beam: Tarjan SCC iterative, brute-force verified

tools/check_lint.py

@@ -26,7 +26,7 @@ WHY THIS EXISTS
 
   This tool grades the line so a fault is visible before it costs an iteration.
 
-WHAT IT MEASURES -- SEVEN CLASSES, EACH WITH ITS RULE PRINTED
+WHAT IT MEASURES -- EIGHT CLASSES, EACH WITH ITS RULE PRINTED
   MISSING      no line matches `^\\s*CHECK:`; the runner calls this RED at once.
   EMPTY        a CHECK line with nothing after the colon.
   TAUTOLOGY    the whole command is `true`, `:`, `exit 0`, `echo <literal>`,
@@ -35,6 +35,12 @@ WHAT IT MEASURES -- SEVEN CLASSES, EACH WITH ITS RULE PRINTED
                true, `:`, `exit 0`, `echo`, `printf`, or a terminal statement
                `; true` / `; :` / `; exit 0` / `; echo ...`. Every path returns
                0, so the runner's GREEN carries no information.
+  LABEL_PREFIX the command opens with a label, `word:` -- the line was written
+               `CHECK: selftest: bash CHECK-x` or `CHECK: line: bash CHECK-x`.
+               bash takes the whole first word as the command name, so the line
+               dies with `word:: command not found` (rc 127) before anything is
+               tested. Added 2026-11-02 after measuring the class on a live run:
+               the same line shape RED'd one of this lane's own iterations.
   TRAILING_CR  a literal carriage return survives at end of line; it reaches
                bash as part of the last word.
   PROSE_TAIL   prose follows the command: an em dash, ` -- `, a trailing
@@ -61,7 +67,7 @@ WHAT IT DOES NOT DO
 USAGE
   python3 check_lint.py --check "$CHECK"
   python3 check_lint.py --ralph-dir /root/.hermes/wizard-outputs/ralph
-  python3 check_lint.py --selftest          # 42 cells, one per class and edge
+  python3 check_lint.py --selftest          # 51 cells, one per class and edge
 """
 
 from __future__ import annotations
@@ -112,6 +118,22 @@ CATCH_ALL_RE = re.compile(
     r"|(?:(?<!&)(?<!\|)[;&]\s*(?:true|:|exit\s+0|echo\b|printf\b)(?:\s|$))"
 )
 
+# A leading label: `selftest: python3 v.py`, `line: python3 v.py`, `Step-2: ...`.
+# bash takes the WHOLE first word as the command name, so a label makes the line
+# exit 127 with `word:: command not found` before a single test runs. Measured
+# 2026-11-02 rather than reasoned: a ralph iteration died on
+# `bash: line 1: selftest:: command not found` while this rubric read the line
+# SOUND, and the runner's own 2026-10-27 forge log carries the sibling shape
+# (`line: python3 ...`) one iteration before that CHECK was rewritten without it.
+#
+# The rule is anchored to the FIRST word and needs no space after the colon
+# (`ok:true` is the same defect), and it deliberately does not match:
+#   `PATH=/a:/b python3 x.py`  -- the `=` breaks the word class before the colon
+#   `python3 x.py: --selftest` -- the colon is not in the first word
+#   `'note: x' cmd`            -- a quoted first token is not a bare word
+# The last two are pinned as SOUND cells so the boundary is measured, not assumed.
+LABEL_RE = re.compile(r"^(?P<label>[A-Za-z0-9][A-Za-z0-9_.+-]*):")
+
 PROSE_MARKERS = [
     (re.compile(r"\s+[\u2014\u2013]\s+"), "em dash"),
     (re.compile(r"\s--\s+"), "spaced --"),
@@ -128,7 +150,12 @@ PROSE_MARKERS = [
 # zero-exit constructs outrank the mechanical ones because they are the faults
 # that make a GREEN meaningless, while TRAILING_CR/PROSE_TAIL only make one
 # noisy (they break the run, loudly, and cost an iteration instead of lying).
-SEVERITY = ["EMPTY", "TAUTOLOGY", "CATCH_ALL", "TRAILING_CR", "PROSE_TAIL"]
+# LABEL_PREFIX sits with the zero-exit group in rank, not in kind: it cannot
+# return 0 either, but it fails the other way round -- the run REDs for a reason
+# that has nothing to do with what was tested, so it is the first thing a reader
+# should be told when a line carries both a label and a prose tail.
+SEVERITY = ["EMPTY", "TAUTOLOGY", "CATCH_ALL", "LABEL_PREFIX", "TRAILING_CR",
+            "PROSE_TAIL"]
 
 
 def prose_tail(core: str):
@@ -167,6 +194,14 @@ def classify(cmd: str | None):
         faults.append("CATCH_ALL")
         detail["catch_all_match"] = m_catch.group(0)
 
+    m_label = LABEL_RE.match(core.lstrip())
+    if m_label:
+        faults.append("LABEL_PREFIX")
+        detail["label"] = m_label.group("label")
+        # The cut point, printed the way PROSE_TAIL prints its own: what the
+        # line would say with the label removed, which is what bash needs.
+        detail["command_after_label"] = core.lstrip()[m_label.end():].strip()
+
     pt = prose_tail(core)
     if pt:
         faults.append("PROSE_TAIL")
@@ -302,6 +337,10 @@ def grade_run(run: dict) -> dict:
         cross = "GREEN_BY_CONSTRUCTION"
     elif final == "GREEN" and primary in ("PROSE_TAIL", "TRAILING_CR"):
         cross = "GREEN_DESPITE_FORMATTING"
+    # LABEL_PREFIX is deliberately in neither arm: a labelled line exits 127 at
+    # the shell, so a runner cannot have reported it GREEN. If one ever did, the
+    # run is reported with cross="none" and the class printed, rather than being
+    # folded into a bucket whose meaning it would dilute.
 
     per_iteration = []
     for n in sorted(iters):
@@ -393,6 +432,22 @@ FIXTURES = [
     ("CHECK: python3 x.py\n\nA CHECK: line is required.\n", "SOUND"),
     ("CHECK: python3 x.py\n\n CHECK: line \u2014 one shell command that verifies "
      "your work and exits 0\n", "PROSE_TAIL"),
+    # LABEL_PREFIX cells (2026-11-02, volt). The first two are the measured
+    # shapes: cell 1 is the line this lane's own ralph iteration died on
+    # (`bash: line 1: selftest:: command not found`), cell 2 is the one the
+    # runner's 2026-10-27 forge log carries, one iteration before that CHECK was
+    # rewritten without the label. The four controls pin the boundary of the
+    # rule: a colon that is not in the FIRST word, a colon inside an assignment,
+    # and a quoted first token are all SOUND.
+    ("CHECK: selftest: python3 tools/check_lint.py --selftest\n", "LABEL_PREFIX"),
+    ("CHECK: line: python3 tools/check_lint.py --selftest\n", "LABEL_PREFIX"),
+    ("CHECK: Verdict: bash CHECK-x\n", "LABEL_PREFIX"),          # capitalised
+    ("CHECK: step-2: python3 x.py\n", "LABEL_PREFIX"),           # hyphenated
+    ("CHECK: 2026-11-02: python3 x.py\n", "LABEL_PREFIX"),       # date label
+    ("CHECK: ok:true\n", "LABEL_PREFIX"),                        # no space after
+    ("CHECK: python3 x.py: --selftest\n", "SOUND"),              # colon, word 2
+    ("CHECK: PATH=/a:/b python3 x.py\n", "SOUND"),               # assignment
+    ("CHECK: 'note: x' cmd\n", "SOUND"),                         # quoted token
 ]
 
 EXTRACT_FIXTURES = [
@@ -435,6 +490,7 @@ def print_rules():
     print("  EMPTY        a CHECK line with nothing after the colon")
     print("  TAUTOLOGY    whole command is true/: /exit 0/echo/printf/sleep")
     print("  CATCH_ALL    a || arm or terminal ; statement that exits 0 regardless")
+    print("  LABEL_PREFIX a leading `word:` label -- bash exits 127 on it")
     print("  TRAILING_CR  a literal carriage return survives at end of line")
     print("  PROSE_TAIL   prose follows the command (CANDIDATE, cut point printed)")
     print("  SOUND        none of the above -- not a promise about its meaning")

tools/selftest_baseline.tsv

@@ -12,7 +12,7 @@
 append_only_audit.py	18	cells
 append_only_owners.py	15	cells
 census_row_delta.py	37	cells
-check_lint.py	42	cells
+check_lint.py	51	cells
 checkin_check.py	36	cells
 citation_hash_class.py	19	fixtures
 citation_resolve_check.py	14	fixtures