Geometry Drift Detection — Test Runbook

Test Objective

Verify that check_geometry_integrity catches geometry mutations made outside of the normal save() path — simulating tampering via direct DB write, bulk update(), or admin bypass. This is the primary detection mechanism for Layer 2 of the geometry integrity controls.

Why This Test Matters

Farm.save() recomputes geometry_hash automatically whenever geolocation changes. The threat model is not a UI edit — the audit log catches those. The threat is a back-channel write that bypasses save(): a direct DB UPDATE, a Django admin bulk action, an ORM update() call, or a compromised import pipeline. This test simulates exactly that scenario.

Prerequisites

Procedure

Step 1 — Open the Django shell

python manage.py shell

Step 2 — Select a mapped farm

from apps.suppliers.models import Farm
f = Farm.objects.exclude(geolocation=None).first()
print(f.pk, f.name, f.geometry_hash[:16])

Note the farm pk and the first 16 characters of the stored hash. You will verify these appear correctly in the drift report.

Step 3 — Inject drift via QuerySet.update()

import copy
new_geo = copy.deepcopy(f.geolocation)
new_geo['coordinates'][0][0][0] += 0.000001   # shift first longitude point by ~0.1m
Farm.objects.filter(pk=f.pk).update(geolocation=new_geo)
print('Drift injected on Farm', f.pk, f.name)

update() writes directly to the DB, bypassing save() and leaving geometry_hash pointing at the old polygon. This is the mutation the check is designed to catch.

Step 4 — Exit the shell and run the check

python manage.py check_geometry_integrity

Step 5 — Verify expected output

Checking N farm(s)…

  ✓ Clean:   N-1
  ⚠ Missing: 0
  ✕ Drifted: 1

Farms with DRIFTED hash (geometry may have changed):
  Farm <pk> [<Company>] <Farm Name>
    stored:   <first 32 chars of original hash>…
    computed: <first 32 chars of recomputed hash>…

1 drifted hash(es) detected. Investigate before using --fix.

Confirm:

Step 6 — Check /ops/geometry/ (production only)

If running against the production DB, load /ops/geometry/ in the ops dashboard. The drifted farm should appear in the red table with stored vs computed hash columns and a link to the audit log. Note: this step only applies if the test was run against production — the ops dashboard reads the production DB, not local dev.

Step 7 — Restore clean state

python manage.py check_geometry_integrity --fix

Output should confirm the hash was rewritten:

    → hash updated

Run the check one more time to confirm clean state:

python manage.py check_geometry_integrity
# Expected: All geometry hashes are clean.

Expected Results Summary

Check Expected
Drift count Increments by 1
Farm identified Correct pk and name
Hash mismatch shown stored ≠ computed, both printed
--fix restores clean Second run shows 0 drifted
Ops dashboard (prod) Farm appears in red table with audit log link

When to Run This Test

Notes