Repository navigation
[agent-cli] status names the EAS session's platform when no app is connected, and says the session is not checked (llp/0021) - #208
Conversation
46f584e to
73c50e9
Compare
| // session call, so the why says the session is recorded, not checked. | ||
| command: `${smokeCommand(cloudSession)} --eas`, | ||
| why: 'a dev server is running with an app connected and this project has an EAS Simulator session, so the gate runs against that session — a plain "smoke" would look for a simulator on this machine', | ||
| why: 'an EAS Simulator session is on record (not checked) and an app is connected, so the gate runs on that session, not on this machine; if the session ended, smoke starts one for this run and stops it after, which bills', |
There was a problem hiding this comment.
[suggestion] The new why says that if the recorded session ended, smoke starts one for this run, stops it after, and bills. smoke --eas only starts a session when it can install an app. For Expo Go it does. For a development build, ensureEasSessionAsync looks up a finished development-simulator build and returns before eas simulator when there is none (src/smoke/smokeAsync.ts noBuildReason, src/dev/openAppEas.ts). The same absolute claim is in the cloudSession note just above. The test regex locks the overclaim in.
Suggestion: Say smoke reuses the session when it is still up, and starts a new one (then stops it, which bills) only when it can install Expo Go or a finished simulator build. Otherwise it fails the start-session phase and names dev --<p> --eas. Point the JSDoc and the test at that same condition.
smoke --eas starts a session only when it can install Expo Go or a finished simulator build; otherwise it fails at start-session and names dev --<p> --eas. The why, the cloudSession JSDoc, the tests and the CHANGELOG now say that. Addresses Kudo's review on #208. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
smoke --eas starts a session only when it can install Expo Go or a finished simulator build; otherwise it fails at start-session and names dev --<p> --eas. The why, the cloudSession JSDoc, the tests and the CHANGELOG now say that. Addresses Kudo's review on #208. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
7b59d7c to
9b46bd2
Compare
smoke --eas starts a session only when it can install Expo Go or a finished simulator build; otherwise it fails at start-session and names dev --<p> --eas. The why, the cloudSession JSDoc, the tests and the CHANGELOG now say that. Addresses Kudo's review on #208. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
9b46bd2 to
e69c861
Compare
…nnected, and says the session is not checked (llp/0021) Problem (N2-26): on a worktree with a recorded EAS Simulator session, `status` beside a dev server with no app connected named `navigate / --eas` with no platform flag. With an app connected, it named `smoke --<p> --eas` with a why that read as if the session was known to be up, but status makes no session call. Change: the no-app step is `navigate / --<p> --eas`, with the platform of the recorded session, the same way the app-connected step derives it. The platform is the session's, not the report's: `status --ios` with only an Android session names `--android`. The app-connected step's why now says the session is on record (not checked), and that if it ended, smoke starts one for this run and stops it after, which bills. The `cloudSession` docstring now covers both steps. The app-connected step stays `smoke --<p> --eas`: #187 kept it on review ("and a live EAS binding, next stays smoke --android --eas"), and this commit only makes its why honest. Red before: sections-test "names the command that acts on the ios/android session" (got `navigate / --eas`), "names the cloud when the device probe could not answer" (same), "keeps the gate on the cloud when the app that is connected is there" and "names the platform of an Android session in the gate" (why did not match /on record \(not checked\).*starts one for this run/). Guard: "names the session's platform, not the report's" (0 and 1 apps connected) passes before and after. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
smoke --eas starts a session only when it can install Expo Go or a finished simulator build; otherwise it fails at start-session and names dev --<p> --eas. The why, the cloudSession JSDoc, the tests and the CHANGELOG now say that. Addresses Kudo's review on #208. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
e69c861 to
2a373e8
Compare
Problem
With an EAS Simulator session on record,
statussuggestednavigate / --easwith no platform when no app was connected (dogfood N2-26). Its reason forsmoke --<p> --easread as if the session were known to be live.Fix
With no app connected, the next step is
navigate / --<p> --eas, with the platform of the recorded session. With an app connected, the step stayssmoke --<p> --eas, and its reason says the session is on record but not checked, and that smoke starts a session for the run if that one ended.🤖 Generated with Claude Code