nisse-dev/src/nisse_dev/control.py

@@ -130,6 +130,8 @@ class MilestonePublisher(Protocol):
 
     def close_pull_request(self, index: int, note: str) -> None: ...
 
+    def published_pull_request(self, task_id: str, milestone: int) -> Any: ...
+
 
 class PlanBoard(Protocol):
     """The Gitea side of plan gating; the concrete one is in
@@ -323,6 +325,27 @@ class Orchestrator:
                 "restarted while waiting on you: "
                 + pending.get("prompt", "my last question still stands")
             )
+        elif state.state == EXECUTING and (published := self._published_pr()):
+            # A3 step 4: Gitea shows this milestone's PR open, so it was
+            # published and only PR_OPEN went unrecorded. Adopt the PR rather
+            # than ask, and never publish it again. Nothing watched it yet,
+            # so every comment on it is new.
+            milestone = int(self._data.get("milestone", 1))
+            self._data["pr"] = {
+                "number": published.number,
+                "url": published.url,
+                "milestone": milestone,
+                "last_comment": self._watch(
+                    "review", published.number, self.handle_review_comment, after=0
+                ),
+            }
+            self._unwatch("plan")
+            self._transition(PR_OPEN, cause=f"boot: milestone {milestone} published")
+            self._transport.send_message(
+                f"restarted after milestone {milestone} was published: PR "
+                f"#{published.number} is open ({published.url}); watching it for "
+                "review comments"
+            )
         elif state.state == PR_OPEN:
             # Waiting on review is a committed state like a gate (A3 step
             # 8): resume, and watch the PR again from its cursor.
@@ -980,6 +1003,20 @@ class Orchestrator:
             f"task on {project} abandoned{closed}. Say 'use <project>' to start another"
         )
 
+    def _published_pr(self) -> Any:
+        """The open PR for the current milestone, if Gitea shows one; None
+        when there is none or Gitea cannot say (then the restart question
+        is asked as before)."""
+        task_id = self._data.get("task_id")
+        if self._publisher is None or not task_id:
+            return None
+        try:
+            return self._publisher.published_pull_request(
+                str(task_id), int(self._data.get("milestone", 1))
+            )
+        except GiteaError:
+            return None
+
     def _return_to_pr_open(self, cause: str) -> None:
         self._transition(PR_OPEN, cause=cause)
         pr = self._data.get("pr") or {}

nisse-dev/src/nisse_dev/publish.py

@@ -84,6 +84,19 @@ class MilestonePublisher:
             cause=f"publishing milestone {milestone}",
         )
         self._git.push(self._remote, branch, token=self._token)
+        # Idempotent: if a PR for this branch is already open (a crash after
+        # opening it, or a response lost on the way back), the push above
+        # has updated it; adopt it rather than fail on a second.
+        existing = self.published_pull_request(task_id, milestone)
+        if existing is not None:
+            self._store.commit_intent(
+                intent_id, cause=f"adopted open PR #{existing.number}"
+            )
+            return PublishResult(
+                existing,
+                f"PR #{existing.number} was already open for this milestone; "
+                f"pushed to it: {existing.url}",
+            )
         pr = self._client.open_pull_request(
             self._repo,
             head=branch,
@@ -94,6 +107,17 @@ class MilestonePublisher:
         self._store.commit_intent(intent_id, cause=f"opened PR #{pr.number}")
         return PublishResult(pr, f"opened PR #{pr.number}: {pr.url}")
 
+    def published_pull_request(
+        self, task_id: str, milestone: int
+    ) -> PullRequest | None:
+        """The open PR for a milestone's branch, if there is one. A closed one
+        does not count: the operator closed it, so a new one may be opened.
+        Raises GiteaError when Gitea cannot say."""
+        pr = self._client.find_pull_request(
+            self._repo, self.branch_for(task_id, milestone)
+        )
+        return pr if pr is not None and pr.state == "open" else None
+
     def comment(self, index: int, body: str) -> None:
         """Reply on the PR. Courtesy only — the pushed commit is the real
         answer — so a failure here must not derail the review loop."""

nisse-dev/tests/test_publish.py

@@ -16,6 +16,7 @@ class FakeGitea:
     def __init__(self, *, fail: bool = False) -> None:
         self.opened: list[dict[str, Any]] = []
         self.existing: list[str] = []
+        self.closed: set[str] = set()
         self._fail = fail
 
     def open_pull_request(
@@ -33,7 +34,8 @@ class FakeGitea:
 
     def find_pull_request(self, repo: str, head: str) -> PullRequest | None:
         if head in self.existing:
-            return PullRequest(number=1, state="open", head=head, url="u")
+            state = "closed" if head in self.closed else "open"
+            return PullRequest(number=1, state=state, head=head, url="u")
         return None
 
 
@@ -212,3 +214,32 @@ def test_a_followup_the_executor_committed_itself_is_pushed(
     # Nothing new since that push: no second push, and it says so.
     assert publisher.push_followup(task_id="t", milestone=1, message="m") is None
     assert len(git.pushed) == 2
+
+
+def test_publishing_again_adopts_the_open_pr_instead_of_opening_another(
+    store: StateStore, tmp_path: Path
+) -> None:
+    # E.g. Gitea opened the PR but the response was lost, and the milestone
+    # was published again: the PR is adopted and updated, never duplicated.
+    git = FakeGit(tmp_path / "work")
+    client = FakeGitea()
+    publisher = build(store, git, client)
+    publisher.publish(task_id="t", milestone=1, project="demo", plan="p")
+    (tmp_path / "work" / "more.txt").write_text("more")
+    result = publisher.publish(task_id="t", milestone=1, project="demo", plan="p")
+    assert len(client.opened) == 1
+    assert result.pull_request is not None and result.pull_request.number == 1
+    assert "already open" in result.note
+    assert git.pushed == ["t/milestone-1", "t/milestone-1"]  # updated it
+    assert store.pending_intents() == []
+
+
+def test_a_closed_pr_is_not_adopted(store: StateStore, tmp_path: Path) -> None:
+    git = FakeGit(tmp_path / "work")
+    client = FakeGitea()
+    client.existing.append("t/milestone-1")
+    client.closed.add("t/milestone-1")
+    publisher = build(store, git, client)
+    assert publisher.published_pull_request("t", 1) is None
+    publisher.publish(task_id="t", milestone=1, project="demo", plan="p")
+    assert len(client.opened) == 1  # the operator closed it; a new one opens

nisse-dev/tests/test_quint_counterexamples.py

@@ -134,6 +134,12 @@ class Publisher:
             PullRequest(number=20, state="open", head=branch, url="u"), "opened PR #20"
         )
 
+    def published_pull_request(self, task_id: str, milestone: int) -> Any:
+        branch = f"{task_id}/milestone-{milestone}"
+        if branch in self._gitea.prs:
+            return PullRequest(number=20, state="open", head=branch, url="u")
+        return None
+
     def push_followup(self, **kwargs: Any) -> str | None:
         return "abc12345"
 
@@ -249,7 +255,6 @@ def test_an_afterthought_rejects_the_next_plan(tmp_path: Path) -> None:
     assert agent.gate().get("rev") == 2  # arch was rev-1; not re-planned
 
 
-@XFAIL
 def test_crash_after_opening_the_pr_publishes_again(tmp_path: Path) -> None:
     """crashAfterOpeningThePrPublishesAgainTest"""
     gitea = FakeGitea()
@@ -261,9 +266,11 @@ def test_crash_after_opening_the_pr_publishes_again(tmp_path: Path) -> None:
     assert agent.state == EXECUTING  # the PR exists; PR_OPEN was never recorded
 
     agent = Agent(tmp_path, gitea)  # restart: the intent is verified
-    agent.say("continue", "start milestone 1")
-    # There is a PR for this milestone: the agent must end up watching it,
-    # not trying to open another.
+    # There is a PR for this milestone: the agent adopts it on boot and ends
+    # up watching it, never trying to open another.
+    assert agent.state == PR_OPEN
+    assert "PR #20 is open" in agent.transport.outbound[-1].text
+    agent.say("continue", "start milestone 1")  # no longer means anything
     assert agent.publisher.attempts == 0
     assert agent.state == PR_OPEN
 
@@ -321,3 +328,22 @@ def test_the_harness_reaches_pr_open_without_faults(tmp_path: Path) -> None:
     assert agent.state == PR_OPEN
     assert agent.publisher.attempts == 1
     _ = (AWAIT_INPUT, MILESTONE_PLANNING)
+
+
+def test_boot_asks_as_before_when_gitea_cannot_say(
+    tmp_path: Path, monkeypatch: pytest.MonkeyPatch
+) -> None:
+    gitea = FakeGitea()
+    agent = Agent(tmp_path, gitea)
+    agent.say("use demo", "approve", "approve")
+    agent.publisher.crash_after_opening = True
+    with pytest.raises(SystemExit):
+        agent.say("start milestone 1")
+
+    def unreachable(self: Publisher, task_id: str, milestone: int) -> Any:
+        raise GiteaError("gitea is down")
+
+    monkeypatch.setattr(Publisher, "published_pull_request", unreachable)
+    agent = Agent(tmp_path, gitea)
+    assert agent.state == AWAIT_INPUT
+    assert "restarted mid-EXECUTING" in agent.transport.outbound[-1].text

specs/orchestrator.qnt

@@ -67,6 +67,7 @@ module orchestrator {
     prThread: List[PrComment],
     nextId: int,
     giteaPr: int -> str,      // PR key -> "open" / "closed"
+    opened: int -> int,       // PR key -> how many PRs Gitea ever opened for it
     openIssues: Set[int],     // plan issues still open, by task
     // -- volatile: lost in a crash --
     queue: List[Job],
@@ -83,7 +84,6 @@ module orchestrator {
     answered: Set[int],       // ... and answered (pushed, asked, or reported)
     badApproval: bool,
     staleFeedback: bool,
-    dupPublish: bool,
   }
 
   // ---- helpers ------------------------------------------------------------
@@ -97,6 +97,11 @@ module orchestrator {
   pure def prIds(s: Sys, key: int): int =
     s.prThread.foldl(0, (acc, c) => if (c.key == key and c.id > acc) c.id else acc)
   pure def max(a: int, b: int): int = if (a > b) a else b
+  // Gitea opens a PR for a branch.
+  pure def openPr(s: Sys, key: int): Sys =
+    { ...s, giteaPr: s.giteaPr.put(key, "open"),
+      opened: s.opened.put(key, (if (s.opened.keys().contains(key)) s.opened.get(key) else 0) + 1) }
+
   pure def queued(s: Sys, src: str): bool = s.queue.foldl(false, (acc, j) => acc or j.src == src)
 
   // _await_input
@@ -260,6 +265,11 @@ module orchestrator {
       awaitInput(s0, { kind: "boot", g: "", interrupted: s.st })
     else if (s.st == "AWAIT_INPUT" and s.pending.kind != "")
       s0
+    else if (s.st == "EXECUTING" and giteaUp and prState(s, prKeyOf(s)) == "open")
+      // A3 step 4: the milestone's PR is open, so it was published and only
+      // PR_OPEN went unrecorded. Adopt it; nothing watched it, so from 0.
+      { ...s0, st: "PR_OPEN", prOn: true, prKey: prKeyOf(s), prLast: 0,
+        prWatch: { on: true, floor: 0 }, planWatch: OFF }
     else if (s.st == "PR_OPEN")
       { ...s0, prWatch: if (s.prOn) { on: true, floor: s.prLast } else OFF }
     else if (s.st == "IDLE" or (SIGNOFF.contains(s.st) and s.gate.on))
@@ -278,11 +288,11 @@ module orchestrator {
     boardOn: false, boardPosted: false, boardLast: 0,
     intents: Set(),
     planThread: List(), prThread: List(), nextId: 1,
-    giteaPr: Map(), openIssues: Set(),
+    giteaPr: Map(), opened: Map(), openIssues: Set(),
     queue: List(), work: "none", workGate: "", workCid: 0, pubStage: 0, cancel: false,
     planWatch: OFF, prWatch: OFF,
     shown: NO_PLAN, taken: Set(), answered: Set(),
-    badApproval: false, staleFeedback: false, dupPublish: false,
+    badApproval: false, staleFeedback: false,
   }
 
   // Each action has a deterministic form taking its choices as parameters,
@@ -372,8 +382,9 @@ module orchestrator {
 
   // Publishing, one step at a time, so a crash can land between them
   // (MilestonePublisher.publish and control._publish_milestone):
-  // 0 commit, begin_intent; 1 push, open the PR; 2 commit_intent;
-  // 3 record the PR, watch it, PR_OPEN.
+  // 0 commit, begin_intent; 1 push, then adopt an open PR for the branch or
+  // open one ("lost": Gitea opened it but the response never came back);
+  // 2 commit_intent; 3 record the PR, watch it, PR_OPEN.
   action publishStep(outcome: str): bool = {
     val key = prKeyOf(s)
     all {
@@ -389,10 +400,11 @@ module orchestrator {
           else if (outcome == "fail" or outcome == "nochanges")
             awaitInput(endWork(s), { kind: "milestone_stopped", g: "", interrupted: "" })
           else if (prState(s, key) == "open")
-            // Gitea refuses a second PR for the branch: a GiteaError.
-            awaitInput(endWork({ ...s, dupPublish: true }),
+            { ...s, pubStage: 2 }  // adopted, never a second
+          else if (outcome == "lost")
+            awaitInput(endWork(openPr(s, key)),
                        { kind: "milestone_stopped", g: "", interrupted: "" })
-          else { ...s, giteaPr: s.giteaPr.put(key, "open"), pubStage: 2 }
+          else { ...openPr(s, key), pubStage: 2 }
         else if (s.pubStage == 2)
           { ...s, intents: s.intents.exclude(Set(key)), pubStage: 3 }
         else
@@ -423,7 +435,7 @@ module orchestrator {
     nondet up = Set(true, false).oneOf()
     nondet planOutcome = Set("submit", "ask", "error", "stopped", "exception").oneOf()
     nondet workOutcome = Set("completed", "stopped", "escalate", "exception").oneOf()
-    nondet pubOutcome = Set("ok", "nochanges", "fail", "exception").oneOf()
+    nondet pubOutcome = Set("ok", "nochanges", "fail", "lost", "exception").oneOf()
     any {
       send(kind), commentPlan(ckind), commentPr, pollPlan, pollPr, runJob(up),
       finishPlanning(planOutcome, up), finishExec(workOutcome), publishStep(pubOutcome),
@@ -449,8 +461,8 @@ module orchestrator {
   // Rejection feedback only ever lands on the plan it was written about.
   val feedback_only_on_what_was_seen = not(s.staleFeedback)
 
-  // Never a second attempt to open a PR for a milestone that has one.
-  val one_pr_per_milestone = not(s.dupPublish)
+  // Never more than one PR opened for a milestone.
+  val one_pr_per_milestone = s.opened.keys().forall(k => s.opened.get(k) <= 1)
 
   // A review comment the agent took up is answered, or is being worked on.
   val review_comments_answered =

specs/orchestrator_scenarios.qnt

@@ -1,7 +1,8 @@
 // Counterexamples found in orchestrator.qnt, written out as short traces.
 // Each ends by expecting the property it breaks to be false; once the design
 // is fixed, the expectation is flipped and the trace kept as a regression.
-// Fixed: 1-4 (replies bind to the revision up when they arrived).
+// Fixed: 1-4 (replies bind to the revision up when they arrived), 5 (boot
+// and publishing adopt an open PR).
 //
 // Run:  quint test specs/orchestrator_scenarios.qnt   (runs every *Test)
 module orchestrator_scenarios {
@@ -77,22 +78,32 @@ module orchestrator_scenarios {
       .expect(s.st == "AWAIT_MILESTONE_SIGNOFF" and feedback_only_on_what_was_seen)
 
   // 5. A crash after the PR is opened but before PR_OPEN is recorded.
-  //    Boot finds the intent's PR and marks it done, but the state still
-  //    says EXECUTING, so it asks the restart question. 'continue' returns
-  //    to the approved plan, and starting it publishes again: Gitea refuses
-  //    a second PR for the branch, and there is no path back to PR_OPEN.
+  //    Boot found the intent's PR and marked it done, but the state still
+  //    said EXECUTING, so it asked the restart question; 'continue' and
+  //    'start' then published again, which Gitea refused, with no path back
+  //    to PR_OPEN. Fixed: boot sees the open PR and adopts it.
   run crashAfterOpeningThePrPublishesAgainTest =
     toMilestoneApproved
       .then(send("start")).then(runJob(true))
       .then(finishExec("completed"))
       .then(publishStep("ok"))   // begin_intent
       .then(publishStep("ok"))   // push, open the PR
-      .then(crash(true))         // intent verified and committed; state EXECUTING
-      .then(send("continue")).then(runJob(true))
+      .then(crash(true))         // adopted on boot
+      .expect(s.st == "PR_OPEN" and s.prOn and one_pr_per_milestone)
+
+  // 5b. Gitea opens the PR but the response is lost; the operator starts
+  //     the milestone again. Publishing adopts the open PR.
+  run lostResponseThenRetryAdoptsThePrTest =
+    toMilestoneApproved
+      .then(send("start")).then(runJob(true))
+      .then(finishExec("completed"))
+      .then(publishStep("ok")).then(publishStep("lost"))
+      .then(send("continue")).then(runJob(true))   // back to the approved plan
       .then(send("start")).then(runJob(true))
       .then(finishExec("completed"))
       .then(publishStep("ok")).then(publishStep("ok"))
-      .expect(s.st == "AWAIT_INPUT" and not(one_pr_per_milestone))
+      .then(publishStep("ok")).then(publishStep("ok"))
+      .expect(s.st == "PR_OPEN" and one_pr_per_milestone)
 
   // 6. A crash while the executor answers a review comment. The comment's
   //    cursor was saved before the work started (at most once), and a