mirror of
https://github.com/stablyai/orca.git
synced 2026-10-25 08:04:05 +00:00
Jinwoo Hong
3f9582bc90
* fix(mobile): removing a host deletes its page cache through the one process store Host removal never deleted the removed host's page generation cache, so it sat on disk until four other hosts were activated and a re-pair could reopen the old tree. Every caller also minted its own GenerationStore with its own queue, so a removal's index or update-failure-log write could drop the mounted session's write landing inside it. The store is now one per process (processGenerationStore, with a reset-for-tests seam); the shell runtime, host removal and Troubleshoot all share it, and removal deletes the host's cache after the metadata commits, fire-and-forget beside the failure forget. Claude-Session: https://claude.ai/code/session_01JNnE9qzUZMMnqpZWCqM3nb * test(mobile): the cold-start store per mount is the test's, not production's Claude-Session: https://claude.ai/code/session_01JNnE9qzUZMMnqpZWCqM3nb
48 lines
2.0 KiB
TypeScript
48 lines
2.0 KiB
TypeScript
import * as ExpoCrypto from 'expo-crypto'
|
|
import { encodeBase64Url } from '../transport/mobile-endpoint-supervisor-support'
|
|
import { BRIDGE_READY_RETRY_MAX_MS } from './bridge/bridge-client-init-handshake'
|
|
import type { GenerationStore } from './generation-store'
|
|
import { processGenerationStore } from './process-generation-store'
|
|
|
|
/**
|
|
* Everything the shell session touches that a test cannot: entropy, the clock, the filesystem.
|
|
*
|
|
* It lives beside the hook rather than inside it so the suite can drive all three without a
|
|
* simulator, and so the one number the shell waits on has somewhere honest to be stated.
|
|
*/
|
|
|
|
/** 32 bytes, base64url: the session id scopes the view's private origin, so two mounts must never
|
|
* share one and a remount must never reuse the one that was just on screen. */
|
|
const SESSION_ID_BYTES = 32
|
|
|
|
/**
|
|
* How long a finished document has to say `ready` before the shell gives up on it.
|
|
*
|
|
* Five times the page's own 2 s retry ceiling (`BRIDGE_READY_RETRY_MAX_MS`), so no device is slow
|
|
* enough for the backoff to outlast the wait: whatever the page is doing, several of its asks fit
|
|
* inside this. What it bounds is a page that will never ask at all, because a route module threw
|
|
* while the bundle was being evaluated and nothing downstream of that import ever ran.
|
|
*/
|
|
export const PAGE_READY_DEADLINE_MS = BRIDGE_READY_RETRY_MAX_MS * 5
|
|
|
|
export type MobileWebShellRuntime = {
|
|
createStore(): GenerationStore
|
|
mintSessionId(): string
|
|
now(): number
|
|
/** Runs `run` once, `delayMs` from now, and returns the cancel. The seam the deadline test drives
|
|
* instead of waiting ten seconds for a real one. */
|
|
setTimer(run: () => void, delayMs: number): () => void
|
|
}
|
|
|
|
export function createMobileWebShellRuntime(): MobileWebShellRuntime {
|
|
return {
|
|
createStore: processGenerationStore,
|
|
mintSessionId: () => encodeBase64Url(ExpoCrypto.getRandomBytes(SESSION_ID_BYTES)),
|
|
now: Date.now,
|
|
setTimer: (run, delayMs) => {
|
|
const handle = setTimeout(run, delayMs)
|
|
return () => clearTimeout(handle)
|
|
}
|
|
}
|
|
}
|