Console: Add device… from the catalog keeps one more warm
メンテナーはふだん 1 日以内に返信
まだ誰も着手していません。
評価
- 難易度
- 4/5
- 見積もり時間
- 3〜5日
- 初心者へのやさしさ
- 48/100
- issue の種類
- 機能追加
- 明瞭さ
- 明確に書かれている
- 活発さ
- 活発
- 技術スタック
- typescript
調査の方向性
Start with ui/src/model/actions.ts and ui/src/state/actions.ts, then read the ActionConfirm flow in ui/src/ds/confirm-dialog.tsx and the listed Actions guidance in ui/AGENTS.md. Run the model and state tests named in the issue, then the browser lane in e2e/console/actions.spec.ts. Done means the specified catalog-driven dialog and warm-pool update work on all three surfaces, with the described errors and tests passing.
索引モデルが issue の本文から書いたものです。
説明
Part of #1137.
Scope
After this PR an operator on a worker's console adds a simulator or emulator of a model, OS and mode picked from the worker's catalog. Add device… sits beside #1133's Enrol phone… on every surface that offers it (the Fleet table's worker row, the worker page, ⌘K). It raises that kind's runtime count by one, and the warm pool creates and boots the device on its next pass. Nothing new on the daemon: it is #1143's warm.set through #1144's route.
In short
Add device… reads GET /v1/catalog, then opens a dialog built from it. The operator picks a Platform, a Model from that platform's models, an OS from the runtimes the catalog pairs with the model, and on iOS a Mode. Add then reads GET /v1/warm-pool and sends one PUT /v1/warm-pool/targets. That PUT raises the matching row by one, or adds a row at the end counted from a matching learned kind's peak, else from 0 (#1137's "Count a client raises from"), and keeps the rest of the runtime list and runtime.learn. The matching row is, in order: the effective row whose key is the one Add would write (iOS: platform, model, OS, mode; Android: platform, model, OS); else the first effective row whose platform, model and resolved equal the picked kind; else none. Android's mode counts as full when matching resolved and learned kinds. The effective list holds one row per key (#1137), so the added row never repeats one. The toast says how many of the kind the pool now keeps.
flowchart LR
A["Add device… (new), beside Enrol phone…, worker mode"] --> B["GET /v1/catalog"]
B --> C[Dialog: Platform, Model, OS, Mode]
C -->|Add| D[GET /v1/warm-pool]
D --> E[PUT /v1/warm-pool/targets, one row raised]
E --> F[Toast]
style A stroke-width:3px
style B stroke-width:3px
style C stroke-width:3px
style D stroke-width:3px
style E stroke-width:3px
Words, exactly (<kind> is the Kind spelling in #1137's Shared values for the picked platform, model, OS and, on iOS, mode; <n> the count sent; <model> the picked model; <message> the daemon's or the client's):
| Where | Text |
|---|---|
| action label | Add device… |
| dialog title | Add a device |
| its body | The warm pool keeps one more of this kind booted and ready, and boots it on its next pass. |
| its fields | Platform (the catalog's platforms that list a model, iOS first and chosen when listed), Model (the platform's models, in catalog order, none chosen), OS (the model's modelRuntimes, in catalog order, the first chosen), Mode (iOS only, Full chosen, Slim; Android shows no Mode field) |
| while the catalog is read | Add device… is busy, as any action in flight; no dialog yet |
| a catalog read that fails | no dialog; error toast Couldn't read the catalog, with the message below and Retry |
| a catalog with no platform that lists a model (a driver whose catalog read failed is left out of a 200 answer) | no dialog; error toast This worker's catalog lists no devices |
| a model with no paired runtime | No runtime is installed for <model>., Add disabled |
| the fresh read at submit finds the matching row's base at 256 or more | no PUT; error toast Couldn't add <kind>, description This kind is already kept at 256. |
| buttons | Cancel, Add (busy: Adding…) |
| success toast | Keeping <n> <kind> warm, description The pool boots one on its next pass. |
| error toast title | Couldn't add <kind>, with the message below and Retry |
Which row Add raises (learnedCount and resolved from warm.list):
| Picked | Effective list | Learned | Sends |
|---|---|---|---|
| iOS iPhone 17 26.0 Full | runtime {ios, iPhone 17, 26.0, mode full, count 1} |
— | that row, 2 |
| iOS iPhone 17 26.0 Full | configured {ios, iPhone 17, osVersion 26.0, count 2}, resolved 26.0, full |
— | a runtime row of that key (no mode), 3 |
| iOS iPhone 16 18.5 Full | none | ios iPhone 16 18.5 (full), peak 2 |
new row {ios, iPhone 16, 18.5, mode full}, 3 |
| Android Pixel 8 35 | none | — | new row {android, Pixel 8, 35}, 1 |
| Android Pixel 8 35 | none | android Pixel 8 35 (full, google_apis), peak 2 |
new row, 1 (a learned kind with an image tag is not this kind) |
| iOS iPhone 17 26.0 Full, the second time before any pass | runtime {ios, iPhone 17, 26.0, mode full, count 1}, no resolved yet |
— | that row by its key, 2 |
| iOS iPhone 17 26.0 Full | configured {ios, iPhone 17, 26.0} counts 2 and 1 in config.json, one row of count 3 in warm.list |
— | a runtime row of that key, 4 |
If this goes wrong, an operator sees Add device… keep two more instead of one, or offer an OS the catalog does not pair with the model.
Technical spec
Facts
Commit ce8fea85.
| # | Fact | Source |
|---|---|---|
| F1 | GET /v1/catalog (role agent) answers {platforms: [{platform, models, runtimes, modelRuntimes, …}]} |
src/http/route-declarations.ts:44-52, src/contract/operations.ts:81-87, src/contract/schemas.ts:683-713 |
| F2 | iOS modelRuntimes pairs each model with the installed runtimes it fits; Android pairs each model with the API levels of the images that fit its profile, none for a model with no profile |
src/drivers/ios/index.ts:1215-1217, src/drivers/android/index.ts:1446-1451 |
| F3 | The console reads /v1/catalog nowhere today |
search v1/catalog in ui/src (no hit); e2e/http-api.test.ts:98-114 is the daemon's test |
| F4 | A request with no mode gets the platform's default mode, ios.defaultMode (default full), configurable |
src/core/config.ts:365-370, :989 |
| F5 | A target's mode is slim or full; Android makes no slim device |
src/core/warm-pool/config.ts:63-64; search slim in src/drivers/android/ (no hit) |
| F6 | useActions().run sends one write and shows the success or error toast with Retry |
ui/src/state/actions.ts:72-108 |
| F7 | The Pager already uses a native <select> |
ui/src/ds/pager.tsx:29 |
| F8 | DESIGN.md styles a select at 40 px tall, in mono | docs/internal/DESIGN.md:238-243 |
| F9 | The browser lane's "the console acts" drives dialogs and toasts against a real daemon with a fake driver | e2e/console/actions.spec.ts:78, e2e/console/fixtures.ts:52 |
| F10 | Learned kinds carry an image tag that targets cannot name | src/core/warm-pool/learned.ts:6-16, src/core/warm-pool/config.ts:20-28 |
| F11 | With no platform query, the daemon leaves out a platform whose catalog read fails and still answers 200 |
src/core/driver-catalog.ts:127-146, src/daemon/dispatcher.ts:346 |
| F12 | ds/ imports nothing from the rest of the console; a confirm's spec is fixed when it opens |
ui/AGENTS.md (Layers), ui/src/ds/confirm-dialog.tsx:31-56 |
#1133 (not merged at this commit) adds enrolActions, the form fields on ActionConfirm (a choice and a text field) and the worker-mode gate this task builds on; #1145 adds PUT to the console's write, the ['warm-pool'] read and the raise rule.
Modules touched
Searches: enrolActions (#1133), ActionConfirm, confirm-dialog, v1/catalog, <select in ui/src/.
ui/src/model/actions.ts: idadd-device; Add device… joins the listenrolActionsanswers; a pureaddDeviceTargets(list, picked)(new) that answers the runtime list andlearnto send and the count<n>.ui/src/ds/confirm-dialog.tsxand its CSS: a select field whose options come from a pure function of the earlier fields' values, given in the spec, and a line under the fields that keeps the confirm disabled (F12).ui/src/state/actions.ts: Add device… readsGET /v1/catalog(not polled), then opens the confirm with the fields built from it; at submit it readsGET /v1/warm-pooland builds the PUT's body from that answer. Its in-flight key iswarm-pool, shared with #1145's controls.ui/AGENTS.md(Actions),docs/internal/DESIGN.md(the select field in the dialog),website/src/content/docs/docs/worker/console.mdx.
flowchart LR
SV[surfaces: Fleet row, worker page, ⌘K] --> MA[model/actions *]
SV --> AC[state/actions *]
AC --> CAT[catalog read *]
AC --> CD[ds/confirm-dialog *]
AC --> API[data/api]
sequenceDiagram
participant O as Operator
participant C as Console
participant D as Daemon (worker)
O->>C: Add device…
C->>D: GET /v1/catalog
alt error or no answer
C-->>O: error toast Couldn't read the catalog, message, Retry
else 200, no platform lists a model
C-->>O: error toast This worker's catalog lists no devices
else 200
O->>C: pick, Add
C->>D: GET /v1/warm-pool
C->>D: PUT /v1/warm-pool/targets
alt 200
C-->>O: toast Keeping n kind warm
else error (either call)
C-->>O: error toast Couldn't add kind, message, Retry
end
end
Contract and event changes
none: the console sends #1143's warm.set input through #1144's route and reads catalog.get as it is.
Other code on the same state
Searches: writers of the runtime list (#1144's CLI, #1145's tab, this dialog).
| Component | Both act | Wins |
|---|---|---|
#1145's tab or a CLI warm set between this dialog's GET /v1/warm-pool and its PUT |
the dialog sends the list it read at submit with its one change | the later PUT (#1137's Assumption on two writers) |
| A second Add while the first is in flight | Add is disabled while busy | the first |
Failure modes
| Call | Outcome | What happens | Limit, counted from |
|---|---|---|---|
GET /v1/catalog |
200 | the dialog opens with its fields | 10 s, from the click |
GET /v1/catalog |
200 with no platform that lists a model | no dialog; This worker's catalog lists no devices |
10 s |
GET /v1/catalog |
error or no answer | no dialog; Couldn't read the catalog with the message (The daemon did not answer. with no answer) and Retry |
10 s |
GET /v1/warm-pool at submit |
error or no answer | error toast Couldn't add <kind>, nothing sent; Retry reads again |
10 s, from Add |
PUT /v1/warm-pool/targets |
200 | success toast | 10 s, from the read's answer |
PUT |
400, 501, no answer | error toast with the message and Retry | 10 s |
The console writes no state of its own.
Rules in play
- ADR 0037 §3: Add device… is one entry of the list
enrolActionsanswers, read by every surface that offers Enrol phone…. - ADR 0037 §4: offered only in worker mode.
- ADR 0037 §5: the toasts.
- ADR 0056 §6: "Add device…" raises one row's count by one.
- Documentation rule 9:
worker/console.mdxdescribes Add device….
Tests
Seam: enrolActions and addDeviceTargets (ui/src/model/actions.test.ts), useActions().run (ui/src/state/actions.test.tsx), and the browser lane's "the console acts" (e2e/console/actions.spec.ts) for the dialog.
- An operator sees Add device… beside Enrol phone… on the Fleet row, the worker page and ⌘K only in worker mode: mode
worker/gateway/ unknown. - An operator opens Add device… and picks platform, model, OS and mode from the worker's catalog: iOS offers full and slim, Android no Mode field; a model lists only its runtimes; a model with no runtime shows
No runtime is installed for <model>.and Add stays disabled; Add disabled until all are picked; a catalog read that fails or lists no device opens no dialog and shows its toast; Cancel sends nothing. - An operator adds a device and sees one
PUT /v1/warm-pool/targetsraising that kind's count by one, then a toast: a kind with a runtime row; the same kind again before a pass, matched by key; a kind with only a configured target written with no mode, matched byresolved; a kind only learned, peak 2, sends 3; a new iOS kind writes its mode, a new Android kind none; a runtimelearnkept; the warm-pool read at submit fails (error toast, nothing sent, Retry reads again); a matching row already at 256 sends nothing and showsThis kind is already kept at 256.; 400 and 501 show the daemon's message.
Done when
pnpm exec vitest run --project unit ui/src/model/actions.test.ts ui/src/state/actions.test.tsxpasses, each Tests line but the dialog's a test there.pnpm test:console --project=chromium e2e/console/actions.spec.tspasses with the dialog's Tests line and one more test: on a worker, Add device… with a picked kind sends onePUT /v1/warm-pool/targetsand the toast shows.website/src/content/docs/docs/worker/console.mdxdescribes Add device….
Out of scope
- A device of an Android image type.
- Creating a device outside the warm pool, or one that is not kept warm.
- Add device… on a gateway's console.
Depends on
- #1145
- #1133
- #1135
Written by an agent.
- 主要言語
- TypeScript
- スター
- 705
- フォーク
- 40
- 平均マージ
- 12時間 9分
- マージ済み PR(30日)
- 376
環境構築
このプロジェクトには開発コンテナ、Dockerfile、コントリビューションガイドがありません。まず README を読み、一般的な手順ははじめてのコントリビューションガイドを参照してください。
はじめの一歩
- issue を最後まで読み、次にプロジェクトのコントリビューションガイドを読みます。
- 着手することを issue にコメントします — 二人が同じ作業をするのを防げます。
- リポジトリをフォークし、ブランチを切って変更します。
- issue 番号を参照したプルリクエストを送ります。
callstackincubator/simlock のほかの issue
-
request:new
難易度 2/5 1〜3時間 初心者へのやさしさ 72/100
callstackincubator/simlock#1806 ·
メンテナーはふだん 1 日以内に返信
-
bug:new
難易度 2/5 1〜3時間 初心者へのやさしさ 78/100
callstackincubator/simlock#1780 · コメント 1 件 ·
メンテナーはふだん 1 日以内に返信
-
task:draft
難易度 2/5 1〜3時間 初心者へのやさしさ 72/100
callstackincubator/simlock#1441 ·
メンテナーはふだん 1 日以内に返信
-
bug:new
難易度 2/5 1〜3時間 初心者へのやさしさ 45/100
callstackincubator/simlock#1805 · コメント 1 件 ·
メンテナーはふだん 1 日以内に返信
-
bug:new
難易度 2/5 1〜3時間 初心者へのやさしさ 45/100
callstackincubator/simlock#1804 · コメント 1 件 ·
メンテナーはふだん 1 日以内に返信
callstackincubator/simlock の issue をすべて見る
似ている issue
-
難易度 1/5 1時間未満 初心者へのやさしさ 82/100
メンテナーはふだん 1 日以内に返信
-
Update project links after the repository renames対応中かも @JonathanTSilva が今日担当しました。 オープンai-generated priority: p2 size: xs type: bug
難易度 2/5 1〜3時間 初心者へのやさしさ 85/100
JonathanTSilva/jts-website#119 ·
メンテナーはふだん 1 日以内に返信
-
難易度 2/5 1〜3時間 初心者へのやさしさ 72/100
-
難易度 2/5 1〜3時間 初心者へのやさしさ 74/100
promplate/code-review#158 · コメント 1 件 · リアクション 1 件 ·
メンテナーはふだん 1 日以内に返信
-
[Docs] CONTRIBUTING points the Backend Remake milestone at a branch that does not exist対応中かも @franpts2 が今日担当しました。 オープンdocs
難易度 1/5 1時間未満 初心者へのやさしさ 90/100
メンテナーはふだん 3 日以内に返信