zeph: Q60 self-scan leg (KNOWN_FIXTURE + masked diagnostics + self-scan cell)

wizard-work/zeph/q60_enforce.py

@@ -24,11 +24,25 @@ What it does, on a checkout at a NAMED head:
      8 hex chars match a KNOWN full value's prefix but whose value differs
      FAILS the gate. The known invented placeholder for beae0b1f is excluded
      from this leg by construction: that prefix has no KNOWN full value.
+     Two qualifications from reply 113 / interview 117: (a) the selftest's
+     own fixture value classifies TEST-FIXTURE, not FAIL, and is built by
+     concatenation so no literal for it exists in this source; (b) FAIL
+     diagnostics are masked (stem + length + site, never the full value),
+     because a report quoting the fault re-commits it at the next head.
+     The beae0b1f stem rule keys on the 8-char stem: any future token
+     sharing that stem classifies exhausted without a look (reply 113
+     section 1 boundary note).
+
+Quoting rule (interview 117 Q2): a verdict is only ever quoted at a head
+that contains the artifact it grades. Green readings at heads predating
+this file are census readings, not evidence for the check.
 
 Exit status: 0 when no FAIL line is printed, 1 otherwise.
 ``--selftest`` builds fixture trees and asserts the negative cells: an
-unknown truncated token fails, a complete-but-wrong value fails, and a
+unknown truncated token fails, a real complete-but-wrong value fails, the
-self-quoting report file excluded from the population collapses the count.
+fixture value classifies TEST-FIXTURE, a self-quoting report file excluded
+from the population collapses the count, and the check over the checker's
+own directory passes (cell 7, the self-scan leg).
 
 Usage:
   python3 q60_enforce.py --tree <checkout-path> [--exclude <relpath> ...]
@@ -79,6 +93,18 @@ KNOWN_EXTERNAL = {
 KNOWN_FULL = dict(KNOWN_IN_TREE)
 KNOWN_FULL.update(KNOWN_EXTERNAL)
 
+# Test fixture, not an archive claim (reply 113 / interview 117 section 0.7):
+# the complete-but-wrong value used by selftest cell 4's predecessor. It is
+# built by concatenation on purpose: no UUID-shaped literal for this value
+# may exist in this source, or the file would fail its own leg (the fault
+# that kept Q60 open at 36c6bba). Occurrences classify TEST-FIXTURE and do
+# not fail -- but the line is still printed with its site, so a re-commit
+# of the same value elsewhere stays visible in the report rather than
+# silent. Value alone cannot distinguish fixture from re-commit; site can.
+KNOWN_FIXTURE = {
+    "3d79ba82": "3d79ba82-" "1111-2222-3333-444444444444",
+}
+
 # Terminal justified-unrecovered state (reply 100 section 4): exhausted by
 # full-history pickaxe, and the only longer string anywhere is Timmy's
 # invented placeholder, which is not a recovery and must not count as one.
@@ -150,7 +176,10 @@ def run_check(tree, excludes):
             for rel, text in contents.items():
                 for m in pat.finditer(text):
                     cand = m.group(0)
-                    if cand != PLACEHOLDER:
+                    # The placeholder and the test fixture are not
+                    # recoveries: skip both, or the recovery line would
+                    # name the fixture value as a recovery.
+                    if cand != PLACEHOLDER and cand not in KNOWN_FIXTURE.values():
                         longer = (cand, rel)
                         break
                 if longer:
@@ -167,11 +196,20 @@ def run_check(tree, excludes):
             lines.append("FAIL TOKEN %s: no longer string in tree, not in known tables" % stem)
             failed = True
     # Complete-but-wrong leg: full values whose 8-prefix is known.
+    # Diagnostics are MASKED (interview 117 section 0.7): the offending
+    # value is never printed in full, because every report that quotes the
+    # fault re-commits the fault at the next head. Stem + length + site is
+    # what a reader needs to locate it; the known value is already in the
+    # tables above, so printing it adds no new value to the record.
     for rel, text in contents.items():
         for val in set(FULL_RE.findall(text)):
             prefix = val[:8]
             if prefix in KNOWN_FULL and val != KNOWN_FULL[prefix]:
-                lines.append("FAIL COMPLETE-BUT-WRONG %s at %s (known %s)" % (val, rel, KNOWN_FULL[prefix]))
+                if val == KNOWN_FIXTURE.get(prefix):
+                    lines.append("TEST-FIXTURE %s-\u2026 (len %d) at %s" % (prefix, len(val), rel))
+                else:
+                    lines.append("FAIL COMPLETE-BUT-WRONG %s-\u2026 (len %d) at %s (known %s)"
+                                 % (prefix, len(val), rel, KNOWN_FULL[prefix]))
                     failed = True
     lines.append("VERDICT=%s" % ("FAIL" if failed else "PASS"))
     return lines, failed
@@ -188,8 +226,14 @@ def selftest():
 
     with tempfile.TemporaryDirectory() as d:
         # Cell 1: clean tree passes (one harmless abbreviation + its full value).
+        # Positive control, not a tautology (interview 117 Q3): a scanner
+        # returning [] for every input would pass a bare rc assertion, so
+        # the cell also asserts the RECOVERED-IN-TREE line is present.
         with open(os.path.join(d, "a.md"), "w") as fh:
             fh.write("see 3d79ba82-\u2026 and full 3d79ba82-926b-48c0-9d15-8791a364ea81\n")
+        lines, _ = run_check(d, [])
+        assert any(l.startswith("TOKEN 3d79ba82: RECOVERED-IN-TREE") for l in lines), \
+            "clean tree missing its recovery line (gutted scanner would pass rc alone)"
         check(d, [], False, "clean-tree-passes")
         # Cell 2: a self-quoting report excluded collapses the count.
         with open(os.path.join(d, "report.md"), "w") as fh:
@@ -204,9 +248,11 @@ def selftest():
             fh.write("dead link deadbeef-\u2026 with no longer string\n")
         check(d, ["report.md"], True, "unknown-token-fails")
         os.remove(os.path.join(d, "b.md"))
-        # Cell 4: complete-but-wrong fails (same prefix, different suffix).
+        # Cell 4: a REAL complete-but-wrong value fails. It must not be the
+        # fixture value (cell 6 owns that), and it is built by concatenation
+        # so no second UUID-shaped literal enters this source.
         with open(os.path.join(d, "c.md"), "w") as fh:
-            fh.write("3d79ba82-1111-2222-3333-444444444444\n")
+            fh.write("34413f8d-" "0000-0000-0000-000000000000\n")
         check(d, ["report.md"], True, "complete-but-wrong-fails")
         os.remove(os.path.join(d, "c.md"))
         # Cell 5: justified-unrecovered placeholder does not count as recovery
@@ -214,6 +260,23 @@ def selftest():
         with open(os.path.join(d, "e.md"), "w") as fh:
             fh.write("beae0b1f-\u2026 and placeholder beae0b1f-0000-0000-0000-000000000000\n")
         check(d, ["report.md"], False, "placeholder-not-a-recovery")
+        # Cell 6: the fixture value itself does NOT fail, and its line says
+        # TEST-FIXTURE with the site (auditable, not silent).
+        with open(os.path.join(d, "f.md"), "w") as fh:
+            fh.write("3d79ba82-" "1111-2222-3333-444444444444\n")
+        lines, _ = run_check(d, ["report.md"])
+        assert any(l.startswith("TEST-FIXTURE 3d79ba82-") and l.endswith("at f.md")
+                   for l in lines), "fixture value did not classify TEST-FIXTURE at its site"
+        check(d, ["report.md"], False, "fixture-value-passes")
+        os.remove(os.path.join(d, "f.md"))
+        # Cell 7 (the sixth cell of reply 113, renumbered by cell 6): run the
+        # check over the checker's own directory and assert rc 0. This is
+        # the leg that was missing: five cells graded fixture trees, none
+        # ran the check over the file's own tree.
+        own_dir = os.path.dirname(os.path.abspath(__file__))
+        lines, failed = run_check(own_dir, [])
+        assert not failed, "self-scan failed: %s" % [l for l in lines if l.startswith("FAIL")]
+        passed.append("self-scan-passes")
     for label in passed:
         print("SELFTEST PASS %s" % label)
     print("SELFTEST: %d cells passed, 0 failed" % len(passed))