nisse/tests/integration/test_end_to_end_loop.py

"""The end-to-end test on the new loop (Development Plan M2.1, PR D): the
whole slice, from an issue assigned in Gitea to a merged PR and a closed work
item, with real models, the Gitea sandbox and a real build host.

A scripted operator plays the operator's part: in Gitea through its own test
account (NISSE_OPERATOR_TOKEN; an operator only in nisse-test.yaml), and in
the item's Matrix thread through the in-memory transport. Run with
scripts/e2e.sh after each milestone.
"""

import contextlib
import os
import time
import uuid
from collections.abc import Iterator
from dataclasses import replace
from fnmatch import fnmatch
from pathlib import Path
from typing import Any

import pytest
from e2e_support import base_branch, operator_gitea, task

from nisse_dev.config import AgentConfig
from nisse_dev.execution import SSHHost
from nisse_dev.gitea import GiteaClient, GiteaError
from nisse_dev.loop.build import RunningLoop, build_loop_agent
from nisse_dev.loop.items import ANSWER, GO_AHEAD, MERGE, WAITING, loop_of
from nisse_dev.transport import InMemoryTransport

pytestmark = [pytest.mark.integration, pytest.mark.e2e]

OPERATOR = "@dongar:dongar.net"  # an allowlisted Matrix operator in nisse-test.yaml
DEADLINE_S = 1200
MAX_ANSWERS = 5  # per wait for something else, before giving up


@pytest.fixture
def running(
    live_config: AgentConfig,
    ssh_host: SSHHost,
    tmp_path: Path,
    monkeypatch: pytest.MonkeyPatch,
) -> Iterator[tuple[RunningLoop, InMemoryTransport, str]]:
    from test_end_to_end import build_host

    run = uuid.uuid4().hex[:8]
    stack = contextlib.ExitStack()
    # Kept outside pytest's temp directories, which pytest prunes, so a
    # failed run's transcript can be read afterwards (nisse_dev.dump).
    keep = Path(os.environ.get("NISSE_E2E_RUNS", "/tmp/nisse-e2e-runs")) / run
    keep.mkdir(parents=True, exist_ok=True)
    print(f"\ne2e run {run}: state store in {keep}")
    operator = operator_gitea(live_config.gitea_url)
    base = stack.enter_context(base_branch(operator, live_config.gitea_repo, run))
    config = replace(
        live_config,
        loop="new",
        base_branch=base,
        state_db=keep / "state.db",
        workspace=Path(f"/tmp/nisse-e2e-{run}/sandbox"),
        build_host=build_host(ssh_host),
    )
    transport = InMemoryTransport()
    monkeypatch.setattr("nisse_dev.loop.build.MatrixTransport", lambda cfg: transport)
    agent = None
    try:
        agent = build_loop_agent(config)
        # The test drives intake and PR watching itself, not on a timer.
        agent.loop._interval = 3600
        agent.start()
        yield agent, transport, run
    finally:
        if agent is not None:
            agent.stop()
        host = SSHHost(
            config.build_host.address,
            key=config.build_host.key,
            user=config.build_host.user,
            port=config.build_host.port,
            state_dir=tmp_path / "ssh",
        )
        host.execute(f"rm -rf /tmp/nisse-e2e-{run}")
        stack.close()  # the run's base branch


def is_artifact(path: str) -> bool:
    parts = path.split("/")
    return (
        "__pycache__" in parts
        or ".pytest_cache" in parts
        or fnmatch(parts[-1], "*.py[co]")
    )


def merge(gitea: GiteaClient, repo: str, pr: int, wait_s: float = 60) -> None:
    """Merge, waiting out Gitea's "Please try again later": it answers that
    while it is still checking a freshly pushed PR can merge."""
    deadline = time.monotonic() + wait_s
    while True:
        try:
            gitea._request(
                "POST", f"/repos/{repo}/pulls/{pr}/merge", json={"Do": "merge"}
            )
            return
        except GiteaError as exc:
            if "try again later" not in str(exc) or time.monotonic() > deadline:
                raise
            time.sleep(3)


class Operator:
    def __init__(self, agent: RunningLoop, transport: InMemoryTransport) -> None:
        self.agent = agent
        self.transport = transport

    def settle(self) -> None:
        self.agent.worker.call(lambda: None, timeout=DEADLINE_S)

    def tick(self) -> None:
        """Intake and PR watching, as the agent's timer would do."""
        self.agent.worker.call(self.agent.loop.intake, timeout=DEADLINE_S)

    def item(self) -> Any:
        items = self.agent.worker.call(self.agent.store.work_items)
        return items[0] if items else None

    def say(self, text: str) -> None:
        self.transport.receive(OPERATOR, text, thread=loop_of(self.item()).thread)
        self.settle()

    def drive_until(self, wait: str) -> Any:
        """Answer whatever the agent asks until it waits for ``wait``."""
        deadline = time.monotonic() + DEADLINE_S
        answers = 0
        while time.monotonic() < deadline:
            item = self.item()
            assert item is not None, "no work item"
            loop = loop_of(item)
            if item.position == WAITING and loop.wait == wait:
                return item
            if item.status == "closed":
                raise AssertionError(f"closed early: {self.last()}")
            if item.position == WAITING and loop.wait == GO_AHEAD:
                self.say("go")
            elif item.position == WAITING and loop.wait == ANSWER:
                answers += 1
                # Answering forever is how a stuck agent looks: give up.
                assert answers <= MAX_ANSWERS, f"stuck asking: {self.last()}"
                self.say("Use your best judgement and proceed.")
            else:
                self.settle()
        raise AssertionError(f"never waited for {wait}: {self.last()}")

    def last(self) -> str:
        return self.transport.outbound[-1].text if self.transport.outbound else ""


def test_the_slice_end_to_end(
    running: tuple[RunningLoop, InMemoryTransport, str],
) -> None:
    agent, transport, run = running
    operator = Operator(agent, transport)
    gitea, repo = agent.gitea, agent.config.gitea_repo
    as_operator = operator_gitea(agent.config.gitea_url)
    issue = None
    pr = 0
    try:
        # --- the operator files an issue and assigns it to the agent (D1) ---
        issue = as_operator.create_issue(repo, f"vowels_{run}", task(run))
        as_operator._request(
            "PATCH",
            f"/repos/{repo}/issues/{issue}",
            json={"assignees": [agent.account]},
        )
        operator.tick()
        item = operator.item()
        assert item is not None and loop_of(item).issue == issue

        # --- restatement, go-ahead, milestone, PR (L3, D4) ---
        item = operator.drive_until(MERGE)
        pr = loop_of(item).pr
        files = gitea._request("GET", f"/repos/{repo}/pulls/{pr}/files")
        changed = [f["filename"] for f in files]
        assert changed and not [p for p in changed if is_artifact(p)], changed
        body = gitea._request("GET", f"/repos/{repo}/pulls/{pr}")["body"]
        assert f"Closes #{issue}" in body

        # --- a review comment is answered with a fix (D5) ---
        # A request that can't already be met, so a fix must follow.
        marker = f"reviewed-{uuid.uuid4().hex[:8]}"
        before = gitea._request("GET", f"/repos/{repo}/pulls/{pr}")["head"]["sha"]
        as_operator.post_comment(
            repo, pr, f"Please add the comment line `# {marker}` above vowels_{run}."
        )
        operator.tick()
        operator.drive_until(MERGE)
        after = gitea._request("GET", f"/repos/{repo}/pulls/{pr}")["head"]["sha"]
        assert after != before, "no fix was pushed for the review comment"
        diff = gitea._http.get(f"/repos/{repo}/pulls/{pr}.diff").text
        assert marker in diff, "the pushed fix doesn't do what the comment asked"

        # --- the operator approves and merges; the agent closes the item ---
        with contextlib.suppress(GiteaError):  # approval may not be required
            as_operator._request(
                "POST",
                f"/repos/{repo}/pulls/{pr}/reviews",
                json={"event": "APPROVED", "body": "looks good"},
            )
        merge(as_operator, repo, pr)
        operator.tick()
        deadline = time.monotonic() + DEADLINE_S
        nudges = 0
        while operator.item().status != "closed" and time.monotonic() < deadline:
            if loop_of(operator.item()).wait == ANSWER:
                nudges += 1
                assert nudges <= MAX_ANSWERS, f"never closed: {operator.last()}"
                operator.say("It's merged; please close the item.")
            operator.settle()
        assert operator.item().outcome == "completed"
        state = gitea._request("GET", f"/repos/{repo}/issues/{issue}")["state"]
        assert state == "closed"
    finally:
        # Leave the sandbox tidy: whatever is still open gets closed.
        item = operator.item()
        if item is not None and item.status == "open":
            operator.say("abandon")
        if issue is not None:
            with contextlib.suppress(GiteaError):
                as_operator.close_issue(repo, issue)
        if pr:
            with contextlib.suppress(GiteaError):
                head = gitea._request("GET", f"/repos/{repo}/pulls/{pr}")["head"]["ref"]
                as_operator._request("DELETE", f"/repos/{repo}/branches/{head}")