap so reviewers can compare them and select a single implementation rather than spending time resolving avoidable conflicts.

Thanks for the heads-up and for reviewing the overlap. I wasn’t aware of PR #190 when I started working on issue #188.

My implementation includes regression tests for impossible February and April dates, plus a valid leap day. I understand that only one implementation should be merged, and I’m happy to defer to the maintainers on which solution is best.