feat(subprocess): reshape the seam Node-ward for multi-consumer use

Review direction (tianyicui, PR #660): make the interface closer to Node's
API so the other process-running places can adopt it. The spec gains
per-stream stdio dispositions — 'pipe' (raw Readable/Writable for protocol
streams), 'inherit' (diagnostics to the parent), and collect mode ({maxBytes,
spill?} — the old bounded tail-keep shape, now with spill optional for
diagnostic tails). SubprocessOutcome carries exit facts only; collected
output stays readable through handle.collected after settlement (spill fds
are sealed at the settle boundary). The handle grows Node-style kill(signal)
(single signal, tree-scoped, no-op after settlement), terminate() (the
SIGTERM→grace→SIGKILL escalation, also driven by the spec signal),
waitForExit() (tree liveness, not just the direct child), and dispose()
(the cooperative stdin-EOF→SIGTERM→SIGKILL ladder from subagent-subprocess,
graces caller-supplied). Tree semantics are platform-correct: POSIX detached
groups with direct-child fallback; Windows taskkill /T with an injectable
runner. scrubbedParentEnv/SENSITIVE_ENV_PATTERN move to the seam as the one
shared scrub definition.

bash-local maps its config onto collect modes and batch stdin and reads
results through the collected readers; its kill() maps to terminate() so
task_kill keeps escalation semantics.
This commit is contained in:
Tianyi Cui
2026-07-26 14:07:42 +08:00
parent 5e6edce4ed
commit 12a7e38417
8 changed files with 786 additions and 261 deletions

View File

@@ -1,14 +1,17 @@
/**
* The subprocess seam (`ctx.subprocess`): spawn fully-specified
* commands into managed process groups with bounded, spill-backed output and
* escalated kills. Command defaulting, shell semantics, deadlines, and
* presentation belong to consumers — the bash executor seam is the owning
* The subprocess seam (`ctx.subprocess`): spawn fully-specified commands into
* managed process trees with Node-shaped stdio dispositions — raw pipes for
* protocol streams, inherit for diagnostics, bounded spill-backed collection
* for batch output — plus tree-scoped signalling and a cooperative dispose
* ladder. Command defaulting, shell semantics, deadlines, framing, and
* presentation belong to consumers; the bash executor seam is the owning
* template. The local implementation lives in
* `@deepseek-ai/dsh-subprocess-local`.
* @module @deepseek-ai/dsh-subprocess
*/
import { Context, Service } from 'cordis'
import { DSH_ENV_PREFIX } from './types.ts'
import type { SubprocessHandle, SubprocessSpawnSpec } from './types.ts'
export { DSH_ENV_PREFIX } from './types.ts'
@@ -16,13 +19,48 @@ export type {
CollectedOutput,
DshEnvironment,
DshEnvironmentKey,
SubprocessCollect,
SubprocessCollectedOutputs,
SubprocessDisposeGraces,
SubprocessHandle,
SubprocessOutcome,
SubprocessOutputMode,
SubprocessOutputRead,
SubprocessOutputReader,
SubprocessSpawnSpec,
SubprocessStdinMode,
SubprocessStdio,
} from './types.ts'
/**
* Credential-shaped environment names are NOT forwarded to children (the
* harness's own `DEEPSEEK_API_KEY`/secrets must not leak into a spawned
* process implicitly). One heuristic for every in-repo spawner; a
* deliberately supplied entry survives because explicit env layers merge
* after the scrub.
*/
export const SENSITIVE_ENV_PATTERN = /KEY|SECRET|TOKEN/i
/**
* The ambient parent environment minus credential-shaped names and minus all
* `DSH_*` names — the canonical base every harness child starts from. `PATH`,
* `HOME`, locale, and proxy variables survive, so child CLIs run normally;
* harness identity never leaks implicitly (a child that needs current `DSH_*`
* facts receives them through {@link SubprocessSpawnSpec.dshEnv}, and a
* deliberately forwarded credential goes through an explicit env layer, which
* merges after this scrub). Exported as a plain function so spawners that
* cannot route through the service (node-pty backends, SDK-managed
* transports) share the one scrub definition.
* @returns a fresh environment object safe to hand to a child spawn.
*/
export function scrubbedParentEnv(): Record<string, string> {
const env: Record<string, string> = {}
for (const [key, value] of Object.entries(process.env)) {
if (value !== undefined && !SENSITIVE_ENV_PATTERN.test(key) && !key.startsWith(DSH_ENV_PREFIX)) env[key] = value
}
return env
}
declare module 'cordis' {
interface Context {
subprocess: SubprocessService
@@ -37,13 +75,17 @@ declare module 'cordis' {
*
* Implementations must honor these semantics:
* - {@link spawn} returns immediately with a live handle; `done` resolves at
* process close and rejects only for spawn-level failures.
* - Output readers are offset-based and non-consuming, so independent readers
* never consume one another's output; lossy reads report truncation and the
* spill file holding the complete stream when one exists.
* - {@link SubprocessHandle.kill} and the spec's abort signal escalate
* SIGTERM→grace→SIGKILL across the whole process group.
* - Disposal kills all still-running managed processes and awaits their exit.
* process close with exit facts and rejects only for spawn-level failures.
* - Collect-mode readers are offset-based and non-consuming, so independent
* readers never consume one another's output; lossy reads report truncation
* and the spill file holding the complete stream when one exists. Piped
* streams are handed to the caller raw and never buffered here.
* - {@link SubprocessHandle.kill} signals without escalation,
* {@link SubprocessHandle.terminate} (and the spec's abort signal) escalates
* SIGTERM→grace→SIGKILL, and {@link SubprocessHandle.dispose} runs the
* cooperative EOF-first ladder — all tree-scoped on every platform.
* - Disposal of the service terminates all still-running managed processes
* and awaits their exit.
*/
export abstract class SubprocessService extends Service {
constructor(ctx: Context) {
@@ -53,8 +95,8 @@ export abstract class SubprocessService extends Service {
/**
* Start one managed child process from a fully-specified spec; this seam
* applies no defaults.
* @param spec - argv, directory, limits, grace, cancellation, and environment.
* @returns the live process handle (readers, kill, outcome promise).
* @param spec - argv, directory, stdio dispositions, grace, cancellation, and environment.
* @returns the live process handle (streams/readers, signalling, outcome promise).
*/
abstract spawn(spec: SubprocessSpawnSpec): SubprocessHandle
}