Files
paperclip/scripts/__tests__/runner-release-crate-cache.test.mjs
T
DottaandClaude Opus 5.5 b5b972ceb1 fix(runner): stop provider tool processes that leave the owned process group (#19727)
## Thinking Path

> - Paperclip is the open source app people use to manage AI agents for
work
> - The local runner (`paperclip-runnerd`) supervises each provider
process, such as the ACPX sidecar for Claude, in its own process group
> - Run teardown stops that process group, and nothing else
> - Claude Code starts each Bash tool command in a new session
(`setsid`). A command such as `nohup ./bench.sh & echo $!` leaves a job
in that session, and the wrapper exits at once. The job is then a child
of init and outside the owned group
> - In a restart chaos storm, a run failed after a restart resume. Its
tool job kept running beside the task's next run, in the same workspace,
for more than four minutes
> - This pull request makes the supervisor record every process that its
provider starts, including processes that leave the group, and stop them
all before teardown reports success
> - The benefit is that a run's tools cannot outlive the run and cause
duplicate side effects in the next run

## Linked Issues or Issue Description

**What happened?**
A restart chaos storm ran a local ACPX Claude task. After a restart
resume, the agent ran `nohup ./bench.sh <nonce> > log 2>&1 & echo
"started pid $!"` and then ended its turn without a result. The run
failed with `native_finalization_missing`. The runner tore down the
sidecar group, and the guardian killed the provider group. `bench.sh`
was in the session that Claude Code made for the tool (pgid and sid =
the exited wrapper, ppid = 1). It was never signalled. It kept forking
`sleep 1` while the task's next run started and finished in the same
workspace.

**Steps to reproduce**
1. Run a local ACPX Claude task whose agent runs a Bash tool command
such as `nohup ./bench.sh x > log 2>&1 & echo $!`.
2. End the run (any outcome, including failure after a restart resume).
3. Run `ps -axo pid,ppid,pgid,command | grep bench.sh`. The job is still
running with ppid 1.

**Agent adapter(s) involved**
Claude Code (local ACPX). The gap is in the shared supervisor, so any
provider whose tools leave the group is affected.

**Root cause**
`SupervisedProcess`
(`packages/paperclip-runner/runner/crates/runner-core/src/process_supervisor.rs`)
spawns with `process_group(0)` and stops only that group in
`terminate_group`, `terminate_group_immediately`, `wait` and `Drop`. Its
own comment said that a caller that must exclude escaped or re-parented
descendants needs a separate lifetime fence. No caller had one. The ACPX
sidecar transport (`acpx_sidecar_transport.rs`, `shutdown` and `poison`)
and the ACPX provider guardian (`installation-integrity.ts`, which kills
only its own group) therefore left every `setsid` tool job behind. The
failure path after the resume was not special: the same gap applies to
every teardown.

**Expected behavior**
When a run ends for any reason, every process that its provider or tools
started stops, and teardown proves that they are gone. Identity comes
from PID, group, session and start time, never from a command name.

## What Changed

- Add the `paperclip-process-table` crate. It reads PID, parent, process
group and start time for every process (libproc on macOS, `/proc` on
Linux). `runner-core` forbids unsafe code, so the two checked libproc
calls live in this small crate.
- Add `process_lineage.rs` to `runner-core`. A sampler thread (every 50
ms) records each descendant of the supervised process by PID and start
time. It also records each group and session that a descendant leads,
with the leader's start time. A process belongs to the lineage when its
parent belongs, or when it is in a recorded group or session and started
no earlier than its leader. Groups and sessions that the supervised
process only inherited (runnerd's session) are never recorded. A
recorded ID whose PID now names a different process is dropped.
- `SupervisedProcess` samples the lineage just before it signals, and
after the leader is reaped it stops every live member: TERM, then KILL
after the shutdown grace. It fails with an error until it proves that no
member is alive (grace plus 5 s). This applies to `terminate_group`,
`terminate_group_immediately`, `wait` and `Drop`.
- If the lineage sampler cannot start, the supervisor stops the process
it just started and reports the failure, so a provider never runs
without tool cleanup.
- Table reads and state updates happen under one lock, so an older read
cannot erase members that a newer read recorded.
- The process table reports processes that exist but cannot be read. A
recorded member that cannot be read stays unresolved, and teardown fails
until its exit is proven.
- The Runner workspace now has two members.
`scripts/runner-release-crate-cache.mjs` and
`scripts/runner-debug-crate-cache.mjs` cache and mark current the
process-table lib unit with the Runner crate (otherwise Cargo rebuilds
the Runner crate after every restore). The debug contract accepts only
this library-only member and still refuses any other workspace change.
The workflow cache path lists and the debug clean step are updated to
match.
- `terminate_group_immediately` now accepts `EPERM` from the group
signal when the leader has already exited. macOS returns `EPERM` for a
group whose only member is the unreaped leader. The lineage pass then
proves that no other member remains.

## Verification

- `cargo test -p paperclip-runner-core --lib -- process_lineage tool_job
immediate_group_fence` (macOS, local).
- New real-process tests in `process_supervisor.rs`: a supervised shell
starts a wrapper with `setsid` (Perl `POSIX::setsid`), the wrapper runs
`nohup <loop> &`, writes the job PID and exits. The test asserts that
the job is outside the owned group and re-parented, then calls
`terminate_group`, `terminate_group_immediately` or drops the process,
and asserts that the job is gone when the call returns.
- Red: with the tracker disabled, all three fail with `tool job <pid>
outlived its provider`.
  - Green: all three pass.
- Unit tests in `process_lineage.rs` cover a re-parented job, a job that
left its group but stayed in the recorded session, a session made after
the first sighting, PID reuse, and no claim through the supervisor's
inherited session.
- `node --test scripts/__tests__/runner-release-crate-cache.test.mjs
scripts/__tests__/runner-debug-crate-cache.test.mjs` (new case: the
process-table library is cached with the crate, and any other member
still needs review).
- `cargo test -p paperclip-process-table` (includes an exited process
reported as gone), and `cargo test -p paperclip-runner-core --lib` (570
passed).
`acpx_provider_backend::tests::unconfirmed_suspension_state_becomes_recoverable_only_after_cleanup_and_suspend`
fails sometimes under full-suite load. It fails at the same rate with
the tracker disabled, so it is a flake that already existed.
- Sampler cost: one table read takes about 0.4 ms for 1,200 processes on
macOS (about 0.8% of one core for each supervised process).
- The restart-chaos replay (`--replay w-20261107T023921-s7106/t67`) did
not run. The local session was not permitted to load provider
credentials. It is not proven against a live Claude run.

## Risks

- Teardown now stops all the processes that a provider started, also
intentional background jobs. This is the intended contract: nothing
outlives the run.
- A wrapper that lives less than one sample (50 ms) and orphans a job
before the sampler sees it is not recorded. Claude Code's wrapper
sources a shell snapshot and lives for seconds. On Linux, a
child-subreaper would close this gap. That is not in this change.
- When `runnerd` itself is killed with SIGKILL, this code does not run.
The provider guardian still kills its own group, but escaped tool
sessions are not stopped in that case.
- Teardown can return a new error when it cannot prove the lineage
stopped. Callers already handle `terminate_group` errors as provider
shutdown failures.
- The server does not yet gate the next run on the tool lineage of the
prior run directly. It relies on the runner finishing its teardown
before the runner exits.

## Model Used

- Claude Opus 5.5 (`claude-opus-5-5`), Anthropic, with tool use and code
execution in Claude Code.

## Checklist

- [x] I have included a thinking path that traces from project context
to this change
- [x] I have specified the model used (with version and capability
details)
- [x] I have checked ROADMAP.md and confirmed this PR does not duplicate
planned core work
- [x] I have searched GitHub for duplicate or related PRs and linked
them above
- [x] I have either (a) linked existing issues with `Fixes: #` / `Closes
#` / `Refs #` OR (b) described the issue in-PR following the relevant
issue template
- [x] I have not referenced internal/instance-local Paperclip issues or
links (only public GitHub `#NNN` / `github.com/paperclipai/paperclip`
URLs)
- [x] My branch name describes the change (e.g. `docs/...`, `fix/...`)
and contains no internal Paperclip ticket id or instance-derived details
- [x] I have run tests locally and they pass
- [x] I have added or updated tests where applicable
- [x] I have updated relevant documentation to reflect my changes
- [x] I have considered and documented any risks above
- [ ] All Paperclip CI gates are green
- [ ] Greptile is 5/5 with no open P2s, recommendations, or follow-ups
- [x] I will address all Greptile and reviewer comments before
requesting merge

🤖 Generated with [Claude Code](https://claude.com/claude-code)

---------

Co-authored-by: Claude Opus 5.5 <[email protected]>
2026-11-07 08:24:41 -06:00

625 lines
34 KiB
JavaScript

import assert from "node:assert/strict";
import { execFileSync, spawnSync } from "node:child_process";
import { chmodSync, mkdirSync, mkdtempSync, readdirSync, readFileSync, rmSync, statSync, utimesSync, writeFileSync } from "node:fs";
import os from "node:os";
import path from "node:path";
import test from "node:test";
import { fileURLToPath } from "node:url";
import { CACHED_PATHS, HASHED_INPUTS, REQUIRED, RUNNER_PACKAGE, computeScope, markCurrent } from "../runner-release-crate-cache.mjs";
const repoRoot = path.resolve(path.dirname(fileURLToPath(import.meta.url)), "../..");
const script = path.join(repoRoot, "scripts/runner-release-crate-cache.mjs");
const readWorkflow = (name) => readFileSync(path.join(repoRoot, ".github/workflows", name), "utf8");
const prWorkflow = readWorkflow("pr-trusted.yml");
const releaseWorkflow = readWorkflow("release-verify.yml");
// Slice one job out of a workflow, from its two-space-indented key to the next.
function job(workflow, name) {
const start = workflow.indexOf(`\n ${name}:\n`);
assert.ok(start >= 0, `workflow is missing the ${name} job`);
const body = workflow.slice(start + 1);
const header = ` ${name}:\n`.length;
const next = body.slice(header).search(/\n [a-z0-9_-]+:\n/);
return next === -1 ? body : body.slice(0, header + next + 1);
}
// Slice one step out of a job by its name, up to the next step.
function step(body, name) {
const marker = ` - name: ${name}\n`;
const start = body.indexOf(marker);
assert.ok(start >= 0, `missing step: ${name}`);
const rest = body.slice(start + marker.length);
const next = rest.search(/\n - name: /);
return marker + (next === -1 ? rest : rest.slice(0, next + 1));
}
const RELEASE_DIR = `${RUNNER_PACKAGE}/runner/target/release`;
const KEY_STEP = "Compute the Runner release crate cache key";
const RESTORE_STEP = "Restore the Runner release crate (read only)";
const MARK_STEP = "Mark the restored Runner release crate current";
const LOOKUP_STEP = "Check for this push's Runner release crate";
const BUILD_STEP = "Build the Runner release crate for its cache";
const SAVE_STEP = "Save the Runner release crate";
const RUST_RESTORE_STEP = "Restore Runner Rust dependencies (read only)";
// Every PR job that builds the Runner release binary, and the command each
// one runs cargo through. The crate must be restored and marked before it.
const READERS = [
["typecheck_release_registry", "run: pnpm run typecheck:build-gaps"],
["verify_paperclip_runner", "run: pnpm --filter @paperclipai/paperclip-runner ${{ matrix.command }}"],
["build", "run: pnpm build"],
["canary_dry_run", "- name: Release canary dry run via release.sh internal build"],
];
const WRITER_GATE = "matrix.lane == 'rust' && github.repository == 'paperclipai/paperclip' && github.event_name == 'push' && github.ref == 'refs/heads/master' && inputs.ref == github.sha";
const expectedPathBlock = ` path: |\n${CACHED_PATHS.map((entry) => ` ${entry}\n`).join("")}`;
const expectedKey = `key: runner-release-crate-v1-\${{ steps.runner_release_crate_key.outputs.scope }}-\${{ hashFiles(${HASHED_INPUTS.map((entry) => `'${entry}'`).join(", ")}) }}`;
function assertCacheStep(body, name, { lookupOnly = false } = {}) {
assert.match(body, /uses: actions\/cache\/(restore|save)@55cc8345863c7cc4c66a329aec7e433d2d1c52a9 # v6\.1\.0/, name);
assert.ok(body.includes(expectedPathBlock), `${name}: the cached path list must be exactly the script's CACHED_PATHS`);
assert.ok(body.includes(expectedKey), `${name}: the key must be built from the script's scope and HASHED_INPUTS`);
// An exact match or nothing: a prefix match would restore another
// revision's artifacts and the mark step would make cargo keep them.
assert.doesNotMatch(body, /restore-keys:/, name);
assert.equal(/lookup-only: true/.test(body), lookupOnly, name);
assert.doesNotMatch(body, /enableCrossOsArchive|fail-on-cache-miss: true/, name);
}
test("the script's path list and key inputs cover what the release build of the crate reads and writes", () => {
assert.deepEqual(CACHED_PATHS, [
`${RELEASE_DIR}/.fingerprint/paperclip-runner-core-*`,
`${RELEASE_DIR}/deps/libpaperclip_runner_core-*`,
`${RELEASE_DIR}/deps/paperclip_runner_core-*.d`,
`${RELEASE_DIR}/deps/paperclip_runnerd-*`,
`${RELEASE_DIR}/paperclip-runnerd`,
`${RELEASE_DIR}/paperclip-runnerd.d`,
`${RELEASE_DIR}/.fingerprint/paperclip-process-table-*`,
`${RELEASE_DIR}/deps/libpaperclip_process_table-*`,
`${RELEASE_DIR}/deps/paperclip_process_table-*.d`,
]);
assert.deepEqual(HASHED_INPUTS, [
`${RUNNER_PACKAGE}/rust-toolchain.toml`,
`${RUNNER_PACKAGE}/acpx-profiles.json`,
`${RUNNER_PACKAGE}/antigravity-distributions.json`,
`${RUNNER_PACKAGE}/runner/Cargo.toml`,
`${RUNNER_PACKAGE}/runner/Cargo.lock`,
`${RUNNER_PACKAGE}/runner/crates/**`,
`${RUNNER_PACKAGE}/protocol/**`,
`${RUNNER_PACKAGE}/test-fixtures/**`,
`${RUNNER_PACKAGE}/test/fixtures/**`,
"tests/runner-e2e/fixtures/**",
]);
for (const input of HASHED_INPUTS) {
const fixed = input.replace(/\/\*\*$/, "");
assert.ok(statSync(path.join(repoRoot, fixed)), `${input} must exist in the repository`);
}
// The stage step copies the uplifted binary; it has to be in the set.
const stage = readFileSync(path.join(repoRoot, RUNNER_PACKAGE, "scripts/stage-runner-binary.mjs"), "utf8");
assert.match(stage, /"runner", "target", "release", executable/);
assert.ok(CACHED_PATHS.includes(`${RELEASE_DIR}/paperclip-runnerd`));
});
// Every `include_str!` / `include_bytes!` in one Rust source, with the path
// each one names. Three argument forms are resolved: a string literal
// (relative to the source file), and `concat!(...)` of string literals,
// `env!("CARGO_MANIFEST_DIR")` (the directory of the nearest Cargo.toml) and
// macro metavariables (`$name`, which make the resolved path a directory
// prefix rather than a file). Anything else is reported as unsupported, so
// a new include form cannot slip past the guard by not matching a regex.
function crateIncludes(source, text) {
const includes = [];
for (const match of text.matchAll(/include_(?:str|bytes)!\s*\(/g)) {
const start = match.index + match[0].length;
let depth = 1;
let inString = false;
let end = start;
for (; end < text.length && depth > 0; end += 1) {
const ch = text[end];
if (inString) {
if (ch === "\\") end += 1;
else if (ch === '"') inString = false;
} else if (ch === '"') inString = true;
else if (ch === "(") depth += 1;
else if (ch === ")") depth -= 1;
}
assert.equal(depth, 0, `${source}: unbalanced include at offset ${match.index}`);
const argument = text.slice(start, end - 1).trim();
includes.push({ argument, ...resolveIncludeArgument(source, argument) });
}
return includes;
}
function manifestDir(source) {
let dir = path.dirname(source);
while (!statSync(path.join(dir, "Cargo.toml"), { throwIfNoEntry: false })?.isFile()) {
const parent = path.dirname(dir);
assert.notEqual(parent, dir, `${source}: no Cargo.toml above the source`);
dir = parent;
}
return dir;
}
function resolveIncludeArgument(source, argument) {
const literal = argument.match(/^"((?:[^"\\]|\\.)*)"$/);
if (literal) return { form: "literal", resolved: path.resolve(path.dirname(source), literal[1]), exact: true };
const concat = argument.match(/^concat!\s*\(([\s\S]*)\)$/);
assert.ok(concat, `${source}: unsupported include form: ${argument}`);
const parts = splitTopLevel(concat[1]);
// A relative path is joined to the source's directory; one that starts
// with the manifest directory is one string, since its next literal
// starts with the "/" that separates the two.
let base = `${path.dirname(source)}/`;
let text = "";
let exact = true;
let form = "concat";
for (const [index, part] of parts.entries()) {
const string = part.match(/^"((?:[^"\\]|\\.)*)"$/);
if (string) {
assert.ok(exact || !/(^|\/)\.\.(\/|$)/.test(string[1]), `${source}: a path segment after a metavariable must not climb out: ${argument}`);
if (exact) text += string[1];
continue;
}
if (part === 'env!("CARGO_MANIFEST_DIR")') {
assert.equal(index, 0, `${source}: env!("CARGO_MANIFEST_DIR") must be the first part of concat!: ${argument}`);
base = manifestDir(source);
form = "concat-manifest-dir";
continue;
}
if (/^\$[A-Za-z_][A-Za-z0-9_]*$/.test(part)) {
// The fixed prefix before the first metavariable is the directory the
// macro's callers can reach; the suffix cannot climb out of it.
exact = false;
form = form === "concat-manifest-dir" ? form : "concat-metavariable";
continue;
}
assert.fail(`${source}: unsupported concat! part in include: ${part}`);
}
const joined = base + text;
const resolved = path.resolve(exact || text.endsWith("/") ? joined : path.dirname(joined));
return { form, resolved, exact };
}
// Split concat!'s arguments on the commas outside string literals.
function splitTopLevel(text) {
const parts = [];
let current = "";
let inString = false;
for (let index = 0; index < text.length; index += 1) {
const ch = text[index];
if (inString) {
current += ch;
if (ch === "\\") current += text[++index] ?? "";
else if (ch === '"') inString = false;
} else if (ch === '"') {
inString = true;
current += ch;
} else if (ch === ",") {
parts.push(current);
current = "";
} else current += ch;
}
parts.push(current);
return parts.map((part) => part.trim()).filter(Boolean);
}
const underHashedInputs = (resolved) =>
HASHED_INPUTS.map((input) => path.join(repoRoot, input.replace(/\/\*\*$/, ""))).some((root) => resolved === root || resolved.startsWith(`${root}${path.sep}`));
test("the include parser resolves literals, concat! with the manifest directory and with metavariables, and rejects other forms", () => {
const root = mkdtempSync(path.join(os.tmpdir(), "runner-release-crate-includes-"));
try {
const crate = path.join(root, "crates/demo");
mkdirSync(path.join(crate, "src/nested"), { recursive: true });
mkdirSync(path.join(root, "protocol/fixtures"), { recursive: true });
writeFileSync(path.join(crate, "Cargo.toml"), '[package]\nname = "demo"\n');
writeFileSync(path.join(root, "protocol/schema.json"), "{}");
const source = path.join(crate, "src/nested/mod.rs");
const text = [
'let a = include_str!("../../../../protocol/schema.json");',
'let b = include_bytes!(\n "../../../../protocol/schema.json"\n);',
'let c = include_str!(concat!(env!("CARGO_MANIFEST_DIR"), "/../../protocol/", "schema.json"));',
'let d = include_str!(concat!("../../../../protocol/fixtures/", $fixture, ".json"));',
'let e = include_str!(concat!(env!("CARGO_MANIFEST_DIR"), "/../../protocol/fixtures/", $golden, ".summary.json"));',
'let f = include_str!("a (paren) in the name.json");',
].join("\n");
const includes = crateIncludes(source, text);
assert.deepEqual(
includes.map(({ form, resolved, exact }) => [form, path.relative(root, resolved), exact]),
[
["literal", "protocol/schema.json", true],
["literal", "protocol/schema.json", true],
["concat-manifest-dir", "protocol/schema.json", true],
["concat-metavariable", "protocol/fixtures", false],
["concat-manifest-dir", "protocol/fixtures", false],
["literal", "crates/demo/src/nested/a (paren) in the name.json", true],
],
);
const rejected = [
'include_str!(concat!("../protocol/", $x, "/../../../outside.json"))',
'include_str!(concat!("../protocol/", env!("CARGO_MANIFEST_DIR")))',
'include_str!(concat!(env!("OTHER_DIR"), "/schema.json"))',
'include_str!(r"../protocol/schema.json")',
'include_str!(path_from_a_function())',
'include_str!(concat!("../protocol/", some_macro!("x")))',
];
for (const argument of rejected) {
assert.throws(() => crateIncludes(source, argument), /unsupported|must not climb out|must be the first part/, argument);
}
} finally {
rmSync(root, { recursive: true, force: true });
}
});
test("every file the crate includes by path sits under a hashed input, so an edit to it changes the key", () => {
const crateRoot = path.join(repoRoot, RUNNER_PACKAGE, "runner/crates");
const sources = [];
(function walk(dir) {
for (const entry of readdirSync(dir, { withFileTypes: true })) {
const full = path.join(dir, entry.name);
if (entry.isDirectory()) walk(full);
else if (entry.name.endsWith(".rs")) sources.push(full);
}
})(crateRoot);
assert.ok(sources.length > 10, "the crate sources must be found");
const forms = new Set();
let includes = 0;
for (const source of sources) {
for (const { argument, form, resolved, exact } of crateIncludes(source, readFileSync(source, "utf8"))) {
includes += 1;
forms.add(form);
const where = `${path.relative(repoRoot, source)} includes ${argument.replace(/\s+/g, " ")}`;
assert.ok(underHashedInputs(resolved), `${where}, which resolves to ${path.relative(repoRoot, resolved)}, outside every HASHED_INPUTS pattern`);
const info = statSync(resolved, { throwIfNoEntry: false });
assert.ok(exact ? info?.isFile() : info?.isDirectory(), `${where}: ${path.relative(repoRoot, resolved)} must exist`);
}
}
assert.ok(includes > 0, "the crate includes protocol files by path; the guard must see them");
// The crate uses every supported form today, so none of the parser's
// branches is exercised only by the synthetic test above.
assert.deepEqual([...forms].sort(), ["concat-manifest-dir", "concat-metavariable", "literal"]);
});
test("isolated Docker Rust stages preserve the layout of embedded schema and identity inputs", () => {
const crateRoot = path.join(repoRoot, RUNNER_PACKAGE, "runner/crates");
const sources = [];
(function walk(dir) {
for (const entry of readdirSync(dir, { withFileTypes: true })) {
const full = path.join(dir, entry.name);
if (entry.isDirectory()) walk(full);
else if (entry.name.endsWith(".rs")) sources.push(full);
}
})(crateRoot);
for (const [dockerfile, stage] of [["Dockerfile", "runner-build"], ["docker/daytona-runner/Dockerfile", "runnerd-build"]]) {
const docker = readFileSync(path.join(repoRoot, dockerfile), "utf8");
const body = docker.split(/^FROM .+ AS /m).find(part => part.startsWith(`${stage}\n`));
assert.ok(body, `${dockerfile}: missing ${stage}`);
const workdir = /^WORKDIR (.+)$/m.exec(body)?.[1];
assert.ok(workdir?.startsWith("/"), `${dockerfile}: an absolute stage working directory is required`);
const copies = [...body.matchAll(/^COPY (?!-)(.+)$/gm)].flatMap(([, instruction]) => {
const args = instruction.trim().split(/\s+/);
const destination = path.posix.resolve(workdir, args.pop());
return args.map(source => {
const absolute = path.join(repoRoot, source);
return { source: absolute, destination: statSync(absolute).isDirectory() ? destination : path.posix.join(destination, path.basename(source)) };
});
});
const copiedPath = source => {
const copy = copies.find(copy => source === copy.source || source.startsWith(`${copy.source}${path.sep}`));
return copy ? path.posix.join(copy.destination, path.relative(copy.source, source)) : null;
};
let checked = 0;
for (const source of sources) {
for (const { resolved } of crateIncludes(source, readFileSync(source, "utf8"))) {
// Cargo test-only fixture trees are intentionally absent from production
// Docker stages. Their cache coverage is checked by the preceding test.
if (/\/(?:test-fixtures|test\/fixtures|tests\/runner-e2e\/fixtures)(?:\/|$)/.test(resolved)) continue;
const stagedSource = copiedPath(source);
assert.ok(stagedSource, `${dockerfile}: crate source was not copied: ${path.relative(repoRoot, source)}`);
const expected = path.posix.resolve(path.posix.dirname(stagedSource), path.relative(path.dirname(source), resolved));
assert.equal(copiedPath(resolved), expected, `${dockerfile}: include input ${path.relative(repoRoot, resolved)} must retain its crate-relative path`);
checked++;
}
}
assert.ok(checked > 0, `${dockerfile}: the guard must see real Rust include paths`);
}
});
test("every PR reader restores the crate after the dependency cache and marks it before building", () => {
for (const [name, buildCommand] of READERS) {
const body = job(prWorkflow, name);
const positions = [RUST_RESTORE_STEP, KEY_STEP, RESTORE_STEP, MARK_STEP].map((stepName) => body.indexOf(` - name: ${stepName}\n`));
assert.ok(positions.every((position) => position >= 0), `${name}: a crate cache step is missing`);
assert.deepEqual([...positions].sort((a, b) => a - b), positions, `${name}: the crate steps must follow the dependency restore in order`);
const build = body.indexOf(buildCommand);
assert.ok(build > positions.at(-1), `${name}: the crate must be marked before ${buildCommand}`);
const key = step(body, KEY_STEP);
assert.match(key, /id: runner_release_crate_key\n/, name);
assert.match(key, /run: echo "scope=\$\(node scripts\/runner-release-crate-cache\.mjs scope\)" >> "\$GITHUB_OUTPUT"/, name);
const restore = step(body, RESTORE_STEP);
assert.match(restore, /id: runner_release_crate\n/, name);
assertCacheStep(restore, `${name} restore`);
assert.doesNotMatch(restore, /actions\/cache\/save/, name);
const mark = step(body, MARK_STEP);
assert.match(mark, /if: \$\{\{ steps\.runner_release_crate\.outputs\.cache-hit == 'true' \}\}/, name);
assert.match(mark, /run: node scripts\/runner-release-crate-cache\.mjs mark-current/, name);
}
// No other job restores or saves the crate.
const total = prWorkflow.match(/- name: Restore the Runner release crate \(read only\)\n/g)?.length ?? 0;
assert.equal(total, READERS.length);
assert.doesNotMatch(prWorkflow, /actions\/cache\/save@/);
});
test("the Verify Paperclip Runner job restores the crate only in the lanes that build it", () => {
const body = job(prWorkflow, "verify_paperclip_runner");
const lanes = JSON.parse(readFileSync(path.join(repoRoot, ".github/ci-task-matrix.json"), "utf8"))
.verify_paperclip_runner.map((entry) => [entry.lane_label, entry.release_crate === true]);
assert.deepEqual(lanes, [
["static checks", false],
["rust", true],
["vitest package (1/3)", false],
["vitest package (2/3)", false],
["vitest package (3/3)", false],
["vitest native server", true],
]);
for (const stepName of [KEY_STEP, RESTORE_STEP]) {
assert.match(step(body, stepName), /if: \$\{\{ matrix\.release_crate == true \}\}/, stepName);
}
// A lane that skipped the restore has no cache-hit output, so the mark
// step's own condition is enough there.
assert.match(step(body, MARK_STEP), /if: \$\{\{ steps\.runner_release_crate\.outputs\.cache-hit == 'true' \}\}/);
for (const [name] of READERS.filter(([name]) => name !== "verify_paperclip_runner")) {
for (const stepName of [KEY_STEP, RESTORE_STEP]) {
assert.doesNotMatch(step(job(prWorkflow, name), stepName), /^\s+if:/m, `${name}: ${stepName} is unconditional`);
}
}
});
test("only the master Rust lane writes the crate, after a lookup, with the readers' exact path list and key", () => {
const body = job(releaseWorkflow, "verify_paperclip_runner");
const positions = ["Verify Paperclip Runner", "Warm debug dependencies for the shared Runner cache", KEY_STEP, LOOKUP_STEP, BUILD_STEP, SAVE_STEP].map((stepName) => body.indexOf(` - name: ${stepName}\n`));
assert.ok(positions.every((position) => position >= 0), "a writer step is missing");
assert.deepEqual([...positions].sort((a, b) => a - b), positions, "the writer steps must follow the checks and the debug warm-up, in order");
for (const stepName of [KEY_STEP, LOOKUP_STEP]) {
assert.match(step(body, stepName), new RegExp(`if: \\$\\{\\{ ${WRITER_GATE.replace(/[.*+?^${}()|[\]\\]/g, "\\$&")} \\}\\}`), stepName);
}
for (const stepName of [BUILD_STEP, SAVE_STEP]) {
assert.match(
step(body, stepName),
new RegExp(`if: \\$\\{\\{ ${WRITER_GATE.replace(/[.*+?^${}()|[\]\\]/g, "\\$&")} && steps\\.runner_release_crate\\.outputs\\.cache-hit != 'true' \\}\\}`),
stepName,
);
}
assert.match(step(body, KEY_STEP), /run: echo "scope=\$\(node scripts\/runner-release-crate-cache\.mjs scope\)" >> "\$GITHUB_OUTPUT"/);
const lookup = step(body, LOOKUP_STEP);
assert.match(lookup, /id: runner_release_crate\n/);
assert.match(lookup, /uses: actions\/cache\/restore@/);
assertCacheStep(lookup, "writer lookup", { lookupOnly: true });
// The saved artifacts come from the readers' exact command, on a crate
// cleaned of the test-profile variants the lane's checks built.
const build = step(body, BUILD_STEP);
assert.match(build, /working-directory: packages\/paperclip-runner\n/);
assert.match(build, /set -euo pipefail\n\s+cargo clean --release --manifest-path runner\/Cargo\.toml -p paperclip-runner-core\n\s+pnpm build:binary\n/);
const runnerScripts = JSON.parse(readFileSync(path.join(repoRoot, RUNNER_PACKAGE, "package.json"), "utf8")).scripts;
assert.match(runnerScripts["build:binary"], /^cargo build --release --manifest-path runner\/Cargo\.toml --locked -p paperclip-runner-core --bin paperclip-runnerd && /);
const save = step(body, SAVE_STEP);
assert.match(save, /uses: actions\/cache\/save@/);
assertCacheStep(save, "writer save");
// The dependency cache's own writer gate is the same expression.
assert.match(body, new RegExp(`save-if: \\$\\{\\{ ${WRITER_GATE.replace(/[.*+?^${}()|[\]\\]/g, "\\$&")} \\}\\}`));
assert.equal((releaseWorkflow.match(/actions\/cache\/save@/g) ?? []).length, 2, "release and debug workspace caches have separate guarded writers");
});
test("the scope names the platform, the compiler and the checkout root, and survives a missing rustc", () => {
const a = computeScope({ workspace: "/home/runner/_work/paperclip/paperclip", rustcVersion: "1.97.1-abcdef123-x86_64-unknown-linux-gnu" });
const b = computeScope({ workspace: "/home/runner/work/paperclip/paperclip", rustcVersion: "1.97.1-abcdef123-x86_64-unknown-linux-gnu" });
const c = computeScope({ workspace: "/home/runner/_work/paperclip/paperclip", rustcVersion: "1.97.2-0123456ab-x86_64-unknown-linux-gnu" });
assert.notEqual(a, b, "a different checkout root must select a different entry");
assert.notEqual(a, c, "a different compiler must select a different entry");
assert.equal(a, computeScope({ workspace: "/home/runner/_work/paperclip/paperclip/", rustcVersion: "1.97.1-abcdef123-x86_64-unknown-linux-gnu" }));
assert.match(a, new RegExp(`^${os.platform()}-${os.arch()}-1\\.97\\.1-abcdef123-x86_64-unknown-linux-gnu-[0-9a-f]{12}$`));
assert.doesNotMatch(computeScope({ rustcVersion: "weird version/with spaces" }), /[^A-Za-z0-9._-]/);
// The CLI reads rustc from PATH; a stand-in proves which fields it reads,
// and an empty PATH proves a missing compiler still yields a usable key.
const bin = mkdtempSync(path.join(os.tmpdir(), "runner-release-crate-rustc-"));
try {
const stub = path.join(bin, "rustc");
writeFileSync(stub, '#!/bin/sh\nprintf "rustc 1.97.1 (8bab26f4f 2026-06-30)\\nbinary: rustc\\ncommit-hash: 8bab26f4f0123456789abcdef\\ncommit-date: 2026-06-30\\nhost: x86_64-unknown-linux-gnu\\nrelease: 1.97.1\\nLLVM version: 21.1.2\\n"\n');
chmodSync(stub, 0o755);
const withStub = execFileSync(process.execPath, [script, "scope"], { encoding: "utf8", env: { ...process.env, PATH: bin, GITHUB_WORKSPACE: "/home/runner/_work/paperclip/paperclip" } }).trim();
assert.equal(withStub, computeScope({ workspace: "/home/runner/_work/paperclip/paperclip", rustcVersion: "1.97.1-8bab26f4f-x86_64-unknown-linux-gnu" }));
const without = execFileSync(process.execPath, [script, "scope"], { encoding: "utf8", env: { ...process.env, PATH: mkdtempSync(path.join(os.tmpdir(), "empty-path-")), GITHUB_WORKSPACE: "/home/runner/_work/paperclip/paperclip" } }).trim();
assert.match(without, /-no-rustc-/);
} finally {
rmSync(bin, { recursive: true, force: true });
}
});
// The tree cargo leaves behind for the crate's release lib and bin units,
// with the mtimes GitHub's cache restores: older than a fresh checkout.
function layRestoredCrate(root) {
const release = path.join(root, RELEASE_DIR);
const lib = path.join(release, ".fingerprint/paperclip-runner-core-0b6c4654c2710743");
const bin = path.join(release, ".fingerprint/paperclip-runner-core-46cce78dbc4840c2");
const table = path.join(release, ".fingerprint/paperclip-process-table-5a1b2c3d4e5f6071");
mkdirSync(path.join(release, "deps"), { recursive: true });
mkdirSync(lib, { recursive: true });
mkdirSync(bin, { recursive: true });
mkdirSync(table, { recursive: true });
const files = {
[path.join(lib, "lib-paperclip_runner_core")]: "0123456789abcdef",
[path.join(lib, "lib-paperclip_runner_core.json")]: "{}",
[path.join(lib, "dep-lib-paperclip_runner_core")]: "dep",
[path.join(lib, "invoked.timestamp")]: "t",
[path.join(bin, "bin-paperclip-runnerd")]: "fedcba9876543210",
[path.join(bin, "bin-paperclip-runnerd.json")]: "{}",
[path.join(bin, "dep-bin-paperclip-runnerd")]: "dep",
[path.join(release, "deps/libpaperclip_runner_core-0b6c4654c2710743.rlib")]: "rlib",
[path.join(release, "deps/libpaperclip_runner_core-0b6c4654c2710743.rmeta")]: "rmeta",
[path.join(release, "deps/paperclip_runner_core-0b6c4654c2710743.d")]: "d",
[path.join(release, "deps/paperclip_runnerd-46cce78dbc4840c2")]: "ELF",
[path.join(release, "deps/paperclip_runnerd-46cce78dbc4840c2.d")]: "d",
[path.join(release, "paperclip-runnerd")]: "ELF",
[path.join(release, "paperclip-runnerd.d")]: "d",
// The process-table workspace library's lib unit.
[path.join(table, "lib-paperclip_process_table")]: "1111222233334444",
[path.join(table, "dep-lib-paperclip_process_table")]: "dep",
[path.join(release, "deps/libpaperclip_process_table-5a1b2c3d4e5f6071.rlib")]: "rlib",
[path.join(release, "deps/libpaperclip_process_table-5a1b2c3d4e5f6071.rmeta")]: "rmeta",
[path.join(release, "deps/paperclip_process_table-5a1b2c3d4e5f6071.d")]: "d",
// A dependency's artifacts sit beside the crate's and must not be touched.
[path.join(release, "deps/libserde-1234567890abcdef.rlib")]: "dep rlib",
};
const old = new Date(Date.now() - 7 * 24 * 60 * 60 * 1000);
for (const [file, content] of Object.entries(files)) {
writeFileSync(file, content);
utimesSync(file, old, old);
}
for (const dir of [lib, bin, table, path.join(release, "deps"), path.join(release, ".fingerprint"), release]) utimesSync(dir, old, old);
return { release, files: Object.keys(files), old };
}
test("mark-current makes every restored artifact newer than the checkout and leaves dependencies alone", () => {
const root = mkdtempSync(path.join(os.tmpdir(), "runner-release-crate-"));
try {
const { release, files, old } = layRestoredCrate(root);
const checkout = path.join(root, RUNNER_PACKAGE, "runner/crates/runner-core/src/lib.rs");
mkdirSync(path.dirname(checkout), { recursive: true });
writeFileSync(checkout, "// checked out after the artifacts were built\n");
const checkedOutAt = statSync(checkout).mtimeMs;
const result = markCurrent({ root, now: new Date(checkedOutAt + 5000) });
assert.equal(result.action, "marked");
assert.equal(result.files, 19);
assert.equal(result.bytes, files.filter((file) => !file.includes("libserde")).reduce((sum, file) => sum + statSync(file).size, 0));
for (const file of files) {
const mtime = statSync(file).mtimeMs;
if (file.includes("libserde")) assert.equal(Math.floor(mtime / 1000), Math.floor(old.getTime() / 1000), "a dependency artifact must keep its mtime");
else assert.ok(mtime > checkedOutAt, `${path.relative(release, file)} must be newer than the checkout`);
}
for (const dir of [".fingerprint/paperclip-runner-core-0b6c4654c2710743", ".fingerprint/paperclip-runner-core-46cce78dbc4840c2"]) {
assert.ok(statSync(path.join(release, dir)).mtimeMs > checkedOutAt, `${dir} must be newer than the checkout`);
}
assert.ok(statSync(checkout).mtimeMs === checkedOutAt, "sources are never touched");
// The CLI prints the same result as one line of JSON for the job log.
const printed = JSON.parse(execFileSync(process.execPath, [script, "mark-current"], { encoding: "utf8", env: { ...process.env, GITHUB_WORKSPACE: root } }));
assert.equal(printed.schema, "paperclip.runner-release-crate.v1");
assert.equal(printed.action, "marked");
assert.equal(printed.files, 19);
} finally {
rmSync(root, { recursive: true, force: true });
}
});
// One file per REQUIRED entry, as cargo 1.97.1 lays them out for the crate's
// release lib and bin units. Measured: without any one of these cargo
// recompiles the unit (the whole crate for a lib file, a relink for a bin
// file); without any of the files in NOT_NEEDED it reports the crate fresh
// and recreates the uplifted binary from the one under deps.
const REQUIRED_FILES = {
"the lib fingerprint hash": ".fingerprint/paperclip-runner-core-0b6c4654c2710743/lib-paperclip_runner_core",
"the lib dep-info": ".fingerprint/paperclip-runner-core-0b6c4654c2710743/dep-lib-paperclip_runner_core",
"the lib rlib": "deps/libpaperclip_runner_core-0b6c4654c2710743.rlib",
"the lib rmeta": "deps/libpaperclip_runner_core-0b6c4654c2710743.rmeta",
"the bin fingerprint hash": ".fingerprint/paperclip-runner-core-46cce78dbc4840c2/bin-paperclip-runnerd",
"the bin dep-info": ".fingerprint/paperclip-runner-core-46cce78dbc4840c2/dep-bin-paperclip-runnerd",
"the bin under deps": "deps/paperclip_runnerd-46cce78dbc4840c2",
"the process-table lib fingerprint hash": ".fingerprint/paperclip-process-table-5a1b2c3d4e5f6071/lib-paperclip_process_table",
"the process-table lib dep-info": ".fingerprint/paperclip-process-table-5a1b2c3d4e5f6071/dep-lib-paperclip_process_table",
"the process-table lib rlib": "deps/libpaperclip_process_table-5a1b2c3d4e5f6071.rlib",
"the process-table lib rmeta": "deps/libpaperclip_process_table-5a1b2c3d4e5f6071.rmeta",
};
const NOT_NEEDED = [
".fingerprint/paperclip-runner-core-0b6c4654c2710743/lib-paperclip_runner_core.json",
".fingerprint/paperclip-runner-core-0b6c4654c2710743/invoked.timestamp",
".fingerprint/paperclip-runner-core-46cce78dbc4840c2/bin-paperclip-runnerd.json",
"deps/paperclip_runner_core-0b6c4654c2710743.d",
"deps/paperclip_runnerd-46cce78dbc4840c2.d",
"paperclip-runnerd",
"paperclip-runnerd.d",
"deps/paperclip_process_table-5a1b2c3d4e5f6071.d",
];
test("mark-current discards a restore that lacks any file cargo needs for either unit, naming the file, so cargo sees a plain miss", () => {
assert.deepEqual(REQUIRED.map(({ label }) => label), Object.keys(REQUIRED_FILES), "every REQUIRED entry has a missing-file case");
for (const [label, file] of Object.entries(REQUIRED_FILES)) {
const root = mkdtempSync(path.join(os.tmpdir(), "runner-release-crate-"));
try {
const { release } = layRestoredCrate(root);
rmSync(path.join(release, file));
const result = markCurrent({ root });
assert.equal(result.action, "discarded", `${file} missing must discard the restore`);
assert.equal(result.reason, `incomplete: ${label} is missing`);
assert.ok(result.removed > 0, label);
assert.deepEqual(readdirSync(path.join(release, ".fingerprint")), [], label);
assert.deepEqual(readdirSync(path.join(release, "deps")), ["libserde-1234567890abcdef.rlib"], `${label}: only the crate's own artifacts are removed`);
assert.deepEqual(readdirSync(release).sort(), [".fingerprint", "deps"], label);
} finally {
rmSync(root, { recursive: true, force: true });
}
}
// The bin's dep-info alone does not stand in for the bin: the glob that
// finds the binary under deps must not be satisfied by the `.d` beside it.
const root = mkdtempSync(path.join(os.tmpdir(), "runner-release-crate-"));
try {
const { release } = layRestoredCrate(root);
rmSync(path.join(release, "deps/paperclip_runnerd-46cce78dbc4840c2"));
assert.ok(statSync(path.join(release, "deps/paperclip_runnerd-46cce78dbc4840c2.d")).isFile());
assert.equal(markCurrent({ root }).reason, "incomplete: the bin under deps is missing");
} finally {
rmSync(root, { recursive: true, force: true });
}
// A file cargo neither reads nor checks for may be absent without costing
// a rebuild, so its absence must not cost one here either.
for (const file of NOT_NEEDED) {
const intact = mkdtempSync(path.join(os.tmpdir(), "runner-release-crate-"));
try {
const { release } = layRestoredCrate(intact);
rmSync(path.join(release, file));
const result = markCurrent({ root: intact });
assert.equal(result.action, "marked", `${file} missing must not discard the restore`);
assert.equal(result.files, 18, file);
} finally {
rmSync(intact, { recursive: true, force: true });
}
}
// Nothing restored at all is reported the same way and exits cleanly.
const empty = mkdtempSync(path.join(os.tmpdir(), "runner-release-crate-empty-"));
try {
const printed = JSON.parse(execFileSync(process.execPath, [script, "mark-current"], { encoding: "utf8", env: { ...process.env, GITHUB_WORKSPACE: empty } }));
assert.equal(printed.action, "discarded");
assert.equal(printed.reason, "incomplete: the lib fingerprint hash is missing");
assert.equal(printed.removed, 0);
} finally {
rmSync(empty, { recursive: true, force: true });
}
});
test("the CLI prints the lists the workflows embed", () => {
const paths = execFileSync(process.execPath, [script, "paths"], { encoding: "utf8" }).trim().split("\n");
assert.deepEqual(paths, CACHED_PATHS);
const inputs = execFileSync(process.execPath, [script, "hashed-inputs"], { encoding: "utf8" }).trim().split("\n");
assert.deepEqual(inputs, HASHED_INPUTS);
const usage = spawnSync(process.execPath, [script, "nonsense"], { encoding: "utf8" });
assert.equal(usage.status, 2);
assert.match(usage.stderr, /usage: runner-release-crate-cache\.mjs/);
});