[Victsm]

Follow-up to #123.

Summary

Unbounded past windows such as SUM(ORDERS.ORDER_SCORE, -INF, 0) fail the new clock validation because the compiler's -2**60-day sentinel overflows to zero when converted to seconds. Saturate that conversion consistently with the other clock conversions, retaining rejection of genuinely sub-tick windows.

Update the snapshot-ingestion test doubles with timestamp_precision='seconds', matching the real DiskgraphIngestArgs default. No production ingestion fallback or API change is added.

Change map

Maintained change map.

Testing Done

• Native release extension/ingest build passed.
• Rust time tests: 73 passed. The new regression fails on the base with zero instead of i64::MIN.
• Python index-clock suite against the rebuilt extension: 33 passed, including 12 unbounded day/week cases and existing sub-tick rejection cases.
• Seven selected snapshot-ingestion tests: 7 failed before the fixture correction; 7 passed after it. The isolated runner binds this checkout and excludes ancestor service setup.
• Changed snapshot-test file: flake8 and isort checks passed.

Limits: the two original high-level PQL tests were attempted with ancestor setup excluded and stopped at missing fixture column statistics (AGE: COUNT/NUM_NA), before reaching the query path. They remain for normal CI qualification. Whole-file rustfmt reports existing formatting differences outside the changed lines; no unrelated formatting was included.

The baseline duration regression fails with 0 instead of i64::MIN; all seven selected snapshot cases fail before the fixture correction. No GPU/cloud jobs or live sweeps were run.

Checklist

• Breaking changes: none intended; restores unbounded past-window validation.
• No documentation change needed: existing query/default-clock behavior is preserved.
• Focused regression tests added and updated.

[Victsm]

Change map

File	Function / scope	What changed	Why
packages/diskgraph/diskgraph-core/src/time.rs	Seconds::dur_from_days; duration regression	Use saturating multiplication and test positive/negative overflow	The existing unbounded PQL sentinel wraps to zero; #123's tick validation then rejects a valid window. Other clock conversions already saturate.
packages/diskgraph/diskgraph-pybind/python/tests/test_index_clocks.py	test_unbounded_past_windows_are_not_rejected_as_sub_tick	Exercise day/week unbounded past windows on all six index clocks	Guard the Python builder boundary while retaining the existing sub-tick rejection tests.
test/pool/test_activities.py	Snapshot-ingestion SimpleNamespace fixtures	Supply the real argument type's timestamp_precision='seconds' default	#123's worker reads the new field; incomplete test doubles currently fail before exercising ingestion/recovery behavior.

Observed failures: unbounded-window validation raises “shorter than one tick”; snapshot fixture cases raise missing timestamp_precision or time out waiting for the skipped ingest call. The duration multiplication predates #123; its new validation exposes the overflow. Keep the production clock validation and transport dataclass unchanged. Stop after focused duration, Python clock-boundary, and snapshot-fixture verification; no additional clock features, validation framework, or workflow changes.

[copy-pr-bot]

This pull request requires additional validation before any workflows can run on NVIDIA's runners.

Pull request vetters can view their responsibilities here.

Contributors can view more details about this message here.

[coderabbitai]

No actionable comments were generated in the recent review.

Recent review info

Run configuration

• Source mode: Auto
• Repository: NVIDIA/kumo-engine
• Branch: codex/fix-index-clock-regressions
• File: .coderabbit.yaml
• Commit: f17b5d7
• Layers: Repository YAML
• Review profile: QUIET
• Plan: Enterprise
• Run ID: c68d3b74-5880-4bdb-a7e2-3e1b4f956f6a

Commits

Reviewing files that changed from the base of the PR and between 9ac307d and f17b5d7.

Files selected for processing: 3

• packages/diskgraph/diskgraph-core/src/time.rs
• packages/diskgraph/diskgraph-pybind/python/tests/test_index_clocks.py
• test/pool/test_activities.py

Included review availability: This review used your included allowance. Your plan provides up to 12 included reviews per hour; 11 remain after this review.

Walkthrough

Day-to-second conversion now saturates at the i64 limits when multiplication overflows. Tests cover extreme day values and temporal windows across configured clocks and window units. Seven snapshot-ingest test fixtures now set timestamp precision to seconds.

Merge Risk: Minimal · up to f17b5

No actionable merge risk remains in the reviewed changes.

Pre-merge checks: 2 · 1

Failed checks: 1 inconclusive

Check name	Status	Explanation	Resolution
Change Map Design Review		The required conversation comment beginning <!-- kumo-pr-change-map --> is not present in the supplied PR conversation or repository checkout, so the comment-specific design claims (observed failure…	Provide the full change-map conversation comment, including its linked problem, per-change rationale, inspected behavior, verification evidence, and stopping condition. Then reassess the design against the code diff.

Passed checks: 2 passed

Check name	Status	Explanation
Pr Size		The pull request changes 3 files. The limit is 40 files, so the pull request passes this check.
Public Resource Access		The pull request adds no external URL, repository location, or package source location. Added lines only change Rust arithmetic, add Python regression-test logic, and set `timestamp_precision='seconds…

Full details: Public Resource Access

Explanation

The pull request adds no external URL, repository location, or package source location. Added lines only change Rust arithmetic, add Python regression-test logic, and set timestamp_precision='seconds' in test fixtures. The s3:// test fixture reference is unchanged and is excluded by the check.

Full details: Change Map Design Review

Explanation

The required conversation comment beginning <!-- kumo-pr-change-map --> is not present in the supplied PR conversation or repository checkout, so the comment-specific design claims (observed failure, stopping condition, and focused verification) cannot be assessed from the required source.

Resolution

Provide the full change-map conversation comment, including its linked problem, per-change rationale, inspected behavior, verification evidence, and stopping condition. Then reassess the design against the code diff.

• Autofix · Keep fixing CodeRabbit findings and required CI, and resolving merge conflicts

CodeRabbit Verify

• Run CodeRabbit Verify to test the changes in this pull request.