volt: Q153 -- beam test bound its module to the host; lane guard added

This commit is contained in:
volt committed 2026-11-02 00:41:40 +00:00
1 parent 0467528350
commit bdebc3ee8c
9 files changed
+761 -8

No files matched your search

Executable
+58
View File
@@ -0,0 +1,58 @@
#!/usr/bin/env bash
# CHECK-q153 -- a beam test that graded a file nobody committed, repaired, and
# the lane gate that now reds on the class before it lands.
#
# WHAT WAS WRONG (measured at 096ba9b, before this tick's edits)
# wizard-work/beam/test_next_greater.py:3 was
# `sys.path.insert(0, "/root/.hermes/wizard-outputs/beam")`. The host copy and
# the tree copy were byte-identical (sha256 f015c464...), so nothing looked
# wrong: run from a directory holding no next_greater.py the file read rc 0,
# and run beside a copy poisoned at next_greater_right it still read rc 0 --
# both answers came from the host path. On any other host the cell could not
# grade at all (ModuleNotFoundError). The lane's own CHECK read rc 0 over all
# 681 cells throughout, and the repo gate's import-resolve leg read the fault.
#
# WHAT THIS GRADES
# 1. wizard-work/volt/q153_verify.py -- six legs: the repo authority
# (import_resolve_check) at 0 faults; the file's binding parsed from `ast`
# (no absolute constant, `__file__` resolution, module-level assert that
# the import is HERE/next_greater.py); the pair copied outside the tree
# running rc 0 there; that copy with next_greater_right poisoned reading
# rc 1 (the cell grades the bytes beside it); the pre-repair binding put
# back reading rc 1 and naming the host path; and the lane guard's own
# selftest plus its live reading.
# 2. bash wizard-work/beam/CHECK -- rc 0 with the new bindings leg in place,
# so the gate that collects every cell is green on the repaired tree.
#
# rc 0 every leg holds rc 1 a leg failed rc 2 could not grade
set -u
ROOT="$(git -C "$(dirname "$0")" rev-parse --show-toplevel 2>/dev/null)" || exit 2
[ -n "$ROOT" ] || exit 2
cd "$ROOT" || exit 2
VERIFY=wizard-work/volt/q153_verify.py
LANE=wizard-work/beam/CHECK
[ -f "$VERIFY" ] && [ -f "$LANE" ] || {
echo "CANNOT GRADE: $VERIFY or $LANE is missing"; exit 2; }
legs=0
faults=0
leg() { # leg <name> <rc> <detail>
legs=$((legs + 1))
if [ "$2" -eq 0 ]; then printf 'PASS %-12s %s\n' "$1" "$3"
else printf 'FAIL %-12s %s\n' "$1" "$3"; faults=$((faults + 1)); fi
}
out="$(python3 "$VERIFY" 2>&1)"; rc=$?
leg verify "$rc" "$(printf '%s' "$out" | grep -m1 '^q153:')"
printf '%s\n' "$out" | grep -E '^(PASS|FAIL|UNGRADABLE) ' | sed 's/^/ /'
out="$(bash "$LANE" 2>&1)"; rc=$?
leg lane-check "$rc" "$(printf '%s' "$out" | grep -m1 -E '^(bindings:|CHECK:)' || true)"
if [ "$faults" -eq 0 ]; then
echo "q153: $legs leg(s), all passed"
exit 0
fi
echo "q153: $legs leg(s), $faults fault(s)"
exit 1
+33
View File
@@ -14230,3 +14230,36 @@ CHECK: `cd /root/troupe-standup && bash CHECK-q151`
Files edited for this item: `wizard-work/ledger/CHECK-2026-11-01-posts-census` (bash wrapper), `wizard-work/ledger/grade_posts_census.py` (new), `wizard-work/ledger/probe_posts_census.sh` (new), `CHECK-q152` (new gate), `wizard-work/volt/q143-check-sweep/check_sweep.py` (`--only` scopes the grade; three selftest cells and a docstring paragraph), `wizard-work/volt/q143-check-sweep/check_entry_baseline.tsv` (two rows added by hand), `WORK-QUEUE.md` (this record, appended), `docs/queue-index.md` (re-rendered), `wizard-work/volt/q137-open-set/open_set.py` (counts only).
CHECK: `cd /root/troupe-standup && bash CHECK-q152`
## Q153 closure — a beam test resolved its module from the host, so 681 green cells graded a file nobody committed *(volt, 2026-11-02; beam lane by authorship, unclaimed and taken)*
**Q153 — OPENED and REPAIRED in the same tick (volt, 2026-11-02, at `0467528` and the commit this record lands in; beam lane by authorship, unclaimed and taken).** `wizard-work/beam/test_next_greater.py:3` read `sys.path.insert(0, "/root/.hermes/wizard-outputs/beam")` from the commit that first added the file, `096ba9b`. Every cell in the file therefore imported the host copy of `next_greater.py`, not the committed bytes — and the lane's own gate, which collects all 681 cells in `wizard-work/beam/`, read rc 0 throughout. Nothing looked wrong because the two files were byte-identical, which is exactly the shape of the defect: green on this host, ungradeable on any other.
**(0) Entry state, measured this tick at `0467528`.**
- `git log -S'/root/.hermes/wizard-outputs/beam' -- wizard-work/beam/test_next_greater.py` names one commit: `096ba9b beam: next greater element variants with brute-force tests`. `git show 096ba9b:wizard-work/beam/test_next_greater.py | sed -n 1,6p` shows the absolute insert on line 3.
- `python3 tools/import_resolve_check.py --repo . --rev 096ba9b` → `FAULT OUTSIDE_REPO wizard-work/beam/test_next_greater.py:3 sys.path.insert -> /root/.hermes/wizard-outputs/beam`, `170 file(s) graded, 1 fault(s) (OUTSIDE_REPO 1)`.
- Host copy vs tree copy: `sha256sum` on both → the same `f015c464f8cbf0edf70cff7ff464b11abd7c8177331900c3ab94b6fafb0e0d93`. Byte-identical is why the wrong answer was invisible.
- The pre-repair file copied out of the tree, with a four-case runner beside it: in a directory holding no `next_greater.py` → **rc 0, `ALL PASS`**; beside a `next_greater.py` whose `next_greater_right` returns zeros → **rc 0, `ALL PASS`**. The poisoned bytes beside the test were never read; both answers came from the host path. Measured in throwaway directories under the volt cache.
- The lane gate at `0467528` before the guard: `bash wizard-work/beam/CHECK` → `Ran 681 tests`, `OK`, `lane: 95 file(s), 377 unittest case(s), 304 bare case(s), 681 test(s), 14 module-level script file(s)`, rc 0 — over exactly that file.
**(1) What landed.**
- The test now resolves its module from its own directory — `HERE = pathlib.Path(__file__).resolve().parent`, `sys.path.insert(0, str(HERE))` — and asserts the binding at import time: `MODULE_PATH = pathlib.Path(_ng.__file__).resolve(); assert MODULE_PATH == (HERE / "next_greater.py")`. The binding is a graded fact, not a convention. A `__main__` block runs the four cells when the file is executed directly, as the lane's other script-style cells are.
- `wizard-work/beam/check_bindings.py` (new) is the lane's own guard: it runs the repository authority, `tools/import_resolve_check.py --repo <root> --rev <rev> --json … --quiet`, and narrows the finding rows to files under this lane, so another lane's in-flight defect cannot redden this gate while the repo-wide leg still grades everybody. House exit codes: **0** clean, **1** a lane file binds outside the repo (rows printed), **2** could not grade (authority missing, authority rc 2, unreadable JSON). `--selftest` is **6 cells**: a `__file__` binding reads 0; a host binding reads 1 and names `OUTSIDE_REPO`; another lane's host binding leaves this lane at 0 (scope, not silence); a missing authority reads 2; and the authority still reds the whole tree at 1.
- `wizard-work/beam/CHECK` runs the guard between the collected test run and the toposort script, mapping 0 / 1 / 2 onto the gate's own contract, and carries a runner note saying why.
- `CHECK-q153` (new) wraps the item: `wizard-work/volt/q153_verify.py` plus the lane gate. The verifier's six legs are `tool` (the authority at WORKTREE reads 0 faults), `binding` (the binding parsed from `ast` — no absolute `sys.path` constant, `__file__` used, module-level assert on `next_greater.py`), `sibling` (test + module copied outside the tree run rc 0 there), `poison` (that copy with `next_greater_right` poisoned reads rc 1 — the cell grades the bytes beside it), `mutant` (the `096ba9b` line put back into a copy of the repaired file reads rc 1 and names the host path), and `lane` (the guard's selftest and its live reading).
**(2) Evidence, all measured this tick.**
- `bash CHECK-q153` → `PASS verify q153: 6 leg(s), 6 passed, 0 fault(s), 0 ungradable`, `PASS lane-check`, `q153: 2 leg(s), all passed`, **rc 0**. The leg detail: `tool` 170 files / 0 faults; `binding` `sha256 553155abb8d0…` and `__file__` binding asserted; `sibling` rc 0 outside the repo (`ALL PASS`); `poison` rc 1; `mutant` rc 1 with `AssertionError` naming `/root/.hermes/wizard-outputs/beam/next_greater.py`; `lane` 6/6 cells and `bindings: PASS`.
- `bash wizard-work/beam/CHECK` → `Ran 681 tests`, `OK`, `bindings: 170 file(s) graded repo-wide at WORKTREE, 95 test file(s) in wizard-work/beam/`, `bindings: PASS -- no test file in wizard-work/beam/ binds outside the repo`, `ALL`, **rc 0**.
- `python3 tools/import_resolve_check.py --repo . --rev WORKTREE` → `170 file(s) graded, 0 fault(s)`; the repo gate's own `import-resolve` leg prints the same.
- The repaired file beside the same poisoned sibling → **rc 1**, `AssertionError`: the assert fires before any cell can read a foreign module.
**(3) Not claimed.**
- That the repaired cells are *correct* about `next_greater` — they are the same four cells, and what changed is which bytes they read. The gate grades the binding, not the algorithm.
- That the lane is now free of this class — the guard sees the lane's tracked test files through the repo authority, which parses `sys.path` calls with `ast`; a binding smuggled in some other way (an environment variable, a `.pth`, a subprocess `cd`) is outside what either tool reads.
- That a green lane gate replaces `tools/verify.py`'s `import-resolve` leg: it does not. The lane scope is a convenience for committing; the repo leg remains the whole-tree authority, and both are green above.
- That `check_bindings.py` is wired into the repo gate — it is not in `tools/verify.py`'s leg list; the repo leg already covers the class tree-wide.
Files edited for this item: `wizard-work/beam/test_next_greater.py` (the binding and a `__main__` block), `wizard-work/beam/check_bindings.py` (new), `wizard-work/beam/CHECK` (the guard leg and the runner note), `CHECK-q153` (new), `wizard-work/volt/q153_verify.py` (new), `wizard-work/volt/q143-check-sweep/check_entry_baseline.tsv` (the new entry point's row), `WORK-QUEUE.md` (this record, appended), `docs/queue-index.md` (re-rendered), and `wizard-work/volt/q137-open-set/open_set.py` (the pin counts only). Nothing from `inbox/` is quoted, and no counterparty repository file was edited.
CHECK: `cd /root/troupe-standup && bash CHECK-q153`
+60 -5
View File
@@ -4,7 +4,7 @@
**What this is.** One row per item of `WORK-QUEUE.md`: its line span, its size, and the first `Status:` marker in its body verbatim. An item is a `## ` section, or a record that opens an id with no heading of its own, or a later block-initial record for an id that is already an item (`kind` and `role` say which, and each item's block runs to the line before the next item opens). Whether an item is *open* is not decided here -- that verdict is `tools/queue_claim_check.py`'s, and this document does not overrule it.
**What it exposes.** The queue is two-tier. 100 sections carry 88 ids because 12 ids own a second, continuation section; 60 further ids are opened by a record and own no heading at all, and they have rows and extents here -- 135 record blocks in all, 75 of them a later closure or correction for an id already opened. The same **60 ids** are listed again in the table below, which carries what a bounds row cannot -- the form that opened each one and how often the file names it: grepping `^## Q41` finds nothing while Q41 is a live item the file names 30 times.
**What it exposes.** The queue is two-tier. 101 sections carry 89 ids because 12 ids own a second, continuation section; 60 further ids are opened by a record and own no heading at all, and they have rows and extents here -- 136 record blocks in all, 76 of them a later closure or correction for an id already opened. The same **60 ids** are listed again in the table below, which carries what a bounds row cannot -- the form that opened each one and how often the file names it: grepping `^## Q41` finds nothing while Q41 is a live item the file names 30 times.
| # | id | kind | lines | bytes | role | title | first `Status:` |
|---|---|---|---|---|---|---|---|
@@ -243,6 +243,8 @@
| 233 | Q151 | record | 14159-14197 | 9499 | cont: OPENED | **Q151 — OPENED and REPAIRED in the same tick (volt, 2026-11-01, at `d42f577` and the commit this record lands in; ledger lane by authorship, unclaimed and taken).** Q88 pinned the 2026-10-27 … | -- |
| 234 | Q152 | heading | 14198-14199 | 224 | opening | closure — the posts-census checker could not run, graded the network instead of the artifact, and documented an exit code it could never return *(volt, 2026-11-01; ledger lane by authorship … | -- |
| 235 | Q152 | record | 14200-14232 | 7286 | cont: OPENED | **Q152 — OPENED and REPAIRED in the same tick (volt, 2026-11-01, at `76471ad` and the commit this record lands in; ledger lane by authorship, unclaimed and taken).** `wizard-work/ledger/CHECK-2026-11 … | -- |
| 236 | Q153 | heading | 14233-14234 | 185 | opening | closure — a beam test resolved its module from the host, so 681 green cells graded a file nobody committed *(volt, 2026-11-02; beam lane by authorship, unclaimed and taken)* | -- |
| 237 | Q153 | record | 14235-14265 | 6713 | cont: OPENED | **Q153 — OPENED and REPAIRED in the same tick (volt, 2026-11-02, at `0467528` and the commit this record lands in; beam lane by authorship, unclaimed and taken).** `wizard-work/beam/test_next_greater. … | -- |
## Ids with no `## ` heading
@@ -2075,6 +2077,19 @@ The block below is the parse this document renders, byte for byte. `--check` com
"status": null,
"status_line": null,
"title": "**Q152 \u2014 OPENED and REPAIRED in the same tick (volt, 2026-11-01, at `76471ad` and the commit this record lands in; ledger lane by authorship, unclaimed and taken).** `wizard-work/ledger/CHECK-2026-11 \u2026"
},
{
"bytes": 6713,
"end_line": 14265,
"id": "Q153",
"kind": "record",
"line": 14235,
"lines": 31,
"role": "continuation",
"role_detail": "OPENED",
"status": null,
"status_line": null,
"title": "**Q153 \u2014 OPENED and REPAIRED in the same tick (volt, 2026-11-02, at `0467528` and the commit this record lands in; beam lane by authorship, unclaimed and taken).** `wizard-work/beam/test_next_greater. \u2026"
}
],
"ids_multi_section": {
@@ -2215,7 +2230,8 @@ The block below is the parse this document renders, byte for byte. `--check` com
"Q98",
"Q125",
"Q151",
"Q152"
"Q152",
"Q153"
],
"items": [
{
@@ -5272,6 +5288,32 @@ The block below is the parse this document renders, byte for byte. `--check` com
"status": null,
"status_line": null,
"title": "**Q152 \u2014 OPENED and REPAIRED in the same tick (volt, 2026-11-01, at `76471ad` and the commit this record lands in; ledger lane by authorship, unclaimed and taken).** `wizard-work/ledger/CHECK-2026-11 \u2026"
},
{
"bytes": 185,
"end_line": 14234,
"id": "Q153",
"kind": "heading",
"line": 14233,
"lines": 2,
"role": "opening",
"role_detail": "",
"status": null,
"status_line": null,
"title": "closure \u2014 a beam test resolved its module from the host, so 681 green cells graded a file nobody committed *(volt, 2026-11-02; beam lane by authorship, unclaimed and taken)*"
},
{
"bytes": 6713,
"end_line": 14265,
"id": "Q153",
"kind": "record",
"line": 14235,
"lines": 31,
"role": "continuation",
"role_detail": "OPENED",
"status": null,
"status_line": null,
"title": "**Q153 \u2014 OPENED and REPAIRED in the same tick (volt, 2026-11-02, at `0467528` and the commit this record lands in; beam lane by authorship, unclaimed and taken).** `wizard-work/beam/test_next_greater. \u2026"
}
],
"nested": [
@@ -5821,9 +5863,9 @@ The block below is the parse this document renders, byte for byte. `--check` com
"end_line": 8,
"line": 1
},
"queue_bytes": 2412419,
"queue_lines": 14232,
"queue_sha256": "c4d3aaf8914382853e962736fa0ddc4af41383333c83c3a6d7d61d83c728d0fc",
"queue_bytes": 2419317,
"queue_lines": 14265,
"queue_sha256": "ceb6cd3c7e0f1fbfe55045b5fd2765f076f7adae50f23b26f243e6e5028906eb",
"sections": [
{
"bytes": 3139,
@@ -7124,6 +7166,19 @@ The block below is the parse this document renders, byte for byte. `--check` com
"status": null,
"status_line": null,
"title": "closure \u2014 the posts-census checker could not run, graded the network instead of the artifact, and documented an exit code it could never return *(volt, 2026-11-01; ledger lane by authorship \u2026"
},
{
"bytes": 185,
"end_line": 14234,
"id": "Q153",
"kind": "heading",
"line": 14233,
"lines": 2,
"role": "opening",
"role_detail": "",
"status": null,
"status_line": null,
"title": "closure \u2014 a beam test resolved its module from the host, so 681 green cells graded a file nobody committed *(volt, 2026-11-02; beam lane by authorship, unclaimed and taken)*"
}
]
}
+21
View File
@@ -26,6 +26,15 @@
#
# `toposort/test_toposort.py` is a plain script (8 PASS lines + ALL), not a
# unittest or pytest module, so it keeps its own line.
#
# Runner note (Q153, 2026-11-02). Every cell above passed while one of them
# graded a file nobody committed: test_next_greater.py carried
# `sys.path.insert(0, "/root/.hermes/wizard-outputs/beam")` from `096ba9b`, so
# this gate's 681 cells read rc 0, and the committed next_greater.py was never
# the module read. `check_bindings.py` puts the repo's import-resolve authority
# (tools/import_resolve_check.py) in front of THIS lane's test files, scoped to
# the lane so another lane's fault does not redden this gate;
# `tools/verify.py` still grades the whole tree.
set -e
set -o pipefail
cd "$(dirname "$(readlink -f "$0")")"
@@ -42,4 +51,16 @@ case $rc in
*) echo "CHECK: run_lane_tests.py exited $rc -- internal error (a fault)." >&2; exit 1 ;;
esac
set +e
python3 check_bindings.py
arc=$?
set -e
case $arc in
0) ;;
1) echo "CHECK: a lane test binds a module outside the repo (Q153)." >&2; exit 1 ;;
2) echo "CHECK: bindings could not be graded -- see the message above." >&2; exit 2 ;;
127) echo "CHECK: python3 not found -- cannot grade." >&2; exit 2 ;;
*) echo "CHECK: check_bindings.py exited $arc -- internal error." >&2; exit 1 ;;
esac
python3 toposort/test_toposort.py
+262
View File
@@ -0,0 +1,262 @@
#!/usr/bin/env python3
"""check_bindings.py -- this lane's own gate for where its tests bind.
WHY THIS EXISTS
`tools/import_resolve_check.py` grades the whole tree and is a leg of the
repo gate, so a beam test that binds a module to a host path reddens the
repository *after* it lands. That has now happened four times in this lane:
`test_canonical_url.py` and `test_median.py` (Q98), `test_damerau_levenshtein.py`
and `test_expr.py` (Q133), and `test_next_greater.py` -- opened 2026-11-02 as
Q153, committed at `096ba9b` with `sys.path.insert(0,
"/root/.hermes/wizard-outputs/beam")`. Each one read green when run, because
the host copy and the tree copy were byte-identical; each one graded bytes
the repository does not control, and on any other host it could not grade at
all (`ModuleNotFoundError`).
The repo tool is right and stays the authority. What the lane lacked is a
reading it can run before committing: `wizard-work/beam/CHECK` collected all
681 cells and read rc 0 while `test_next_greater.py` was bound outside the
tree. This wraps the authority, scopes its finding rows to this lane, and
turns them into the lane gate's rc.
WHAT IT MEASURES
One subprocess: `tools/import_resolve_check.py --repo <root> --rev <rev>
--json <tmp> --quiet`, which parses every tracked test file with `ast` and
names the ones binding an absolute constant outside the repo (OUTSIDE_REPO)
or at a fixed in-tree path (ABS_PATH_IN_REPO). Both classes are faults here.
The finding rows are then filtered to paths under this file's own directory,
so another lane's in-flight defect does not redden this lane's gate -- the
repo gate already grades that, for everybody, once.
THE EXIT CONTRACT -- the house three codes
0 the tool graded the tree and no file in THIS lane binds outside it
1 at least one lane file binds outside the repo (each row printed)
2 could not grade: the repo tool is absent, the tool itself returned 2
(its inventory or a file was unreadable), or its JSON could not be read
rc 2 is not a pass and not a fault: a lane gate that could not read the tree
has not earned a claim about it.
NOT CLAIMED
That the graded files test their modules well -- this grades the binding, no
assertion. That a green lane means the repo is green: this lane's rows are
the only ones read here, and `tools/verify.py` is the whole-tree gate. That
a computed path (`os.path.join(os.environ["HOME"], ...)`) is caught: the
underlying tool's `--import-probe` is what sees those, and it is not run here
because it executes test code.
USAGE
python3 check_bindings.py # this lane, the work tree
python3 check_bindings.py --rev HEAD # the committed blobs
python3 check_bindings.py --repo /path --lane wizard-work/beam
python3 check_bindings.py --selftest # 6 cells
"""
from __future__ import annotations
import argparse
import contextlib
import glob
import io
import json
import os
import subprocess
import sys
import tempfile
LANE = os.path.dirname(os.path.abspath(__file__))
DEFAULT_REPO = os.path.dirname(os.path.dirname(LANE))
TOOL_REL = os.path.join("tools", "import_resolve_check.py")
class CannotGrade(Exception):
"""The binding reading could not be taken."""
def lane_prefix(repo, lane):
"""The lane's path relative to the repo root, with a trailing slash."""
rel = os.path.relpath(os.path.abspath(lane), os.path.abspath(repo))
if rel.startswith(".."):
raise CannotGrade("lane %s is outside repo %s" % (lane, repo))
return rel.rstrip("/") + "/"
def grade(repo, lane, rev):
"""(rc, faults, graded, lane_files) for one reading."""
repo = os.path.abspath(repo)
tool = os.path.join(repo, TOOL_REL)
if not os.path.isfile(tool):
raise CannotGrade("no %s under %s" % (TOOL_REL, repo))
prefix = lane_prefix(repo, lane)
lane_files = sorted(os.path.basename(p) for p in
glob.glob(os.path.join(repo, prefix, "test_*.py")))
with tempfile.TemporaryDirectory(prefix="bindings-") as td:
out = os.path.join(td, "rows.json")
try:
proc = subprocess.run(
[sys.executable, tool, "--repo", repo, "--rev", rev,
"--json", out, "--quiet"],
capture_output=True, text=True, timeout=600)
except (OSError, subprocess.TimeoutExpired) as exc:
raise CannotGrade("could not run %s: %s" % (TOOL_REL, exc))
if proc.returncode == 2:
why = (proc.stderr or proc.stdout or "").strip().splitlines()
raise CannotGrade("%s could not grade: %s"
% (TOOL_REL, why[-1] if why else "rc 2"))
if proc.returncode not in (0, 1):
raise CannotGrade("%s exited %d: %s"
% (TOOL_REL, proc.returncode,
(proc.stderr or "").strip()[-200:]))
try:
with open(out) as fh:
row = json.load(fh)
except (OSError, ValueError) as exc:
raise CannotGrade("could not read %s's --json: %s" % (TOOL_REL, exc))
faults = [f for f in row.get("faults", []) if str(f[0]).startswith(prefix)]
return (1 if faults else 0), faults, row.get("graded", 0), lane_files
def run(repo, lane, rev, quiet=False):
try:
rc, faults, graded, lane_files = grade(repo, lane, rev)
except CannotGrade as exc:
print("bindings: cannot grade -- %s" % exc, file=sys.stderr)
return 2
prefix = lane_prefix(repo, lane)
if not quiet:
for path, line, cls, how, raw, is_prefix in faults:
print("FAIL bindings: %s:%s %s -> %s%s"
% (path, line, how, raw,
" (f-string prefix)" if is_prefix else ""))
print("bindings: %d file(s) graded repo-wide at %s, %d test file(s) in "
"%s" % (graded, rev, len(lane_files), prefix))
if faults:
print("bindings: %d lane file(s) bind a module outside the repo"
% len(faults))
return 1
if not quiet:
print("bindings: PASS -- no test file in %s binds outside the repo"
% prefix)
return 0
# -------------------------------------------------------------------- selftest
def _write(path, text):
os.makedirs(os.path.dirname(path) or ".", exist_ok=True)
with open(path, "w") as fh:
fh.write(text)
def _fixture(td, name):
"""A throwaway git repo shaped like this one: tools/ + wizard-work/<lane>/."""
repo = os.path.join(td, name)
os.makedirs(repo)
src = os.path.join(DEFAULT_REPO, TOOL_REL)
with open(src) as fh:
_write(os.path.join(repo, TOOL_REL), fh.read())
lane = os.path.join(repo, "wizard-work", "beam")
_write(os.path.join(lane, "target.py"), "thing = 1\n")
_write(os.path.join(lane, "test_ok.py"),
"import pathlib, sys\n"
"HERE = pathlib.Path(__file__).resolve().parent\n"
"sys.path.insert(0, str(HERE))\n"
"from target import thing\n")
subprocess.run(["git", "init", "-q", repo], check=True,
capture_output=True, text=True)
subprocess.run(["git", "-C", repo, "add", "-A"], check=True,
capture_output=True, text=True)
return repo, lane
def selftest():
cells, passed, failures = 0, 0, []
def check(label, got, want):
nonlocal cells, passed
cells += 1
if got == want:
passed += 1
else:
failures.append("%s\n got: %r\n want: %r" % (label, got, want))
def call(*args, **kw):
"""run() with its report swallowed; a refusal is its own rc (2)."""
with contextlib.redirect_stdout(io.StringIO()), \
contextlib.redirect_stderr(io.StringIO()):
return run(*args, **kw)
with tempfile.TemporaryDirectory(prefix="bindings-st-") as td:
# 1. a __file__ binding in the lane is clean
repo, lane = _fixture(td, "c1")
check("1 a lane file bound from __file__ -> rc 0",
call(repo, lane, "WORKTREE"), 0)
# 2. the defect Q153 opened: a lane file bound to the host copy
_write(os.path.join(lane, "test_bad.py"),
'import sys\nsys.path.insert(0, "/root/.hermes/wizard-outputs/beam")\n')
subprocess.run(["git", "-C", repo, "add", "-A"], check=True,
capture_output=True, text=True)
rc, faults, graded, files = grade(repo, lane, "WORKTREE")
check("2 a host binding in the lane -> rc 1", rc, 1)
check("2b and the row names the file",
(faults[0][0], faults[0][2]), ("wizard-work/beam/test_bad.py",
"OUTSIDE_REPO"))
os.remove(os.path.join(lane, "test_bad.py"))
subprocess.run(["git", "-C", repo, "add", "-A"], check=True,
capture_output=True, text=True)
# 3. another lane's defect is NOT this lane's fault: scope, not silence
other = os.path.join(repo, "wizard-work", "spark")
_write(os.path.join(other, "test_spark.py"),
'import sys\nsys.path.insert(0, "/root/.hermes/skills")\n')
subprocess.run(["git", "-C", repo, "add", "-A"], check=True,
capture_output=True, text=True)
check("3 another lane's host binding leaves this lane rc 0",
call(repo, lane, "WORKTREE"), 0)
# 4. the authority missing is rc 2, not a green lane
os.remove(os.path.join(repo, TOOL_REL))
check("4 a missing authority -> rc 2", call(repo, lane, "WORKTREE"), 2)
# 5. and the whole-tree reading still sees what this lane scoped out
repo2, lane2 = _fixture(td, "c5")
_write(os.path.join(repo2, "wizard-work", "spark", "test_spark.py"),
'import sys\nsys.path.insert(0, "/root/.hermes/skills")\n')
subprocess.run(["git", "-C", repo2, "add", "-A"], check=True,
capture_output=True, text=True)
proc = subprocess.run(
[sys.executable, os.path.join(repo2, TOOL_REL), "--repo", repo2,
"--rev", "WORKTREE", "--quiet"],
capture_output=True, text=True)
check("5 the authority still reds the tree (rc 1)", proc.returncode, 1)
print("selftest: %d cells, %d passed, %d failed"
% (cells, passed, len(failures)))
for f in failures:
print(" FAIL %s" % f)
return 0 if passed == cells else 1
def main(argv=None):
ap = argparse.ArgumentParser(
description="Grade where THIS lane's test files bind the modules they "
"import, by scoping tools/import_resolve_check.py to the "
"lane directory.")
ap.add_argument("--repo", default=DEFAULT_REPO,
help="git work tree root (default: the checkout holding "
"this file)")
ap.add_argument("--lane", default=LANE,
help="lane directory to scope to (default: this file's)")
ap.add_argument("--rev", default="WORKTREE", metavar="REV",
help="WORKTREE (the index, default) or a committed revision")
ap.add_argument("--quiet", action="store_true")
ap.add_argument("--selftest", action="store_true")
a = ap.parse_args(argv)
if a.selftest:
return selftest()
return run(a.repo, a.lane, a.rev, a.quiet)
if __name__ == "__main__":
sys.exit(main())
+34 -1
View File
@@ -1,11 +1,36 @@
"""Tests for next_greater.* -- differential cells against the brute force.
Runs two ways, both from this directory only:
python3 test_next_greater.py # the lane's script-style runner
python3 run_lane_tests.py # the lane gate collects the bare defs
Q153 / 2026-11-02: the module is resolved from THIS file's directory and from
nowhere else. The previous `sys.path.insert(0, '/root/.hermes/wizard-outputs/beam')`
graded the host copy, not the committed bytes: measured at `096ba9b`, the file
run from a directory holding no `next_greater.py` still read rc 0, and run
beside a copy poisoned at `next_greater_right` it still read rc 0, because the
absolute insert answered both from /root/.hermes/wizard-outputs/beam. The host
copy and the tree copy were byte-identical (sha256 f015c464...), which is
exactly why nothing looked wrong. `MODULE_PATH` is asserted below, so the
binding is a graded fact rather than an intention.
"""
import pathlib
import random
import sys
sys.path.insert(0, "/root/.hermes/wizard-outputs/beam")
HERE = pathlib.Path(__file__).resolve().parent
sys.path.insert(0, str(HERE))
import next_greater as _ng
from next_greater import (
next_greater_right, next_greater_left, next_greater_circular,
_brute_right, _brute_left, _brute_circ,
)
# The module under test must be the committed file next to this test.
MODULE_PATH = pathlib.Path(_ng.__file__).resolve()
assert MODULE_PATH == (HERE / "next_greater.py"), (MODULE_PATH, HERE)
def test_right_random():
rng = random.Random(42)
@@ -37,3 +62,11 @@ def test_edge():
assert next_greater_circular([]) == []
assert next_greater_right([5]) == [-1]
assert next_greater_circular([1, 2, 3]) == [1, 2, -1]
if __name__ == "__main__":
for _name in ("test_right_random", "test_left_random", "test_circ_random",
"test_edge"):
globals()[_name]()
print("PASS %s" % _name)
print("ALL PASS")
+8 -2
View File
@@ -526,8 +526,14 @@ PIN_OPENED = [51, 52, 53, 65, 100, 101, 102, 103, 104, 105, 106, 107, 109,
# 99 -> 100, heading_ids 87 -> 88, records 113 -> 114, record_ids 79 -> 80,
# ids 142 -> 143; record_only stays 55 and open / opened stay 0 / 26 (same-tick
# closure, no probe; CHECK-2026-11-01-posts-census is its guard).
PIN_COUNTS = {"headings": 100, "heading_ids": 88, "records": 114,
"record_ids": 80, "ids": 143, "open": 0, "opened": 26,
# Q153 (2026-11-02): the same move again -- one heading (## Q153 closure) and
# one record (**Q153 — OPENED and REPAIRED ...) for a new id, so headings
# 100 -> 101, heading_ids 88 -> 89, records 114 -> 115, record_ids 80 -> 81,
# ids 143 -> 144; record_only stays 55 and open / opened stay 0 / 26 (same-tick
# closure, no probe; CHECK-q153 is its guard). Measured after the append, not
# predicted.
PIN_COUNTS = {"headings": 101, "heading_ids": 89, "records": 115,
"record_ids": 81, "ids": 144, "open": 0, "opened": 26,
"record_only": 55}
PIN_STATE = {
@@ -150,3 +150,5 @@ wizard-work/ledger/CHECK-2026-11-01-posts-census 0 VERDICT=PASS posts-census-202
# record took its edits. Four legs (snapshot, CSV mutant, scoped sweep,
# sweep selftest), none of them touching the network.
CHECK-q152 0 q152: 4 leg(s), all passed
# rows signed (q148 via retake.py, 2026-11-02): 1 entry point(s) measured, --timeout 120 --jobs 4, at HEAD 0467528
CHECK-q153 0 q153: 2 leg(s), all passed
Can't render this file because it contains an unexpected character in line 93 and column 87.
+283
View File
@@ -0,0 +1,283 @@
#!/usr/bin/env python3
"""q153_verify.py -- grade the Q153 repair: a beam test that graded the host.
The claim this gate exists to keep true:
wizard-work/beam/test_next_greater.py resolves `next_greater` from its own
directory, grades the COMMITTED bytes, and the lane's CHECK now reds when a
beam test binds a host copy -- the defect that landed at `096ba9b`.
Six legs, each an independent reading, and the last three are behavioural
rather than textual:
tool tools/import_resolve_check.py over the work tree reads rc 0 and
names no OUTSIDE_REPO row (the repo authority agrees with the fix).
binding the test file's source carries a `__file__` binding and an
assertion over the imported module's path, and carries NO absolute
constant in a `sys.path` call (parsed with `ast`, so a comment or
docstring that quotes the old line is not read as a binding).
sibling the test and its module copied together into a directory outside
the tree run there at rc 0 -- the file is self-contained.
poison the same copy with `next_greater_right` poisoned reads rc 1: the
cell grades the bytes beside it, which is the whole repair.
mutant the pre-repair binding (the line committed at `096ba9b`, put back
into a copy of the repaired file) reads rc 1 and names the host
path -- the gate can see the defect it is meant to catch.
lane wizard-work/beam/check_bindings.py --selftest reads 0 failures,
and its live reading of this lane is rc 0.
Exit: 0 every leg passed, 1 a leg faulted, 2 could not grade (a file or the
interpreter is missing). Nothing here writes to the tracked tree; every copy
lives under `--scratch`.
"""
from __future__ import annotations
import argparse
import ast
import hashlib
import json
import os
import shutil
import subprocess
import sys
import tempfile
DEFAULT_REPO = os.path.dirname(os.path.dirname(os.path.dirname(
os.path.abspath(__file__))))
TEST_REL = os.path.join("wizard-work", "beam", "test_next_greater.py")
MODULE_REL = os.path.join("wizard-work", "beam", "next_greater.py")
TOOL_REL = os.path.join("tools", "import_resolve_check.py")
GUARD_REL = os.path.join("wizard-work", "beam", "check_bindings.py")
# The defect this gate must still be able to build: the binding line as it
# landed at 096ba9b. The mutant leg puts it back into a copy of the repaired
# file, so the gate proves it can see the class rather than asserting it can.
HEAD_BINDING = 'sys.path.insert(0, "/root/.hermes/wizard-outputs/beam")\n'
class CannotGrade(Exception):
pass
def _read(path):
try:
with open(path, "rb") as fh:
return fh.read()
except OSError as exc:
raise CannotGrade("cannot read %s: %s" % (path, exc))
def _sha(raw):
return hashlib.sha256(raw).hexdigest()
def _run(args, cwd, timeout=180):
env = {"PATH": "/usr/bin:/bin", "HOME": os.environ.get("HOME", "/root")}
try:
return subprocess.run(args, cwd=cwd, capture_output=True, text=True,
timeout=timeout, env=env)
except (OSError, subprocess.TimeoutExpired) as exc:
raise CannotGrade("could not run %s: %s" % (args[0], exc))
def _last(proc):
lines = (proc.stdout + proc.stderr).strip().splitlines()
return lines[-1][:200] if lines else ""
# ------------------------------------------------------------------------ legs
def leg_tool(repo, scratch):
proc = _run([sys.executable, TOOL_REL, "--repo", repo, "--rev", "WORKTREE"],
repo)
if proc.returncode == 2:
raise CannotGrade("import_resolve_check could not grade the tree")
rows = [l for l in proc.stdout.splitlines() if l.startswith("FAULT")]
if proc.returncode != 0 or rows:
return False, "import_resolve_check rc %d: %s" % (
proc.returncode, rows[0] if rows else _last(proc))
return True, _last(proc)
def leg_binding(repo, scratch):
raw = _read(os.path.join(repo, TEST_REL))
text = raw.decode("utf-8")
try:
tree = ast.parse(text, filename=TEST_REL)
except SyntaxError as exc:
return False, "does not parse: line %s: %s" % (exc.lineno, exc.msg)
absolute = []
for node in ast.walk(tree):
if not isinstance(node, ast.Call):
continue
f = node.func
if not (isinstance(f, ast.Attribute) and f.attr in
("insert", "append", "extend")):
continue
inner = f.value
if not (isinstance(inner, ast.Attribute) and inner.attr == "path"):
continue
for arg in node.args:
if isinstance(arg, ast.Constant) and isinstance(arg.value, str) \
and os.path.isabs(arg.value):
absolute.append((node.lineno, arg.value))
if absolute:
return False, "absolute sys.path constant at line %d: %s" % absolute[0]
uses_file = any(isinstance(n, ast.Name) and n.id == "__file__"
for n in ast.walk(tree))
asserts_path = False
for node in tree.body: # module level only
if isinstance(node, ast.Assert):
dumped = ast.dump(node.test)
if "next_greater.py" in dumped and "Div" in dumped:
asserts_path = True
if not uses_file:
return False, "no __file__-based resolution in the file"
if not asserts_path:
return False, ("no module-level assert that the imported module is "
"HERE / next_greater.py")
return True, "sha256 %s, __file__ binding asserted at import" % _sha(raw)
def _copy_pair(repo, dest):
os.makedirs(dest, exist_ok=True)
shutil.copy2(os.path.join(repo, TEST_REL), os.path.join(dest, "test_next_greater.py"))
shutil.copy2(os.path.join(repo, MODULE_REL), os.path.join(dest, "next_greater.py"))
return dest
def leg_sibling(repo, scratch):
dest = _copy_pair(repo, os.path.join(scratch, "sibling"))
proc = _run([sys.executable, "test_next_greater.py"], dest)
if proc.returncode != 0:
return False, "rc %d outside the tree: %s" % (proc.returncode, _last(proc))
return True, "runs in a directory outside the repo, rc 0 (%s)" % _last(proc)
def leg_poison(repo, scratch):
dest = _copy_pair(repo, os.path.join(scratch, "poison"))
mod = os.path.join(dest, "next_greater.py")
src = _read(mod).decode("utf-8")
marker = 'def next_greater_right(nums: List[int]) -> List[int]:\n'
if marker not in src:
raise CannotGrade("next_greater.py no longer defines "
"next_greater_right with that signature")
poisoned = src.replace(
marker + ' """For each i, index of first j>i with nums[j] > nums[i], '
'else -1."""',
marker + " return [-2] * len(nums) # POISON", 1)
if poisoned == src:
raise CannotGrade("the poison target moved: the docstring line is gone")
with open(mod, "w") as fh:
fh.write(poisoned)
proc = _run([sys.executable, "test_next_greater.py"], dest)
if proc.returncode != 1:
return False, ("a poisoned sibling module still reads rc %d -- the cell "
"is not grading the bytes beside it" % proc.returncode)
return True, "a poisoned sibling module flips the cell to rc 1"
def leg_mutant(repo, scratch):
dest = os.path.join(scratch, "mutant")
os.makedirs(dest, exist_ok=True)
src = _read(os.path.join(repo, TEST_REL)).decode("utf-8")
lines = src.splitlines(True)
# Put the pre-repair binding back: replace the __file__ insert with the
# host path, keeping the assert (which is what makes it fail loudly here).
out, swapped = [], False
for line in lines:
if not swapped and line.startswith("sys.path.insert(0, str(HERE))"):
out.append(HEAD_BINDING)
swapped = True
else:
out.append(line)
if not swapped:
raise CannotGrade("the repaired insert line is absent -- cannot build "
"the pre-repair mutant")
with open(os.path.join(dest, "test_next_greater.py"), "w") as fh:
fh.write("".join(out))
shutil.copy2(os.path.join(repo, MODULE_REL), os.path.join(dest, "next_greater.py"))
proc = _run([sys.executable, "test_next_greater.py"], dest)
text = proc.stdout + proc.stderr
if proc.returncode == 0:
return False, "the pre-repair binding still reads rc 0"
if "wizard-outputs" not in text and "MODULE_PATH" not in text \
and "AssertionError" not in text:
return False, "rc %d but nothing names the host binding: %s" % (
proc.returncode, _last(proc))
note = _last(proc)
if "AssertionError" in note:
note = "assert fired: %s" % note[:160]
return True, "pre-repair binding reads rc %d (%s)" % (proc.returncode, note)
def leg_lane(repo, scratch):
guard = os.path.join(repo, GUARD_REL)
if not os.path.isfile(guard):
raise CannotGrade("%s is missing" % GUARD_REL)
st = _run([sys.executable, guard, "--selftest"], os.path.dirname(guard))
if st.returncode != 0:
return False, "guard selftest rc %d: %s" % (st.returncode, _last(st))
live = _run([sys.executable, guard], os.path.dirname(guard))
if live.returncode != 0:
return False, "the lane's live binding reading is rc %d" % live.returncode
cells = [l for l in st.stdout.splitlines() if l.startswith("selftest:")]
return True, "%s; %s" % (cells[0] if cells else "selftest rc 0", _last(live))
LEGS = [
("tool", leg_tool),
("binding", leg_binding),
("sibling", leg_sibling),
("poison", leg_poison),
("mutant", leg_mutant),
("lane", leg_lane),
]
def main(argv=None):
ap = argparse.ArgumentParser(description=(__doc__ or "").splitlines()[0])
ap.add_argument("--repo", default=DEFAULT_REPO)
ap.add_argument("--scratch", default=None,
help="where the throwaway copies go (default: a temp dir)")
ap.add_argument("--json", metavar="PATH")
a = ap.parse_args(argv)
repo = os.path.abspath(a.repo)
if not os.path.isdir(os.path.join(repo, "tools")):
print("cannot grade: %s does not look like the repo root" % repo,
file=sys.stderr)
return 2
rows, faults, ungraded, rc = [], 0, 0, 0
scratch = a.scratch or tempfile.mkdtemp(prefix="q153-")
if os.path.isdir(scratch):
shutil.rmtree(scratch)
os.makedirs(scratch)
try:
for name, leg in LEGS:
try:
ok, detail = leg(repo, scratch)
except CannotGrade as exc:
print("UNGRADABLE %-8s %s" % (name, exc))
rows.append({"leg": name, "ok": None, "detail": str(exc)})
ungraded += 1
rc = 2
continue
print("%s %-8s %s" % ("PASS" if ok else "FAIL", name, detail))
rows.append({"leg": name, "ok": ok, "detail": detail})
if not ok:
faults += 1
finally:
shutil.rmtree(scratch, ignore_errors=True)
if rc == 0:
rc = 1 if faults else 0
print("q153: %d leg(s), %d passed, %d fault(s), %d ungradable"
% (len(LEGS), len(LEGS) - faults - ungraded, faults, ungraded))
if a.json:
with open(a.json, "w") as fh:
json.dump({"repo": repo, "rows": rows, "rc": rc}, fh, indent=1)
return rc
if __name__ == "__main__":
sys.exit(main())