Public: anyone can read this; the crew can change it.

A guided review's doors: the terminal walks and sets the steps, a how-to, the MCP read and the brief line #14229

opened this

Check · Review · Land · Live — No review yet. Next: reviewer, no review yet.

Ready to land

scout/forge:scout/14002-a-guided-review-s-doors-the-terminal-wal → main 2 commits fork

Review

Sign in to review this change.
10 files changed +328 −3
AGENTS.md +1
@@ -56,6 +56,7 @@ The check queue hands work to a runner through `src/checks/runner.ts` (#13652, R
5656 The capacity setting `children.awake` (boot default `FORGE_CHILD_AWAKE_MAX`) counts forges in use; a wake or a spawn at the ceiling puts the longest-idle forge to sleep first (`roomFor` in `src/spawn/host.ts`), and the boot never opens a child past it. Hosted children the boot opens (so their apps come back) are exempt from `awakeCount` until a credentialed request reaches them or they sleep (`bootOpened` in `src/spawn/host.ts`). The exempt set holds only body ids of existing finished children and is emptied when the parent closes, so it is bounded by the number of children. Cell children are always counted. The boot starts only the forges that were awake and in use when the parent stopped (`awake-at-stop.json` under the data directory, written by `closeHostedChildren`, read and removed once by `bootHostedChildren`, `src/spawn/awake_at_stop.ts`), up to the ceiling; a crash leaves no file, so nothing starts and forges wake on their next request. A sleeping cell forge with a run due has its wake armed. The idle sweeper starts at boot.
5757
5858 An app token's requests go behind people's and repeat reads are 304s (`src/http/app_turn.ts`, `repeat_reads.ts`, `perf/asking.ts`): see `docs/features/app-asking.md`.
59 `src/core/review_steps.ts`: `reads`, per database, what each pull's review script was last read as (the kept row and the description's block, by pull id); `saveSteps` is the only writer of the row and keeps the entry, a new description body reads again; at most 2000 pulls, dropped whole past that, gone on restart. Its terminal door is `src/tui/screens/pull_review.ts` (`V` walks, `E` sets; `readAnswerLine`/`setWords` in `actions/review_steps.ts` give the page and the terminal the same gates).
5960 `src/core/reports.ts`: `closedSet`, the ids of repositories whose owner closed reports from outside (#12321), read once per database and kept by `setReportsClosed`, the only writer, so every repository page's head link costs no statement; at most one id per closed repository, gone on restart (the rows are `report_closed`). `src/core/contribute_token.ts`: `openSet`, the ids of repositories that allow outside agents, read once per database and kept by `setContributeOpen`, the only writer, at most one id per allowing repository, gone on restart; a raw SQL write to `contribute_open` bypasses it. The terminal door for reports is `r` on the outside agents screen (`setOutsideReports`, admins).
6061 `src/spawn/births.ts`: `states`, per forge, the births in flight, the spawns waiting their turn (first in, first out, signed-in people ahead of strangers; at most 4 × `spawn.births_at_once` wait, strangers at most `spawn.births_at_once` of those, the rest are answered at once with their place; each waits 15 s at most, then is answered with its place) and the mean of the births measured (#13822); a birth is counted until its forge answers or 5 minutes pass; at most `spawn.births_at_once` in flight, gone on restart. The limits that bound a surge (`spawn.*`, `forges.wakes_per_forge_hour`) are settings on /admin/limits.
6162 `src/web/studio/waiting_on_me.ts`: `waitingRows`, each viewer's issues and pulls that wait for them, held 10 seconds per viewer (at most 500 viewers, gone on restart) so the board's live lanes share one `decisionsFor` read; the terminal door is `src/tui/screens/waiting_for_me.ts` (palette entry for any person).
docs/features/README.md +1
@@ -221,6 +221,7 @@ Read with the guide page [Ship and undo](../guide-ship-and-undo.md).
221221 - [The daily count of look pages in the check log](look-count.md)
222222 - [A pull opens on its verdict](pull-verdict.md)
223223 - [A pull's looks](pull-looks.md)
224 - [A guided review: steps, answers, the terminal walk](review-steps.md)
224225 - [An approval says what it was given for](approval-base.md)
225226 - [An approved pull that has waited a week](approved-week.md)
226227 - [Build cache](build-cache.md)
docs/features/review-steps.md added +9
@@ -0,0 +1,9 @@
1 # A guided review
2
3 A builder ends a page-changing pull with a fenced `review` block (1 to 12 steps of `{ go, look, expect, ask? }`), or a writer sets the script; a reviewer walks it in the preview and the answers go back as ONE comment on the pull.
4
5 - **Page and API** — `GET /api/v1/repos/:owner/:name/pulls/:number/review` lists the steps with the preview address and the forge's deep link `/<owner>/<name>/pulls/<n>/review/<step>`; a writer sets them with `PUT` on the same path.
6 - **Terminal** — on an app's pull, `V` walks the steps: `o` opens the step's page in the preview (or shows its address to copy), `w` works, `b` broken (a one-line note follows), `s` skip. After the last answer one comment is sent, the same as the page's. `E` lets a writer set the script from one line of JSON; a bad list says which step is wrong and sets nothing.
7 - **MCP** — the `pull` tool carries `reviewSteps`, the same list as the API.
8
9 Both doors use `readAnswerLine`, `setReviewSteps` and `setWords` in `src/actions/review_steps.ts`, so they give the same answers and refusals. The script's shape is `src/core/review_steps.ts`.
src/actions/review_steps.ts +22
@@ -15,6 +15,7 @@ import {
1515 saveSteps,
1616 stepPath,
1717 stepsOf,
18 stepsWords,
1819 type ReviewStep,
1920 type StepsError,
2021 } from "../core/review_steps.ts";
@@ -73,6 +74,27 @@ export function readAnswers(count: number, get: (key: string) => string | null):
7374 return out;
7475 }
7576
77 /**
78 * One answer typed in the terminal: `w` (works), `b <note>` (broken, the note optional), `s`
79 * (skip, a note optional). Anything else is refused, so the terminal and the page share a gate.
80 */
81 export function readAnswerLine(line: string): StepAnswer | undefined {
82 const m = /^\s*([wbs])(?:\s+([\s\S]*))?$/i.exec(line);
83 if (m === null) return undefined;
84 const v: Verdict = { w: "works", b: "broken", s: "skipped" }[m[1]!.toLowerCase() as "w" | "b" | "s"] as Verdict;
85 const read = readAnswers(1, (k) => (k === "v1" ? v : k === "n1" ? (m[2] ?? "") : null));
86 return read?.[0];
87 }
88
89 /** The words for a refused script write, for the page's message and the terminal's toast. */
90 export function setWords(error: StepsSetError): string {
91 if (error === "unauthorized") return "sign in to set the review steps";
92 if (error === "forbidden") return "only a writer sets the review steps";
93 if (error === "not_found") return "no such pull";
94 if (error === "not_open") return "the pull is not open";
95 return stepsWords(error);
96 }
97
7698 const WORD: Readonly<Record<Verdict, string>> = {
7799 works: "works",
78100 broken: "does not work",
src/crew/brief.ts +1
@@ -224,6 +224,7 @@ export function systemPrompt(input: BriefInput): string {
224224 "- The bar is the forge's own: small modules named for one concept, every input validated at the boundary and every output escaped, the security floor in " +
225225 (isForge(repo) ? "`src/http/respond.ts` (`securityFloor`)" : "`src/server.ts`") +
226226 " kept on every response, no dependency without a reason in the pull, and a page that answers in milliseconds. `AGENTS.md` in the repository says how this app is shaped; keep its shape.",
227 "- A pull that changes what a page shows ends its description with a fenced `review` block: a JSON list of 1 to 12 steps `{ \"go\": \"/path\", \"look\": \"css selector\", \"expect\": \"one sentence a reviewer can check\", \"ask\": \"optional question\" }`; the reviewer walks them in the preview.",
227228 ...(input.previewUrl === undefined
228229 ? []
229230 : [
src/http/mcp/tools/pull.ts +3 −1
@@ -12,6 +12,7 @@ import { appView } from "../../../actions/apps.ts";
1212 import { countThread, listComments } from "../../../core/issues.ts";
1313 import { landingLine } from "../../../flow/landing_line.ts";
1414 import { pullLooks } from "../../../looks/readable.ts";
15 import { reviewSteps } from "../../../actions/review_steps.ts";
1516 import { isCrew } from "../../../core/agents.ts";
1617 import { markReaders } from "../../../crew/direct.ts";
1718 import { good, bad, type Tool, REPO_ARG, repoArg, isAnswer } from "../shared.ts";
@@ -23,7 +24,7 @@ const PULL_COMMENTS = 10;
2324
2425 export const tool: Tool = {
2526 name: "pull",
26 doc: "`{ repo, number }`: a pull's title, state, head and base, `summary` (two or three sentences: what changed, what was verified), and for an open pull whether it merges into the base's current head (`mergeable`) and the paths that stop it (`conflicts`), and `next: { who, why }` (open only; who acts next: builder, reviewer, queue or person), `step: { step, why, done }` (check, review, land or live; null once closed unlanded), and `comments` (the newest 10: author, body, at, `untrusted` on an outsider's), `commentsTotal` and `commentsNext` (older: `GET …/issues/:number/comments?cursor=`), and `looks` (newest first: `{ id, width, says, by, commit, url, at }`, the same as `GET …/pulls/:number/looks`), and `eta: { minutes, says }` (approved open pull only, else null), and `check: { state, baseOid }`: the latest check's state and the base tip it was merged with (empty when it merged nothing), `checkStale` (the base has moved since that check) and `checkedAgainst` (that base tip, or null; the same as `GET …/pulls/:number`); `mergeable` and `conflicts` are the same fields as `GET …/pulls/:number`, `check` is `state` and `run.baseOid` from `GET …/pulls/:number/checks`; draws the `ui://forge/pull` card: `#N title`, state with check and conflicts while open, `head → base` linked to the pull, and the preview link when the pull has an environment",
27 doc: "`{ repo, number }`: a pull's title, state, head and base, `summary` (two or three sentences: what changed, what was verified), and for an open pull whether it merges into the base's current head (`mergeable`) and the paths that stop it (`conflicts`), and `next: { who, why }` (open only; who acts next: builder, reviewer, queue or person), `step: { step, why, done }` (check, review, land or live; null once closed unlanded), and `comments` (the newest 10: author, body, at, `untrusted` on an outsider's), `commentsTotal` and `commentsNext` (older: `GET …/issues/:number/comments?cursor=`), and `looks` (newest first: `{ id, width, says, by, commit, url, at }`, the same as `GET …/pulls/:number/looks`), and `eta: { minutes, says }` (approved open pull only, else null), and `check: { state, baseOid }`: the latest check's state and the base tip it was merged with (empty when it merged nothing), `checkStale` (the base has moved since that check) and `checkedAgainst` (that base tip, or null; the same as `GET …/pulls/:number`); `mergeable` and `conflicts` are the same fields as `GET …/pulls/:number`, `check` is `state` and `run.baseOid` from `GET …/pulls/:number/checks`; `reviewSteps` (the guided review's steps, as `GET …/pulls/:number/review`), draws the `ui://forge/pull` card: `#N title`, state with check and conflicts while open, `head → base` linked to the pull, and the preview link when the pull has an environment",
2728 description:
2829 "A pull: title, state, head, base; open: `mergeable`, `conflicts`, `checkStale`, `checkedAgainst`, `check`.",
2930 inputSchema: {
@@ -69,6 +70,7 @@ export const tool: Tool = {
6970 commentsTotal: total,
7071 commentsNext: comments.length > 0 && comments.length < total ? comments[0]!.id : null,
7172 looks: pullLooks(ctx.app, repo, pull, user),
73 reviewSteps: reviewSteps(ctx.app, user, repo, pull),
7274 number: pull.number,
7375 title: pull.title,
7476 state: pull.mergedAt !== null ? "merged" : pull.state,
src/tui/screens/pull.ts +11 −1
@@ -54,6 +54,7 @@ import { openEnvUrl } from "../../actions/apps.ts";
5454 import { listRepoRuns, mayReadRunLog } from "../../actions/runs.ts";
5555 import { pullLooks, type PullLook } from "../../looks/readable.ts";
5656 import { pullLooksScreen } from "./pulllooks.ts";
57 import { pullReviewScreen, setStepsPrompt } from "./pull_review.ts";
5758 import { combinedView, combinedWords, type CombinedView } from "../../core/combined_view.ts";
5859 import { previewEnv } from "../../apps/data.ts";
5960 import { findingsWords, pullPreviewFindings } from "../../core/preview_findings.ts";
@@ -1169,7 +1170,8 @@ export function pullScreen(
11691170 ? [binding(["O"], "open the preview")]
11701171 : []),
11711172 ...(isApp ? [binding(["P"], "open production")] : []),
1172 ...(isApp ? [binding(["I"], "looks")] : []),
1173 ...(isApp ? [binding(["I"], "looks"), binding(["V"], "walk the review steps")] : []),
1174 ...(isApp && mayWrite() ? [binding(["E"], "set the review steps")] : []),
11731175 ...(takeBackKey ? [binding(["T"], "take it back")] : []),
11741176 ...(isApp && preview !== undefined && !takeBackKey
11751177 ? [binding(["T"], "the preview's terminal")]
@@ -1462,6 +1464,14 @@ export function pullScreen(
14621464 if (pull !== undefined) ctx.push(pullLooksScreen(f, ctx, repo, pull));
14631465 return true;
14641466 }
1467 if ((e.key === "V" || e.key === "E") && isApp && repo !== undefined && detail !== undefined) {
1468 const pull = getPull(f.db, repo.id, number);
1469 if (pull === undefined) return true;
1470 if (e.key === "V") ctx.push(pullReviewScreen(f, ctx, repo, pull, () => runAsync(ctx, reload)));
1471 else if (mayWrite() && pull.state === "open")
1472 setStepsPrompt(f, ctx, screen, repo, pull, () => undefined);
1473 return true;
1474 }
14651475 if (e.key === "P" && isApp) {
14661476 openProduction();
14671477 return true;
src/tui/screens/pull_review.ts added +133
@@ -0,0 +1,133 @@
1 /**
2 * `V` on an app's pull: walk its review steps one by one. `o` opens the step's page in the
3 * preview (or shows its address to copy), `w` works, `b` broken with a note, `s` skip; after the
4 * last answer ONE comment goes to the pull, the same as the page's. `setStepsPrompt` (`E`) lets a
5 * writer set the script from a JSON list. Gates are `actions/review_steps.ts`'s.
6 */
7
8 import {
9 readAnswerLine,
10 reviewSteps,
11 sendReviewAnswers,
12 setReviewSteps,
13 setWords,
14 type StepAnswer,
15 } from "../../actions/review_steps.ts";
16 import type { Pull } from "../../core/pulls.ts";
17 import type { Repo } from "../../core/repos.ts";
18 import type { Ctx, Screen } from "../app/app.ts";
19 import { binding } from "../app/keys.ts";
20 import type { Session } from "../session.ts";
21 import { style } from "../term/style.ts";
22 import { emptyState } from "../ui/widgets.ts";
23 import { openLink, promptScreen } from "./common.ts";
24
25 const plain = (text: string): string => text.replace(/[\p{Cc}\p{Cf}\u2028\u2029]/gu, " ");
26
27 export function pullReviewScreen(f: Session, ctx: Ctx, repo: Repo, pull: Pull, onSent: () => void): Screen {
28 const steps = reviewSteps(f.app, f.user, repo, pull);
29 const answers: StepAnswer[] = [];
30 let sending = false;
31 const screen: Screen = {
32 title: `Review · #${pull.number}`,
33 get keys() {
34 return steps.length === 0
35 ? [binding(["esc", "q"], "back")]
36 : [
37 binding(["o"], "open the step"),
38 binding(["w"], "works"),
39 binding(["b"], "broken"),
40 binding(["s"], "skip"),
41 binding(["esc", "q"], "back"),
42 ];
43 },
44 render(area, s) {
45 const c = ctx.theme.c;
46 if (steps.length === 0) {
47 emptyState(s, area, c, {
48 title: "No review steps",
49 body: "A writer sets them with E, or the pull's description carries a review block.",
50 });
51 return;
52 }
53 const step = steps[Math.min(answers.length, steps.length - 1)]!;
54 const rows = [
55 `Step ${step.step} of ${steps.length}`,
56 "",
57 `Look at: ${plain(step.look)}`,
58 `Expect: ${plain(step.expect)}`,
59 ...(step.ask === null ? [] : [`Ask: ${plain(step.ask)}`]),
60 `Page: ${plain(step.url ?? step.link)}`,
61 ];
62 rows.forEach((r, i) => {
63 if (i < area.h) s.text(area.x, area.y + i, r.slice(0, area.w), style(c.fg, c.bg), area.x + area.w);
64 });
65 },
66 key(e) {
67 if (e.key === "esc" || e.key === "q") {
68 ctx.pop();
69 return true;
70 }
71 const step = steps[answers.length];
72 if (step === undefined || sending) return true;
73 if (e.key === "o") {
74 if (step.url === null) ctx.toast("this step's page is not open to you", "attn");
75 else openLink(ctx, step.url);
76 return true;
77 }
78 const take = (a: StepAnswer | undefined): void => {
79 if (a === undefined) return;
80 answers.push(a);
81 ctx.invalidate();
82 if (answers.length < steps.length) return;
83 sending = true;
84 void sendReviewAnswers(f.app, f.user, repo, pull, answers).then((out) => {
85 ctx.toast(out.ok ? "sent your answers as one comment" : "could not send the answers", out.ok ? "success" : "danger");
86 if (out.ok) {
87 onSent();
88 ctx.pop();
89 } else sending = false;
90 });
91 };
92 if (e.key === "w" || e.key === "s") take(readAnswerLine(e.key));
93 else if (e.key === "b")
94 ctx.push(
95 promptScreen(ctx, screen, {
96 title: `Step ${step.step} does not work`,
97 label: "Note",
98 hint: "What went wrong, one line (optional). Enter records it.",
99 onSubmit: (value) => {
100 take(readAnswerLine(`b ${value}`));
101 return undefined;
102 },
103 }),
104 );
105 return true;
106 },
107 };
108 return screen;
109 }
110
111 /** `E`: a writer sets the script from a JSON list of `{ go, look, expect, ask? }`. */
112 export function setStepsPrompt(f: Session, ctx: Ctx, parent: Screen, repo: Repo, pull: Pull, done: () => void): void {
113 ctx.push(
114 promptScreen(ctx, parent, {
115 title: `Review steps · #${pull.number}`,
116 label: "JSON",
117 hint: 'One line: [{"go":"/","look":"main","expect":"…"}]. Enter sets; Esc cancels.',
118 onSubmit: (value) => {
119 let raw: unknown;
120 try {
121 raw = JSON.parse(value);
122 } catch {
123 return "that is not JSON: a list of { go, look, expect, ask? }";
124 }
125 const out = setReviewSteps(f.app, f.user, repo, pull, raw);
126 if (!out.ok) return setWords(out.error);
127 ctx.toast(`${out.value.length} review step${out.value.length === 1 ? "" : "s"} set`, "success");
128 done();
129 return undefined;
130 },
131 }),
132 );
133 }
test/review-steps.test.ts +17 −1
@@ -10,7 +10,7 @@ import { tmpdir } from "node:os";
1010 import { join } from "node:path";
1111 import { openPull } from "../src/actions/pulls.ts";
1212 import { commitToBranch, createRepo } from "../src/actions/repos.ts";
13 import { answersComment, readAnswers } from "../src/actions/review_steps.ts";
13 import { answersComment, readAnswerLine, readAnswers, setWords } from "../src/actions/review_steps.ts";
1414 import { createApp, type App } from "../src/app.ts";
1515 import { AppsRuntime, type EnvStatus } from "../src/apps/runtime.ts";
1616 import { tourJs, tourResponse, TOUR_ASSET_PATH } from "../src/apps/tour.ts";
@@ -79,6 +79,22 @@ describe("the answers", () => {
7979 });
8080 });
8181
82 describe("the terminal's answer line", () => {
83 test("w, b <note> and s are the only answers", () => {
84 expect(readAnswerLine("w")).toEqual({ verdict: "works", note: "" });
85 expect(readAnswerLine(" b the list\u0007 is empty ")).toEqual({ verdict: "broken", note: "the list is empty" });
86 expect(readAnswerLine("s")).toEqual({ verdict: "skipped", note: "" });
87 expect(readAnswerLine("great")).toBeUndefined();
88 expect(readAnswerLine("")).toBeUndefined();
89 expect(readAnswerLine("b" + "x".repeat(600))).toBeUndefined();
90 expect(readAnswerLine("b " + "x".repeat(600))!.note.length).toBe(500);
91 });
92 test("a refused write has words", () => {
93 expect(setWords("forbidden")).toContain("writer");
94 expect(setWords("step_2")).toContain("step 2");
95 });
96 });
97
8298 describe("the tour's tag", () => {
8399 const tour = { repo: "a/b", pull: 3, steps: [{ go: "/", look: "h1", expect: 'say "hi" <b>' }] };
84100 test("is written only with a script, escaped, and points at a content-addressed file", () => {
test/tui-pull-review.test.ts added +130
@@ -0,0 +1,130 @@
1 /**
2 * A guided review from the terminal (#14002): `V` walks the steps and sends ONE comment; `E` sets
3 * them.
4 */
5 import { afterAll, beforeAll, expect, test } from "bun:test";
6 import { mkdtemp, rm } from "node:fs/promises";
7 import { tmpdir } from "node:os";
8 import { join } from "node:path";
9 import { openPull } from "../src/actions/pulls.ts";
10 import { commitToBranch, createRepo } from "../src/actions/repos.ts";
11 import { createApp, type App as Forge } from "../src/app.ts";
12 import { createApp as insertApp } from "../src/core/apps.ts";
13 import { listComments } from "../src/core/issues.ts";
14 import { getPull, type Pull } from "../src/core/pulls.ts";
15 import type { Repo } from "../src/core/repos.ts";
16 import { stepsOf } from "../src/core/review_steps.ts";
17 import { createUser, type User } from "../src/core/users.ts";
18 import { createForgeApp } from "../src/tui/forge.ts";
19 import { pullScreen } from "../src/tui/screens/pull.ts";
20 import { local } from "../src/tui/screens/where.ts";
21 import { headlessTerminal } from "../src/tui/term/terminal.ts";
22 import { seedAdmin } from "./helpers/admin.ts";
23 import { quietLogs } from "./helpers/globals.ts";
24 import { settleApp } from "./helpers/tui.ts";
25
26 quietLogs();
27
28 let dataDir: string;
29 let forge: Forge;
30 let alice: User;
31 let bob: User;
32 let repo: Repo;
33 let pull: Pull;
34
35 beforeAll(async () => {
36 dataDir = await mkdtemp(join(tmpdir(), "forge-tui-review-"));
37 forge = createApp({ dataDir, port: 0, url: "https://forge.test" });
38 const a = await seedAdmin(forge, { username: "alice", email: "[email protected]", password: "a-long-passphrase" });
39 if (!a.ok) throw new Error("signup failed");
40 alice = a.value;
41 const b = await createUser(forge.db, {
42 origin: "seed",
43 trusted: true,
44 username: "bob",
45 email: "[email protected]",
46 password: "a-long-passphrase",
47 });
48 if (!b.ok) throw new Error(b.error);
49 bob = b.value;
50 const r = await createRepo(forge, alice, { name: "tour", visibility: "public" });
51 if (!r.ok) throw new Error("repo creation failed");
52 repo = r.value;
53 await commitToBranch(forge, alice, repo, { message: "seed", files: [{ path: "README.md", content: "# s\n" }] });
54 await commitToBranch(forge, alice, repo, {
55 branch: "feature/one",
56 base: "main",
57 message: "One",
58 files: [{ path: "one.ts", content: "export const one = 1;\n" }],
59 });
60 const o = await openPull(forge, alice, repo, { title: "Pull one", body: "", baseBranch: "main", headBranch: "feature/one" });
61 if (!o.ok) throw new Error(String(o.error));
62 pull = o.value;
63 insertApp(forge.db, { repoId: repo.id, slug: "tour", access: "public" });
64 });
65
66 afterAll(async () => {
67 forge.close();
68 await rm(dataDir, { recursive: true, force: true });
69 });
70
71 async function open(user: User) {
72 const { app, session } = createForgeApp(
73 forge,
74 { sshUsername: user.username, remoteAddress: "127.0.0.1", user },
75 { terminal: headlessTerminal(160, 50), exitHooks: false },
76 );
77 app.start();
78 app.ctx.push(pullScreen(session, app.ctx, local(session, repo), pull.number));
79 await settleApp(app);
80 const press = async (...keys: string[]) => {
81 for (const key of keys) app.dispatch({ type: "key", key, ctrl: false, alt: false, shift: false });
82 await settleApp(app);
83 };
84 return { app, press, text: () => app.screen.toText() };
85 }
86
87 const SCRIPT = JSON.stringify([
88 { go: "/", look: "main h1", expect: "The heading says Hello" },
89 { go: "/about", look: "#team", expect: "Three people", ask: "Readable?" },
90 ]);
91
92 test("a writer sets the steps with E; a bad list says why and sets nothing", async () => {
93 const h = await open(alice);
94 await h.press("E", ..."not json", "enter");
95 expect(h.text()).toContain("not JSON");
96 await h.press("esc");
97 await h.press("E", ...'[{"go":"https://x.example/","look":"a","expect":"b"}]', "enter");
98 expect(h.text()).toContain("step 1 needs go");
99 expect(stepsOf(forge.db, getPull(forge.db, repo.id, pull.number)!)).toEqual([]);
100 await h.press("esc");
101 await h.press("E", ...SCRIPT, "enter");
102 expect(stepsOf(forge.db, getPull(forge.db, repo.id, pull.number)!)).toHaveLength(2);
103 h.app.stop();
104 });
105
106 test("a reader without write cannot set steps", async () => {
107 const h = await open(bob);
108 await h.press("E");
109 expect(h.text()).not.toContain("Review steps");
110 h.app.stop();
111 });
112
113 test("V walks the steps and sends one comment", async () => {
114 const before = listComments(forge.db, pull.id, {}).length;
115 const h = await open(alice);
116 await h.press("V");
117 expect(h.text()).toContain("Step 1 of 2");
118 expect(h.text()).toContain("The heading says Hello");
119 await h.press("w");
120 expect(h.text()).toContain("Step 2 of 2");
121 await h.press("b", ..."list is empty", "enter");
122 await new Promise((r) => setTimeout(r, 50));
123 await h.press();
124 const comments = listComments(forge.db, pull.id, {});
125 expect(comments.length).toBe(before + 1);
126 const body = comments[comments.length - 1]!.body;
127 expect(body).toContain("1 of 2 steps work, 1 do not, 0 not tried.");
128 expect(body).toContain("Note: list is empty");
129 h.app.stop();
130 });

↑↓ navigate ↵ open ⌘↵ its app esc close

Keyboard shortcuts

Docs