[Starosdev]

Summary

Manufacturer SMART failures now send a notification at notify level Warn, and in every other settings combination. Two defects caused the bug. Both are fixed here for 1.x.

1. The when_failed constants did not match smartctl's output. smartctl writes "now" and "past", but 1.x compared the value against "FAILING_NOW" and "IN_THE_PAST". An ATA attribute was therefore never flagged from the drive's own threshold. This PR cherry-picks the fix already on beta (e2e6ddd8, part of fix(smart): detect manufacturer SMART failures for ATA and NVMe and keep them through recalculation #970).
2. ShouldNotify ignored a failed overall health check (smart_status.passed=false) unless one attribute was also flagged. Only the shortcut for Filter=All, Repeat=Always and a level other than Warn checked the device status. So Warn, Filter=Critical, and Repeat=Only-on-change all dropped the failure. The beta code has the same defect.

Type of Change

• Bug fix (non-breaking change that fixes an issue)
• New feature (non-breaking change that adds functionality)
• Breaking change (fix or feature that would cause existing functionality to change)
• Documentation update
• Refactoring (no functional changes)
• CI/CD or infrastructure change

Related Issues

Fixes #1108

Changes Made

• constants.go, smart_ata_attribute.go: match when_failed against "now" / "past" with strings.EqualFold. This is a cherry-pick of e2e6ddd8 from beta. The only conflict was in a test file, where both branches had added tests at the end.
• notify.go: if this submission has DeviceStatusFailedSmart, and the status threshold covers it, ShouldNotify now returns true at every notify level and attribute filter. The check uses this submission's status (smartAttrs.Status), not the stored device status. The stored status keeps failure bits until a fully passing submission resets it. With Threshold=Scrutiny, the failure is still ignored.
• With repeat notifications off, the failure notifies only on a transition, that is, when the stored status before this submission did not have FailedSmart. ShouldNotify gets a new previousDeviceStatus argument. The upload handler passes device.DeviceStatus from ResolveDevice, which is read before reconcileDeviceStatus updates it. No new database read and no schema change.
• New fixture smart-fail2-helium-only.json, provided by the reporter: only Helium_Level (22) fails, with "when_failed": "now". It has no Scrutiny thresholds.

Testing

• I have tested this locally

• I have added/updated tests that prove my fix/feature works

• All existing tests pass (CI green; one test fails only locally, see below)

• go build ./... and go vet on pkg/notify, pkg/models/measurements and pkg/web/handler: clean.

• go test ./pkg/models/measurements/ ./pkg/web/handler/: pass.

• go test ./pkg/notify/: everything passes except TestScriptStreamsOutputAndPreservesEnvironment, which fails locally on macOS (Condition never satisfied). That test exercises sendScriptNotification, which this PR does not change. It passes in CI (Test Backend green at 6b56b400), so the failure is specific to the local macOS environment.

• I confirmed that the new tests detect the bugs. With the ShouldNotify check disabled and the old "FAILING_NOW" constant restored, TestShouldNotify_OverallHealthFailed_QualifiesAtEveryLevelAndFilter, ..._NoRepeatNotifiesOnlyOnTransition, TestPopulateAttributeStatus_WhenFailed and TestFromCollectorSmartInfo_Fail_ManufacturerThresholdOnlyAttribute all fail.

Checklist

• My code follows the project's code style
• I have performed a self-review of my code
• I have commented my code where necessary (particularly complex areas)
• I have updated the documentation if needed (no documentation describes notify-level behavior)
• My changes generate no new warnings
• No console.log, debug statements, or commented-out code

Additional Notes

Review: not required (the change lands on develop and reaches users only through a promotion PR and a manual release. It does not change a security boundary or a shared contract).

Verdict change for existing drives: fixing when_failed changes the verdicts for drives that are already monitored.

• An ATA attribute that smartctl reports as "past" becomes WarningScrutiny. At notify level Warn with repeat notifications on, that drive now sends a notification on every submission.
• An attribute reported as "now" becomes FailedSmart and shows as failed in the UI.

Both results are correct, but users can see new warnings and notifications after they upgrade.

Known limit with repeat notifications off: UpdateDeviceStatus ORs new bits into the stored device status. So FailedSmart stays set until a submission passes fully. Example: the overall health check fails, then passes while a Scrutiny failure is still active, then fails again. The second failure does not send a notification, because the stored status before that submission still has FailedSmart. This case is rare. Fixing it means storing the status per submission, which is out of scope here.

Behavior change: with Filter=Critical, a failed overall health check now notifies even when no critical attribute is flagged. This is intended. The overall health result is the drive's own verdict, not an attribute.

Follow-up for the next master-into-beta sync:

• Expect a small conflict in constants.go, where beta has the same fix in a different form.
• On beta, fix(smart): detect manufacturer SMART failures for ATA and NVMe and keep them through recalculation #970 also sets FailedSmart on the device from an ATA attribute that is not ignored (setFailedSmartFromAttribute). With this change, a non-critical attribute that fails its manufacturer threshold will therefore also notify under Filter=Critical. Decide during the sync whether to keep that.