The queue: parking, the operator's order, blocked items (M2.2)
CI / checks (push) Skipped
CI / checks (pull_request) Successful in 3m26s

Several work items on the new loop (L1, D4):
- Switching parks the item on the workspace: its work is committed and its
  branch pushed; resuming checks the item's branch back out. Branches of
  open items are never set aside by a fresh workspace.
- Next item: the operator's "do #N next" first, otherwise the oldest.
- Issues with open Gitea dependencies are blocked until they close, checked
  at intake and on every round; status shows blocked items and the line.

Tests: queue unit tests, the property walk over several issues with
dependencies and a new parked_work_safe property (it catches parking
disabled), and an end-to-end test with two issues where B depends on A.

Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01PPQ14L3LikyefCMFBqEbKE
This commit is contained in:
dongarandClaude Opus 5.5 committed 2026-10-04 13:34:09 +00:00
1 parent d504588a57
commit cceb30c729
11 files changed
+657 -85

No files matched your search

+110 -15
View File
@@ -1,9 +1,10 @@
"""specs/loop.qnt's properties, checked against the real loop (M2.1, PR D).
The Quint model checks the design; this checks the code. A seeded random
walk interleaves what the operator and Gitea can do (go-aheads, answers,
comments, merges, closes, abandon, a crash and restart) with an adaptive
scripted model, and checks every property after every step. A failure prints
walk interleaves what the operator and Gitea can do (new issues, some
depending on others, go-aheads, answers, "do #N next", comments, merges,
closes, abandon, a crash and restart) with an adaptive scripted model, and
checks every property after every step. A failure prints
the seed and the steps, so it replays exactly.
"""
@@ -19,17 +20,25 @@ from test_loop_milestones import world as world # noqa: PLC0414 -- the fixture
from nisse_dev.cancel import Cancel
from nisse_dev.canonical import Fact, ToolCall
from nisse_dev.gitea import Comment
from nisse_dev.gitea import Comment, Issue
from nisse_dev.loop.agent import LoopAgent
from nisse_dev.loop.driver import Driver
from nisse_dev.loop.items import ANSWER, GO_AHEAD, MERGE, WAITING, loop_of
from nisse_dev.loop.items import (
ANSWER,
BLOCKED,
GO_AHEAD,
MERGE,
RUNNING,
WAITING,
loop_of,
)
from nisse_dev.oplog import OpLog
from nisse_dev.tools import Workspace
from nisse_dev.transport import InboundMessage, InMemoryTransport
from nisse_dev.worker import InlineWorker
SEEDS = range(32)
STEPS = 20
STEPS = 30
class PolicyModel:
@@ -55,6 +64,13 @@ class PolicyModel:
return [call("restate", text=f"restatement {self.n}")]
if "wait_for_review" in offered and self.rng.random() < 0.2:
return [call("wait_for_review", note=f"waiting {self.n}")]
if self.rng.random() < 0.15:
# Half done, then a question: work a switch must park.
return [
call("start_milestone", name=f"m{self.n}"),
call("write_file", path=f"f{self.n}.py", content=f"x = {self.n}\n"),
call("ask_human", question=f"halfway {self.n}?"),
]
merged = any("merged" in t for t in facts[-6:])
if merged and self.rng.random() < 0.8:
return [call("close_item", summary=f"done {self.n}")]
@@ -71,7 +87,8 @@ class Walk:
self.rng = random.Random(seed)
self.seed = seed
self.steps: list[str] = []
self.shown_at_go: list[int] = [] # the revision on screen at each go
self.shown_at_go: list[str] = [] # item:revision on screen at each go
self.next_issue = 15
self.agent, self.transport = self._agent()
def _agent(self) -> tuple[LoopAgent, InMemoryTransport]:
@@ -105,20 +122,30 @@ class Walk:
open_items = [i for i in self.items() if i.status == "open"]
# Weighted towards what an item can actually do, so review paths
# are walked often, not drowned out by intake and crashes.
weights = {"intake": 2 if not open_items else 1, "crash": 1}
for item in open_items:
loop = loop_of(item)
weights = {"intake": 2 if not open_items else 1, "crash": 1, "issue": 1}
item = self.rng.choice(open_items) if open_items else None
loop = loop_of(item) if item else None
if item and loop:
weights.update({"go": 3, "answer": 2, "abandon": 1, "go-stale": 1})
weights.update({"first": 1})
if loop.pr:
weights.update(
{"comment": 4, "merge": 2, "close-pr": 1, "comment-merge": 2}
)
action = self.rng.choices(list(weights), list(weights.values()))[0]
self.steps.append(action)
item = open_items[0] if open_items else None
loop = loop_of(item) if item else None
self.steps.append(f"{action}#{loop.issue}" if loop else action)
if action == "intake":
self.agent.intake()
elif action == "issue":
# A new issue, sometimes depending on one that is still open.
n, self.next_issue = self.next_issue, self.next_issue + 1
forge.issues.append(Issue(n, f"task {n}", f"do task {n}", "don", ()))
live = [i.number for i in forge.assigned_issues("", "") if i.number != n]
if live and self.rng.random() < 0.5:
forge.deps[n] = [self.rng.choice(live)]
self.agent.intake()
elif action == "first" and loop:
self.agent.handle(InboundMessage(DON, f"do #{loop.issue} next"), None)
elif action == "crash":
# Volatile state is lost: a new agent over the same store.
self.agent, self.transport = self._agent()
@@ -127,7 +154,7 @@ class Walk:
elif action in ("go", "answer", "abandon") and loop:
text = {"go": "go", "answer": "an answer", "abandon": "abandon"}[action]
if action == "go" and item.position == WAITING and loop.wait == GO_AHEAD:
self.shown_at_go.append(loop.rev)
self.shown_at_go.append(f"{item.id}:{loop.rev}")
self.transport.receive(DON, text, thread=loop.thread)
elif action == "go-stale" and loop and loop.rev > 1:
# A go sent while an older revision was on screen.
@@ -185,9 +212,21 @@ class Walk:
if isinstance(f.event, Fact)
and f.event.text.startswith("go-ahead on rev-")
]
assert all(rev in self.shown_at_go for rev in gos), (
assert all(f"{item.id}:{rev}" in self.shown_at_go for rev in gos), (
f"go_binds: {gos}: {where}"
)
# Blocked comes from external truth (L1): an item whose issue
# had an open dependency from the start never ran before it closed.
if forge.open_dependencies("don/sandbox", loop.issue):
assert item.position == BLOCKED or item.status == "closed", (
f"blocked_never_runs: #{loop.issue} {item.position}: {where}"
)
ran = [
e
for e in store.transcript(item.id, "driver").read()
if e.actor == "driver"
]
assert not ran, f"blocked_never_runs (driver): #{loop.issue}: {where}"
mine = [p for p in forge.prs.values() if p["head"].startswith(item.id)]
assert sum(p["state"] == "open" for p in mine) <= 1, (
f"milestone_after_merge: {where}"
@@ -199,6 +238,10 @@ class Walk:
assert not [
i for i in store.pending_intents() if i.work_item == item.id
], f"closed_items_leave_nothing (intent): {where}"
assert not [i for i in items if i.position == RUNNING], (
f"one_active (nothing left running): {where}"
)
self._parked_work_safe(where)
heads = [p["head"] for p in forge.prs.values()]
assert len(heads) == len(set(heads)), f"one_pr_per_milestone: {where}"
bare = self.world["bare"]
@@ -216,6 +259,58 @@ class Walk:
)
assert landed, f"review_fixes_land (PR {n}): {where}"
def _parked_work_safe(self, where: str) -> None:
"""L1, D4: nothing of a parked item lives only on the build VM: what
every branch of an open item that isn't on the workspace points at is
in the forge."""
root, bare = self.world["repo"].root, self.world["bare"]
current = _git(root, "rev-parse", "--abbrev-ref", "HEAD")
ids = [i.id for i in self.items() if i.status == "open"]
for line in _git(
root, "for-each-ref", "--format=%(refname:short)", "refs/heads"
).splitlines():
if line == current or not any(line.startswith(f"{i}-") for i in ids):
continue
local = _git(root, "rev-parse", line)
pushed = subprocess.run(
[
"git",
"--git-dir",
str(bare),
"cat-file",
"-e",
f"{local}^{{commit}}",
],
capture_output=True,
check=False,
)
assert pushed.returncode == 0, (
f"parked_work_safe: {line} isn't pushed: {where}"
)
# And no write is lost: every file an open item's driver wrote is on
# the workspace while it's that item's, or in the forge.
for item in self.items():
if item.status != "open":
continue
written = {
str(e.event.arguments.get("path"))
for e in self.world["store"].transcript(item.id, "driver").read()
if isinstance(e.event, ToolCall) and e.event.tool == "write_file"
}
here = current.startswith(f"{item.id}-")
for path in written:
if here and (root / path).exists():
continue
history = _git(bare, "log", "--all", "--format=%H", "--", path)
assert history, f"parked_work_safe: {path} of #{item.id} lost: {where}"
def _git(root: Path, *args: str) -> str:
where = ["--git-dir", str(root)] if root.suffix == ".git" else ["-C", str(root)]
return subprocess.run(
["git", *where, *args], capture_output=True, text=True, check=True
).stdout.strip()
class _Null:
def write(self, _: str) -> None: ...