[TheWitness]

Summary

Fixes a materialization bug where the newest cold partition is never cached, so a Filter keeps showing Required even immediately after a successful materialization. On an install with 1,523 partitions this showed as NAT/Non-NAT shards capping at 1,521 while there were 1,522 cold partitions — exactly one cold partition (the most recently rotated) perpetually missing.

Root cause

A partition is cached only when its shard is a full scan, which requires the materialization range to lie strictly outside the partition's data (flowview_is_full_scan). flowview_get_materialize_range padded the start by a full day but the end by only one second.

A flow is written to the partition of the time it was captured, but its end_time can fall after that partition's nominal window:

• a flow still open when the partition rotates is captured in it but ends in the next one, and
• an exporter whose clock runs ahead of the collector reports an end time past the window.

The newest rotated partition is the one most likely to hold such a boundary-spanning flow (and is also the one still on InnoDB before the poller converts it to Aria). Its MAX(end_time) fell a few minutes past the one-second end pad, so it was never a full scan and never cached — and re-materializing never helped, because it is still the newest cold partition with the same too-tight range end.

Fix

Pad the end by a day too, symmetric with the start ([$start - 86400, $end + 86400]). flowview_get_materialize_range is only used by the materializer, so report ranges are unaffected. The live partition is still excluded from full scans regardless of range.

How I verified it

I reproduced it on a scratch Cacti 1.2.32 + FlowView install:

1. Created 6 daily partitions (5 cold + 1 live), the newest cold one left on InnoDB, all data inside its window → all 5 cold partitions cache, status would be Up to date.
2. Added one flow to the newest cold partition ending 5 minutes past its window → that partition drops to full_scan = NO, cached 4/5, perpetual Required (matches the report).
3. Applied the fix → the newest cold partition is full_scan = YES again, cached 5/5.

Testing

• Updated the flowview_get_materialize_range test for the new end pad and added a regression test that a partition whose newest flow ends after its window is a full scan (and that the live partition never is).
• Unit suite: 240 passed locally; the only failures are 4 pre-existing background-exec tests that fail the same way on unmodified develop under Windows (shell quoting).

[sonarqubecloud]

Quality Gate passed

Issues
0 New issues
0 Accepted issues

Measures
0 Security Hotspots
0.0% Coverage on New Code
0.0% Duplication on New Code

See analysis details on SonarQube Cloud

[Copilot]

🟢 Approval recommended

The localized fix addresses the materialization boundary condition with targeted regression coverage and no unresolved findings.

No open issues 🎉

What changed in this PR

Fixes FlowView materialization skipping the newest rotated partition when a flow’s end time extends beyond its nominal window, leaving the Filter marked Required.

Changes:

• Extends the materialization end padding from one second to one day, matching the start padding.
• Updates range expectations and adds regression coverage for boundary-spanning flows and live-partition exclusion.
• Documents the fix in the function comment and changelog.

File	Description
tests/​Unit/​CoreFunctionsTest.php	Tests the wider range and full-scan eligibility while preserving live-partition exclusion.
includes/​functions.php	Adds one-day end padding and explains its purpose.
CHANGELOG.md	Records the materialization fix and its user-visible effect.

_{Review effort: Balanced}