mirror of
https://github.com/paperclipai/paperclip.git
synced 2026-11-16 03:31:44 +01:00
## Thinking Path > - Paperclip is the open source app people use to manage AI agents for work > - The Runner is the part that executes agent work: a daemon (`paperclip-runnerd`) and a provider pack of agent tools > - Parts 2 and 3 of this stack build the Runner alone as an image with two variants, and check it on pull requests. Nothing publishes it > - Downstream sandbox and VM images can copy the Runner only from a registry, and they must be able to verify where it came from > - A published image must be exactly the bytes that were checked, and its name must never point at something else later > - This pull request publishes both variants from `master` as `ghcr.io/paperclipai/paperclip-runner`, one time for each distinct set of inputs, and attests them > - Nothing consumes the image in this pull request. The application image build does not change > - The benefit is one verified, attested Runner that later changes can copy by digest ## Linked Issues or Issue Description Refs #21204 (part 2 of this stack) and #21096 (the merge-queue image build). No issue exists for this change. **Problem or motivation** The Runner image exists only inside a pull-request check. A consumer cannot pull it. There is no rule for which commit publishes an image, no protection for its tags, and no way to verify a copied Runner. **Proposed solution** Publish from a `master` push in this repository, and from nothing else. Build only when no image has the digest of the commit's inputs. Let the build legs hold no write access. Push the bytes that the legs checked by digest, attest each multi-platform index, and write the tags last. Write the full tag after the slim tag, so that the full tag means "both exist". **Alternatives considered** - Publish from `docker.yml`. That workflow holds one concurrency slot for each ref, and the Runner image needs its own signer identity, so that its attestation cannot stand in for the application image's. - Resolve the lockfile in the workflow before a `master` build, as `docker.yml` does. Then the published Runner depends on the state of the npm registry at run time, and the digest cannot be computed again from a commit. - Hand each variant to the publish job as its own OCI archive. The two archives hold the two shared layers two times: 1.27 GB for each architecture in place of 0.85 GB. - Write the full tag first. Then a consumer can see a full image whose slim image does not exist yet. **Roadmap alignment** This is build infrastructure. It adds no product feature and does not overlap an item in `ROADMAP.md`. This is part 4 of 4. Each part is on top of the one before it: 1. #21228 (merged): an optional field in the provider pack manifest 2. #21204: build and check the Runner image, with no publication 3. #21229: layers by change rate, and the slim variant 4. #21230: publish both variants from `master` (this pull request) ## What Changed - `.github/workflows/runner-image.yml`: adds the trigger `push` to `master` and the job `publish`. - `identity` checks the committed lockfile against the Runner's manifests, computes the digest in the build context, and asks the registry whether these inputs are published. - `build` is the job of part 3. It also records what it built, on every event. On a canonical `master` push it uploads that. It cannot write to the registry on any event. The only token in the job can read packages, and only the lookup step receives it. - `publish` runs only for a `master` push in this repository. It builds nothing and runs no dependency code. - Lockfile rule. A `master` push builds from the committed `pnpm-lock.yaml` and resolves nothing, so the digest is a function of the commit. When the committed lockfile is behind the manifests that the Runner build reads, the run writes a notice, publishes nothing, and stays green. The commit that merges the lockfile refresh publishes the image. Pull requests and manual runs resolve the lockfile first, always build, and never publish. - Hand-off. Each leg joins its two OCI archives into one OCI layout that names the images `slim` and `full`. The layout stores the two shared layers one time. `record` then reads the layout back, each image by its name, and checks every blob against its digest. The leg's record binds the layout to both image digests, the architecture, the inputs digest, and the commit. Those digests are the digests of the archives that part 3 unpacks and checks, so the tree that is checked and the image that is pushed are the same bytes. The publish job accepts only records for its own inputs and commit, and only a layout that still has the recorded bytes. - `record` runs on pull requests too. Only the upload is for a `master` push. A pull request therefore runs the hand-off on real build output. - Publication order. The publish job pushes the four images by digest (`skopeo copy --preserve-digests`). It makes a candidate index for the slim variant, then one for the full variant that names the slim index in the annotation `io.paperclip.runner.slim-digest`. It verifies that each index names exactly the recorded images, attests both, and then writes `inputs-<digest>-slim` and, last, `inputs-<digest>`. - The full tag is the completion marker. Every "is it published" lookup reads the full tag. `publication-state` reports one of: | State | Registry | What a run does | | --- | --- | --- | | `absent` | Neither tag | Builds and publishes | | `incomplete` | The slim tag alone | Builds and publishes. It replaces the slim tag with its own image, then writes the full tag | | `published` | The full tag, and the slim tag at the digest that the full index names | Nothing | | broken | The full tag, and a slim tag that is missing or points at another image | Fails, with the steps for an operator | | `unknown` | No clear answer | Builds. The publish job fails before it writes a final tag | - Recovery rule. A run that stops between the two tags leaves the state `incomplete`. It does not block those inputs. The next run never keeps the earlier slim image: it writes its own slim tag immediately before its full tag, verifies both against its own index digests, and then reads the pair back as every later run will see it. The broken state cannot be repaired by a run, because no run moves a full tag. An operator must delete the two package versions and re-run the workflow run of a `master` push with those inputs. - Pairing rule for consumers, in the contract: resolve the full tag first, take the slim digest from the annotation of the full index, and pull the slim image by that digest. - `verify-index` accepts `linux/arm64` with the variant `v8`, which some BuildKit versions record, and no other variant. - `scripts/runner-image-publish.mjs` (533 lines): the commands `lock-state`, `record`, `verify-legs`, `publication-state` and `verify-index`. It reads archives with the reader of `scripts/runner-image-oci.mjs` (607 lines here), which can now read one named image of a layout. `lock-state` refuses a checkout, because pnpm can rewrite the lockfile of the tree that it checks. The workflow gives it an export that no build reads. - The workflow test pins the build arguments of the image build. Build arguments are not part of the image identity. - `doc/runner-image-contract.md` (197 lines): the section "Publication": the lockfile rule, the order, the pairing rule, the recovery rule, the create-only tags, the internal `candidate-` tags, the command that verifies provenance, and the step to make the package public. `doc/DOCKER.md` says where the image is published. ## Verification Tests, and the build jobs of the workflow on GitHub. Nothing in this pull request ran against a registry, and no publish job has run. - Linux (the pinned `node:24.22.3-bookworm` base with GNU tar 1.34 and zstd): `runner-image-publish.test.ts` (75), `runner-image-workflow.test.ts` (12), the three other Runner image test files, the provider pack archive tests and the existing Dockerfile tests: 372 tests pass in 10 files. The same files pass on macOS. - The tests of the leg record failed on Linux in the first runs of this pull request: their fixtures named the archive members `./index.json`, which bsdtar finds and GNU tar does not. The fixtures now name the members as buildx does, and the publish script uses the one reader of part 3. - Leg records: both variants are accepted for this run's inputs and commit; other inputs, another commit, a changed layout, a record without a variant, an archive that lacks a blob, a full image that does not start with the slim image's layers, and a build that is not a publication build are refused. - Publication state: all five states, and the half-finished publication end to end. A slim tag alone lets the next run publish. A full image that is left with another run's slim tag is reported as broken. - Index: exactly the recorded images of one variant; the full index must name its slim index, and the slim index must name none; `v8` is accepted on arm64 only. - Registry lookups: a missing tag only when the registry says so; every refused, failed, or unreachable request is `unknown`. - Workflow: job permissions; the publish job and the upload are behind the canonical `master` push; the build steps receive no credential; the full tag is written last, after a strict lookup and both attestations; one export and a time limit for each build command; the two build arguments; actions pinned by commit; no expression inside a shell script. - `record` on the two OCI archives of a real `linux/arm64` build from part 3: the joined layout is 850,149,376 bytes, and both images read back by name with every blob under its digest. The two archives together are 1,273,646,592 bytes. The image digests in the record are the digests that `unpack` reports for the trees that were checked. - `tests/runner-e2e` workflow tests: 39 pass. `node --test .github/scripts/tests/*.test.mjs`: 799 pass. `node --test` on the image contract scripts: 33 pass. `docker buildx build --check` reports nothing. The image identity is the same as in part 3 (`a544fc9dde780963419ede0f34252f3a57de43a760d41d3eb8fcd4f577504f6a`), because this pull request changes no input of the image. GitHub, at this head: - The `Runner image` workflow passes with these changes: `Build and check (amd64)` in 18m30s and `Build and check (arm64)` in 7m48s. Each leg builds both variants, checks them, and writes the leg record. `Publish and attest` is skipped, as it must be on a pull request. Never run: - The publish job. It runs only for a `master` push. - The publish path: the registry lookups with a workflow token, `skopeo copy` from `oci-archive:<layout>:<name>` by digest, `docker buildx imagetools create` with index annotations, and the tag writes. - The attestations, and the `gh attestation verify` command in the document. - `actionlint`. It is not installed on the local host. - The full test suite and the full typecheck. CI must supply them. ## Risks - **Manual step.** GitHub creates the `paperclip-runner` package private on the first publication. A maintainer must make it public one time. Until then, consumers cannot pull the image without credentials. - The first `master` runs of the workflow are the first test of the publish path. A failure before the tags publishes nothing. A failure between the two tags leaves the `incomplete` state, and the next run with those inputs repairs it. - A slim tag can move while its publication is incomplete. A consumer must resolve the full tag first and take the slim digest from the annotation of the full index. It must never resolve `inputs-<digest>-slim` by tag on its own. The contract says so. - `candidate-` tags are written before the attestations and can name an index that is not attested. They are internal. - The broken state needs an operator, and it fails every `master` run with those inputs until then. It can arise only when someone changes or deletes a tag outside the workflow. - **Known residual.** Two `master` runs that both start before a new digest is published can each build the image. Each build leg and the publish job look again before they work, and the publish job runs one at a time for each digest. The cost is runner time, not correctness. - A commit whose lockfile is behind the Runner's manifests has no published Runner image until the lockfile refresh merges. A lockfile that the refresh cannot fix leaves every `master` run green with only a notice, for as long as it lasts. - The tags are create-only because the workflow never moves a tag of a complete publication. The registry does not enforce that. Consumers must pin the digest and verify the attestation. - `payload.runnerSourceRevision` in a pack from this image is the commit that published the image. Code that reads it as the application's commit must change before it consumes the image. > For core feature work, check [`ROADMAP.md`](ROADMAP.md) first and discuss it in `#dev` before opening the PR. Feature PRs that overlap with planned core work may need to be redirected — check the roadmap first. See `CONTRIBUTING.md`. ## Model Used - Claude Fable 5.2 (`claude-fable-5-2`), Anthropic, used through Claude Code with tool use (shell, file edits, local Docker builds). ## 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 - [x] All Paperclip CI gates are green - [x] 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 Fable 5.2 <noreply@anthropic.com>
608 lines
28 KiB
JavaScript
608 lines
28 KiB
JavaScript
#!/usr/bin/env node
|
|
// OCI archives of the Runner image, as `docker buildx build --output type=oci`
|
|
// writes them for one platform. doc/runner-image-contract.md is the contract.
|
|
//
|
|
// unpack writes the file tree of the image in an archive: its layers,
|
|
// in order. The layout check and the smoke matrix then read
|
|
// the bytes of the archive, not a second export of the build,
|
|
// so what they accept is what a registry would receive.
|
|
// check-layers verifies the layers of the two variants: they are split by
|
|
// how often their content changes, the full image is the slim
|
|
// image plus the transport layers, and nothing else is in
|
|
// those.
|
|
//
|
|
// Usage:
|
|
// node scripts/runner-image-oci.mjs unpack IMAGE.oci.tar --out DIR
|
|
// node scripts/runner-image-oci.mjs check-layers [--slim SLIM.oci.tar] [--full FULL.oci.tar]
|
|
//
|
|
// The archive and its layers are read here, not with the system's tar, so
|
|
// every platform reads them the same way. Each blob is checked against its
|
|
// digest as it is read. The image is `FROM scratch` and every layer only adds
|
|
// files, so `unpack` refuses a whiteout, a path that is absolute or has `..`,
|
|
// a path written twice, an entry behind a link, and every entry that is not a
|
|
// file, a directory or a link.
|
|
//
|
|
// The image build does not run this file, so it is not an input of the image.
|
|
|
|
import { createHash } from "node:crypto";
|
|
import {
|
|
chmodSync,
|
|
closeSync,
|
|
createReadStream,
|
|
linkSync,
|
|
lstatSync,
|
|
mkdirSync,
|
|
openSync,
|
|
readSync,
|
|
readdirSync,
|
|
symlinkSync,
|
|
utimesSync,
|
|
writeSync,
|
|
} from "node:fs";
|
|
import path from "node:path";
|
|
import { Transform } from "node:stream";
|
|
import { pipeline } from "node:stream/promises";
|
|
import { pathToFileURL } from "node:url";
|
|
import { createGunzip } from "node:zlib";
|
|
import { RUNNER_IMAGE_PATHS, canonicalJson, parseLayoutArguments, summarize } from "./runner-image-layout.mjs";
|
|
|
|
const BLOCK = 512;
|
|
const DIGEST = /^sha256:[0-9a-f]{64}$/;
|
|
const GZIP_LAYER = /^application\/vnd\.(?:oci\.image\.layer\.v1\.tar\+gzip|docker\.image\.rootfs\.diff\.tar\.gzip)$/;
|
|
const ENTRY_KINDS = Object.freeze({ 0: "file", "\0": "file", 1: "hardlink", 2: "link", 5: "directory" });
|
|
|
|
/** One tar header block, or null for the zero block that ends an archive. */
|
|
function parseTarHeader(block) {
|
|
if (block.every((byte) => byte === 0)) return null;
|
|
const text = (start, length) => {
|
|
const field = block.subarray(start, start + length);
|
|
const end = field.indexOf(0);
|
|
return field.subarray(0, end === -1 ? length : end).toString("utf8");
|
|
};
|
|
const octal = (start, length, label) => {
|
|
const value = text(start, length).trim();
|
|
if (!/^[0-7]*$/.test(value)) throw new Error(`tar header has a ${label} this reader does not support`);
|
|
return value === "" ? 0 : Number.parseInt(value, 8);
|
|
};
|
|
let sum = 0;
|
|
for (let index = 0; index < BLOCK; index += 1) sum += index >= 148 && index < 156 ? 32 : block[index];
|
|
if (sum !== octal(148, 8, "checksum")) throw new Error("tar header checksum does not match");
|
|
// Only the POSIX format has a name prefix; GNU tar keeps other fields there.
|
|
const prefix = text(257, 6) === "ustar" && block[263] === 48 ? text(345, 155) : "";
|
|
return {
|
|
name: prefix ? `${prefix}/${text(0, 100)}` : text(0, 100),
|
|
mode: octal(100, 8, "mode"),
|
|
size: octal(124, 12, "size"),
|
|
mtime: octal(136, 12, "time"),
|
|
type: String.fromCharCode(block[156]),
|
|
linkname: text(157, 100),
|
|
};
|
|
}
|
|
|
|
/** The records of a pax extended header: `<length> <key>=<value>\n`. */
|
|
function parsePaxRecords(data) {
|
|
/** @type {Record<string, string>} */
|
|
const records = {};
|
|
for (let offset = 0; offset < data.length;) {
|
|
const space = data.indexOf(32, offset);
|
|
const length = Number.parseInt(data.subarray(offset, space).toString("latin1"), 10);
|
|
if (space === -1 || !Number.isSafeInteger(length) || length <= 0 || offset + length > data.length) throw new Error("tar pax header is malformed");
|
|
const record = data.subarray(space + 1, offset + length - 1).toString("utf8");
|
|
const equals = record.indexOf("=");
|
|
if (equals === -1) throw new Error("tar pax header is malformed");
|
|
records[record.slice(0, equals)] = record.slice(equals + 1);
|
|
offset += length;
|
|
}
|
|
return records;
|
|
}
|
|
|
|
/** Header types that only describe the next entry: pax (`x`, and `g` for the whole archive) and GNU long names. */
|
|
const DESCRIBING = new Set(["x", "g", "L", "K"]);
|
|
|
|
/** Applies describing headers and returns the entry each other header stands for. */
|
|
function tarEntryReader() {
|
|
/** @type {Record<string, string>} */
|
|
let pax = {};
|
|
/** @type {{ name?: string, linkname?: string }} */
|
|
let gnu = {};
|
|
const cString = (data) => data.subarray(0, data.indexOf(0) === -1 ? data.length : data.indexOf(0)).toString("utf8");
|
|
return {
|
|
describe(header, data) {
|
|
if (header.type === "x") pax = parsePaxRecords(data);
|
|
else if (header.type === "L") gnu.name = cString(data);
|
|
else if (header.type === "K") gnu.linkname = cString(data);
|
|
},
|
|
entry(header) {
|
|
const entry = {
|
|
name: pax.path ?? gnu.name ?? header.name,
|
|
linkname: pax.linkpath ?? gnu.linkname ?? header.linkname,
|
|
size: pax.size === undefined ? header.size : Number(pax.size),
|
|
mtime: pax.mtime === undefined ? header.mtime : Number(pax.mtime),
|
|
mode: header.mode & 0o7777,
|
|
type: header.type,
|
|
};
|
|
pax = {};
|
|
gnu = {};
|
|
if (!Number.isSafeInteger(entry.size) || entry.size < 0) throw new Error(`tar entry ${entry.name} has no usable size`);
|
|
return entry;
|
|
},
|
|
};
|
|
}
|
|
|
|
/** Where each member of an uncompressed tar file is, by its exact name. */
|
|
function indexTarFile(file) {
|
|
const fd = openSync(file, "r");
|
|
try {
|
|
/** @type {Map<string, { offset: number, size: number }>} */
|
|
const members = new Map();
|
|
const reader = tarEntryReader();
|
|
const block = Buffer.alloc(BLOCK);
|
|
for (let offset = 0; readSync(fd, block, 0, BLOCK, offset) === BLOCK;) {
|
|
const header = parseTarHeader(block);
|
|
if (header === null) break;
|
|
offset += BLOCK;
|
|
if (DESCRIBING.has(header.type)) {
|
|
const data = Buffer.alloc(header.size);
|
|
if (readSync(fd, data, 0, header.size, offset) !== header.size) throw new Error(`${file} ends inside a header`);
|
|
reader.describe(header, data);
|
|
offset += Math.ceil(header.size / BLOCK) * BLOCK;
|
|
continue;
|
|
}
|
|
const entry = reader.entry(header);
|
|
if (members.has(entry.name)) throw new Error(`${file} holds ${entry.name} twice`);
|
|
members.set(entry.name, { offset, size: entry.size });
|
|
offset += Math.ceil(entry.size / BLOCK) * BLOCK;
|
|
}
|
|
return members;
|
|
} finally {
|
|
closeSync(fd);
|
|
}
|
|
}
|
|
|
|
/**
|
|
* Calls `visit(entry, content)` for every entry of a tar stream. `content`
|
|
* yields the entry's bytes; what `visit` leaves unread is skipped.
|
|
* @param {AsyncIterable<Buffer>} source
|
|
* @param {(entry: ReturnType<ReturnType<typeof tarEntryReader>["entry"]>, content: () => AsyncGenerator<Buffer>) => Promise<void> | void} visit
|
|
*/
|
|
async function readTarStream(source, visit) {
|
|
const iterator = source[Symbol.asyncIterator]();
|
|
let buffered = Buffer.alloc(0);
|
|
/** Some bytes, at most `limit`: what is buffered, or the next chunk. Empty at the end of the stream. */
|
|
const some = async (limit) => {
|
|
while (buffered.length === 0) {
|
|
const next = await iterator.next();
|
|
if (next.done) return buffered;
|
|
buffered = next.value;
|
|
}
|
|
const taken = buffered.subarray(0, limit);
|
|
buffered = buffered.subarray(taken.length);
|
|
return taken;
|
|
};
|
|
/** Exactly `length` bytes, or fewer at the end of the stream. */
|
|
const take = async (length) => {
|
|
const parts = [];
|
|
for (let missing = length; missing > 0;) {
|
|
const part = await some(missing);
|
|
if (part.length === 0) break;
|
|
parts.push(part);
|
|
missing -= part.length;
|
|
}
|
|
return parts.length === 1 ? parts[0] : Buffer.concat(parts);
|
|
};
|
|
const exactly = async (length) => {
|
|
const taken = await take(length);
|
|
if (taken.length !== length) throw new Error("tar stream ends inside an entry");
|
|
return taken;
|
|
};
|
|
const reader = tarEntryReader();
|
|
for (;;) {
|
|
const block = await take(BLOCK);
|
|
if (block.length === 0) break;
|
|
if (block.length !== BLOCK) throw new Error("tar stream ends inside a header");
|
|
const header = parseTarHeader(block);
|
|
if (header === null) break;
|
|
if (DESCRIBING.has(header.type)) {
|
|
reader.describe(header, (await exactly(Math.ceil(header.size / BLOCK) * BLOCK)).subarray(0, header.size));
|
|
continue;
|
|
}
|
|
const entry = reader.entry(header);
|
|
let remaining = entry.size;
|
|
const content = async function* () {
|
|
while (remaining > 0) {
|
|
const chunk = await some(remaining);
|
|
if (chunk.length === 0) throw new Error("tar stream ends inside an entry");
|
|
remaining -= chunk.length;
|
|
yield chunk;
|
|
}
|
|
};
|
|
await visit(entry, content);
|
|
for await (const _unread of content()) { /* skip what the visitor left */ }
|
|
await exactly((BLOCK - (entry.size % BLOCK)) % BLOCK);
|
|
}
|
|
// Read to the end, so the caller's digest covers the whole blob.
|
|
while (!(await iterator.next()).done) { /* trailing zero blocks */ }
|
|
}
|
|
|
|
/**
|
|
* One image in an OCI archive: its manifest descriptor and its layers. A
|
|
* member is found by its exact name (`index.json`, `blobs/sha256/<hex>`), as
|
|
* buildx writes them. The manifest is checked against its digest here, and
|
|
* each layer when `readLayer` or `verifyBlobs` reads it.
|
|
*
|
|
* An archive from buildx holds one image. A layout that holds several names
|
|
* each with the annotation `org.opencontainers.image.ref.name`; `ref` then
|
|
* says which one to read.
|
|
* @param {string} archive
|
|
* @param {{ ref?: string | null }} [options]
|
|
*/
|
|
export function readOciArchive(archive, { ref = null } = {}) {
|
|
const members = indexTarFile(archive);
|
|
const located = (name) => {
|
|
const member = members.get(name);
|
|
if (!member) throw new Error(`${archive} has no member ${name}`);
|
|
return member;
|
|
};
|
|
const bytes = (name) => {
|
|
const { offset, size } = located(name);
|
|
const data = Buffer.alloc(size);
|
|
const fd = openSync(archive, "r");
|
|
try {
|
|
if (readSync(fd, data, 0, size, offset) !== size) throw new Error(`${archive} ends inside ${name}`);
|
|
} finally {
|
|
closeSync(fd);
|
|
}
|
|
return data;
|
|
};
|
|
const blobName = (digest) => {
|
|
if (!DIGEST.test(digest ?? "")) throw new Error(`${archive} names a blob by something other than a sha256 digest`);
|
|
return `blobs/sha256/${digest.slice(7)}`;
|
|
};
|
|
const sha256 = (data) => `sha256:${createHash("sha256").update(data).digest("hex")}`;
|
|
const index = JSON.parse(bytes("index.json").toString("utf8"));
|
|
if (!Array.isArray(index.manifests)) throw new Error(`${archive} has no list of manifests`);
|
|
const named = ref === null ? index.manifests : index.manifests.filter((entry) => entry.annotations?.["org.opencontainers.image.ref.name"] === ref);
|
|
if (named.length !== 1) {
|
|
throw new Error(ref === null ? `${archive} must hold exactly one image manifest` : `${archive} must hold exactly one image named ${ref}`);
|
|
}
|
|
const { mediaType, digest, size } = named[0];
|
|
const manifestBytes = bytes(blobName(digest));
|
|
if (sha256(manifestBytes) !== digest) throw new Error(`${archive}: the manifest does not have the digest the index names`);
|
|
const manifest = JSON.parse(manifestBytes.toString("utf8"));
|
|
if (!Array.isArray(manifest.layers)) throw new Error(`${archive} holds an index, not one platform's image; build without attestations`);
|
|
/** The digest of a blob's bytes as the archive holds them. */
|
|
const blobDigest = async (name) => {
|
|
const { offset, size: length } = located(name);
|
|
const hash = createHash("sha256");
|
|
if (length > 0) for await (const chunk of createReadStream(archive, { start: offset, end: offset + length - 1 })) hash.update(chunk);
|
|
return `sha256:${hash.digest("hex")}`;
|
|
};
|
|
for (const layer of manifest.layers) {
|
|
if (!GZIP_LAYER.test(layer.mediaType ?? "")) throw new Error(`${archive}: layer ${layer.digest} is ${layer.mediaType}, not a gzip tar`);
|
|
located(blobName(layer.digest));
|
|
}
|
|
return {
|
|
archive,
|
|
descriptor: { mediaType, digest, size },
|
|
/** @type {{ mediaType: string, digest: string, size: number }[]} */
|
|
layers: manifest.layers.map((layer) => ({ mediaType: layer.mediaType, digest: layer.digest, size: layer.size })),
|
|
/** Fails unless the config and every layer of the image are in the archive with their digests. */
|
|
async verifyBlobs() {
|
|
for (const blob of [manifest.config, ...manifest.layers]) {
|
|
if (await blobDigest(blobName(blob?.digest)) !== blob.digest) throw new Error(`${archive}: blob ${blob.digest} does not have its digest`);
|
|
}
|
|
},
|
|
/** Visits every entry of one layer and fails when the blob does not have its digest. */
|
|
async readLayer(layer, visit) {
|
|
const { offset, size: bytesInArchive } = located(blobName(layer.digest));
|
|
if (bytesInArchive !== layer.size || bytesInArchive === 0) {
|
|
throw new Error(`${archive}: layer ${layer.digest} is ${bytesInArchive} bytes, the manifest says ${layer.size}`);
|
|
}
|
|
const hash = createHash("sha256");
|
|
const hashing = new Transform({ transform(chunk, _encoding, done) { hash.update(chunk); done(null, chunk); } });
|
|
const gunzip = createGunzip({ chunkSize: 1024 * 1024 });
|
|
const feeding = pipeline(createReadStream(archive, { start: offset, end: offset + bytesInArchive - 1 }), hashing, gunzip);
|
|
// A failure below destroys the streams; that rejection is not the cause.
|
|
feeding.catch(() => {});
|
|
try {
|
|
await readTarStream(gunzip, visit);
|
|
} catch (error) {
|
|
gunzip.destroy();
|
|
throw error;
|
|
}
|
|
await feeding;
|
|
if (`sha256:${hash.digest("hex")}` !== layer.digest) throw new Error(`${archive}: layer ${layer.digest} does not have its digest`);
|
|
},
|
|
};
|
|
}
|
|
|
|
/**
|
|
* The path of a layer entry relative to the image root, and its kind. Throws
|
|
* for an entry an additive layer of a `FROM scratch` image cannot have.
|
|
* Returns null for the root directory itself.
|
|
*/
|
|
function layerEntry(entry) {
|
|
const kind = ENTRY_KINDS[entry.type];
|
|
if (!kind) throw new Error(`layer entry ${entry.name} is of type ${JSON.stringify(entry.type)}, not a file, a directory or a link`);
|
|
const clean = (name, label) => {
|
|
if (name.startsWith("/")) throw new Error(`layer ${label} ${name} is absolute`);
|
|
const relative = name.replace(/^(?:\.\/)+/, "").replace(/\/+$/, "");
|
|
if (relative === "" || relative === ".") return null;
|
|
const parts = relative.split("/");
|
|
if (relative.includes("\0") || parts.some((part) => part === "" || part === "." || part === "..")) {
|
|
throw new Error(`layer ${label} ${name} is not a plain relative path`);
|
|
}
|
|
if (parts.some((part) => part.startsWith(".wh."))) throw new Error(`layer ${label} ${name} is a whiteout: the image only adds files`);
|
|
return relative;
|
|
};
|
|
const name = clean(entry.name, "entry");
|
|
if (name === null) {
|
|
if (kind !== "directory") throw new Error(`layer entry ${entry.name} replaces the image root`);
|
|
return null;
|
|
}
|
|
// A link's target is text and is never followed here. A hard link names a file of the image.
|
|
const target = kind === "hardlink" ? clean(entry.linkname, "hard link target") : kind === "link" ? entry.linkname : null;
|
|
if (kind === "hardlink" && target === null) throw new Error(`layer entry ${entry.name} is a hard link to the image root`);
|
|
return { name, kind, target, mode: entry.mode, size: entry.size, mtime: entry.mtime };
|
|
}
|
|
|
|
/**
|
|
* The entries of every layer of the image in an archive, in order.
|
|
* @param {string} archive
|
|
*/
|
|
export async function listOciLayers(archive) {
|
|
const image = readOciArchive(archive);
|
|
const layers = [];
|
|
for (const layer of image.layers) {
|
|
/** @type {NonNullable<ReturnType<typeof layerEntry>>[]} */
|
|
const entries = [];
|
|
await image.readLayer(layer, (raw) => {
|
|
const entry = layerEntry(raw);
|
|
if (entry) entries.push(entry);
|
|
});
|
|
layers.push({ digest: layer.digest, bytes: layer.size, entries });
|
|
}
|
|
return { digest: image.descriptor.digest, layers };
|
|
}
|
|
|
|
/**
|
|
* Writes the file tree of the image in an archive to `destination`, which
|
|
* must be empty or absent: every layer in order. Returns the image digest.
|
|
* @param {string} archive
|
|
* @param {string} destination
|
|
*/
|
|
export async function unpackOciImage(archive, destination) {
|
|
const image = readOciArchive(archive);
|
|
mkdirSync(destination, { recursive: true });
|
|
if (readdirSync(destination).length > 0) throw new Error(`${destination} is not empty`);
|
|
const stat = (file) => {
|
|
try {
|
|
return lstatSync(file);
|
|
} catch (error) {
|
|
if (/** @type {NodeJS.ErrnoException} */ (error).code === "ENOENT") return null;
|
|
throw error;
|
|
}
|
|
};
|
|
/** The place of an entry, after its parents are known to be real directories. */
|
|
const place = (name) => {
|
|
let current = destination;
|
|
const parts = name.split("/");
|
|
for (const part of parts.slice(0, -1)) {
|
|
current = path.join(current, part);
|
|
const parent = stat(current);
|
|
if (parent === null) mkdirSync(current, { mode: 0o755 });
|
|
else if (!parent.isDirectory()) throw new Error(`layer entry ${name} is behind ${path.relative(destination, current)}, which is not a directory`);
|
|
}
|
|
return path.join(current, parts.at(-1));
|
|
};
|
|
/** @type {Map<string, number>} */
|
|
const directoryModes = new Map();
|
|
let files = 0;
|
|
for (const layer of image.layers) {
|
|
await image.readLayer(layer, async (raw, content) => {
|
|
const entry = layerEntry(raw);
|
|
if (entry === null) return;
|
|
const target = place(entry.name);
|
|
const existing = stat(target);
|
|
if (entry.kind === "directory") {
|
|
if (existing === null) mkdirSync(target, { mode: 0o755 });
|
|
else if (!existing.isDirectory()) throw new Error(`a layer writes ${entry.name} again, as a directory`);
|
|
directoryModes.set(target, entry.mode);
|
|
return;
|
|
}
|
|
if (existing !== null) throw new Error(`a layer writes ${entry.name} again: the image only adds files`);
|
|
files += 1;
|
|
if (entry.kind === "link") {
|
|
symlinkSync(/** @type {string} */ (entry.target), target);
|
|
} else if (entry.kind === "hardlink") {
|
|
const source = place(/** @type {string} */ (entry.target));
|
|
if (!stat(source)?.isFile()) throw new Error(`layer entry ${entry.name} is a hard link to ${entry.target}, which is not a file of the image`);
|
|
linkSync(source, target);
|
|
} else {
|
|
const fd = openSync(target, "wx", 0o600);
|
|
try {
|
|
for await (const chunk of content()) writeSync(fd, chunk);
|
|
} finally {
|
|
closeSync(fd);
|
|
}
|
|
chmodSync(target, entry.mode);
|
|
utimesSync(target, entry.mtime, entry.mtime);
|
|
}
|
|
});
|
|
}
|
|
// Last, and children first: a directory may not be writable by its owner.
|
|
for (const [directory, mode] of [...directoryModes].sort(([left], [right]) => right.length - left.length)) chmodSync(directory, mode);
|
|
return { digest: image.descriptor.digest, layers: image.layers.length, files };
|
|
}
|
|
|
|
const SLIM_LAYERS = Object.freeze([
|
|
"provider pack dependency files of 1 MiB or more",
|
|
"everything else",
|
|
]);
|
|
/**
|
|
* The layers of each variant, in order. Content that changes rarely is apart
|
|
* from content that changes with every Runner source change, so a new image
|
|
* shares its large layers with the one before it. The full image is the slim
|
|
* image plus two layers that hold the transport archives and their receipt
|
|
* and nothing else, so a reader can fetch all of them without the pack.
|
|
*
|
|
* "Dependency files" are the pack's files outside `dist`, which the Runner's
|
|
* own build writes. A large file under `dist` changes with the Runner's
|
|
* source, so it and its transport archive are with the content that does.
|
|
*/
|
|
export const RUNNER_IMAGE_LAYERS = Object.freeze({
|
|
slim: SLIM_LAYERS,
|
|
full: Object.freeze([
|
|
...SLIM_LAYERS,
|
|
"transport archives of the files in layer 1",
|
|
"the other transport archives, the overlay archive and the receipt",
|
|
]),
|
|
});
|
|
|
|
/**
|
|
* Verifies the layer rule on OCI archives of one platform's image. With both
|
|
* variants it also verifies that the full image starts with the slim image's
|
|
* layers.
|
|
* @param {{ slim?: string | null, full?: string | null }} archives
|
|
* @returns {Promise<{ problems: string[], layers: Record<string, { role: string, digest: string, bytes: number, entries: number }[]> }>}
|
|
*/
|
|
export async function checkRunnerImageLayers({ slim = null, full = null }) {
|
|
if (slim === null && full === null) throw new Error("check-layers needs --slim, --full or both");
|
|
/** @type {string[]} */
|
|
const problems = [];
|
|
/** @type {Record<string, { role: string, digest: string, bytes: number, entries: number }[]>} */
|
|
const layers = {};
|
|
const pack = RUNNER_IMAGE_PATHS.providerPack;
|
|
const archivesDirectory = RUNNER_IMAGE_PATHS.providerPackArchives;
|
|
const receiptPath = RUNNER_IMAGE_PATHS.providerPackArchiveReceipt;
|
|
const transport = (name) => name === receiptPath || name === archivesDirectory || name.startsWith(`${archivesDirectory}/`);
|
|
// Transport content anywhere else: under another directory, or under another name.
|
|
const transportLike = (name) => name.split("/").some((part) => part === receiptPath || part === archivesDirectory || part.endsWith(".tar.zst"));
|
|
/** @type {Record<string, { digest: string, names: string[], files: string[] }[]>} */
|
|
const images = {};
|
|
for (const [variant, archive] of Object.entries({ slim, full })) {
|
|
if (archive === null) continue;
|
|
const image = await listOciLayers(archive);
|
|
const roles = RUNNER_IMAGE_LAYERS[variant];
|
|
layers[variant] = image.layers.map((layer, position) => ({
|
|
role: roles[position] ?? "unexpected layer", digest: layer.digest, bytes: layer.bytes, entries: layer.entries.length,
|
|
}));
|
|
if (image.layers.length !== roles.length) {
|
|
problems.push(`the ${variant} image has ${image.layers.length} layers, not ${roles.length}`);
|
|
continue;
|
|
}
|
|
const listed = image.layers.map((layer) => ({
|
|
digest: layer.digest,
|
|
names: [...new Set(layer.entries.map((entry) => entry.name))],
|
|
files: [...new Set(layer.entries.filter((entry) => entry.kind !== "directory").map((entry) => entry.name))],
|
|
}));
|
|
images[variant] = listed;
|
|
const [large, rest] = listed;
|
|
// Each layer only adds. A directory may be in several layers; a file or a link is in one.
|
|
/** @type {Map<string, { layers: number, file: boolean }>} */
|
|
const written = new Map();
|
|
for (const layer of listed) {
|
|
const files = new Set(layer.files);
|
|
for (const name of layer.names) {
|
|
const entry = written.get(name) ?? { layers: 0, file: false };
|
|
written.set(name, { layers: entry.layers + 1, file: entry.file || files.has(name) });
|
|
}
|
|
}
|
|
summarize(problems, `${variant}: more than one layer writes`, [...written].filter(([, entry]) => entry.file && entry.layers > 1).map(([name]) => name));
|
|
summarize(problems, `${variant}: the first layer holds entries outside the pack's dependency files`,
|
|
large.names.filter((name) => name !== pack && (!name.startsWith(`${pack}/`) || name === `${pack}/dist` || name.startsWith(`${pack}/dist/`))));
|
|
summarize(problems, `${variant}: transport content is in the layers both variants share`,
|
|
[...large.names, ...rest.names].filter((name) => transport(name) || transportLike(name)));
|
|
if (!rest.files.includes(RUNNER_IMAGE_PATHS.manifest)) problems.push(`${variant}: the second layer does not hold ${RUNNER_IMAGE_PATHS.manifest}`);
|
|
}
|
|
if (images.full) {
|
|
const [large, , stable, changing] = images.full;
|
|
const numbered = (layer) => layer.files.filter((name) => name !== `${archivesDirectory}/overlay.tar.zst` && name !== receiptPath);
|
|
summarize(problems, "full: the third layer holds something other than transport archives",
|
|
stable.names.filter((name) => name !== archivesDirectory && !numbered(stable).includes(name)));
|
|
summarize(problems, "full: the last layer holds something other than transport content", changing.names.filter((name) => !transport(name)));
|
|
summarize(problems, "full: the last layer lacks", [receiptPath, `${archivesDirectory}/overlay.tar.zst`].filter((name) => !changing.files.includes(name)));
|
|
/** @type {{ path: string }[] | null} */
|
|
let files = null;
|
|
if (changing.files.includes(receiptPath)) {
|
|
try {
|
|
files = JSON.parse(await readLayerFile(/** @type {string} */ (full), 3, receiptPath)).files;
|
|
} catch {
|
|
problems.push("full: the receipt in the last layer is not readable");
|
|
}
|
|
}
|
|
if (Array.isArray(files)) {
|
|
// Archive N holds file N of the receipt. The receipt decides the layer.
|
|
const own = (file) => file.path === "dist" || file.path.startsWith("dist/");
|
|
const archiveOf = (file) => `${archivesDirectory}/${files.indexOf(file)}.tar.zst`;
|
|
const expect = (label, actual, expected) => {
|
|
summarize(problems, `full: ${label} lacks`, expected.filter((name) => !actual.includes(name)));
|
|
summarize(problems, `full: ${label} holds, against the receipt,`, actual.filter((name) => !expected.includes(name)));
|
|
};
|
|
expect("the first layer", large.files, files.filter((file) => !own(file)).map((file) => `${pack}/${file.path}`));
|
|
expect("the third layer", numbered(stable), files.filter((file) => !own(file)).map(archiveOf));
|
|
expect("the last layer", numbered(changing), files.filter(own).map(archiveOf));
|
|
}
|
|
}
|
|
if (images.slim && images.full) {
|
|
// The same blobs in the same order, not only the same files.
|
|
const shared = images.slim.map((layer) => layer.digest);
|
|
if (canonicalJson(images.full.slice(0, shared.length).map((layer) => layer.digest)) !== canonicalJson(shared)) {
|
|
problems.push("the full image does not start with the layers of the slim image");
|
|
}
|
|
}
|
|
return { problems, layers };
|
|
}
|
|
|
|
/** The text of one file in one layer of an archive. */
|
|
async function readLayerFile(archive, position, name) {
|
|
const image = readOciArchive(archive);
|
|
/** @type {Buffer[] | null} */
|
|
let chunks = null;
|
|
await image.readLayer(image.layers[position], async (raw, content) => {
|
|
if (layerEntry(raw)?.name !== name) return;
|
|
chunks = [];
|
|
for await (const chunk of content()) chunks.push(chunk);
|
|
});
|
|
if (chunks === null) throw new Error(`${name} is not in layer ${position + 1}`);
|
|
return Buffer.concat(chunks).toString("utf8");
|
|
}
|
|
|
|
const COMMAND_FLAGS = Object.freeze({
|
|
unpack: { out: "value" },
|
|
"check-layers": { slim: "value", full: "value" },
|
|
});
|
|
|
|
const invoked = process.argv[1] ? pathToFileURL(path.resolve(process.argv[1])).href : null;
|
|
if (invoked === import.meta.url) {
|
|
try {
|
|
const [command, ...rest] = process.argv.slice(2);
|
|
if (command !== "unpack" && command !== "check-layers") {
|
|
throw new Error("Usage: runner-image-oci.mjs unpack IMAGE.oci.tar --out DIR | check-layers [--slim TAR] [--full TAR]");
|
|
}
|
|
const { positional, flags } = parseLayoutArguments(command, rest, COMMAND_FLAGS[command]);
|
|
if (command === "unpack") {
|
|
if (positional.length !== 1 || typeof flags.out !== "string") throw new Error("unpack takes exactly one OCI archive and --out DIR");
|
|
const result = await unpackOciImage(path.resolve(positional[0]), path.resolve(flags.out));
|
|
process.stdout.write(`Unpacked ${result.files} files from the ${result.layers} layers of ${result.digest} into ${path.resolve(flags.out)}.\n`);
|
|
} else {
|
|
if (positional.length > 0) throw new Error(`check-layers takes no positional argument: ${positional[0]}`);
|
|
const archive = (name) => (flags[name] === undefined ? null : path.resolve(String(flags[name])));
|
|
const { problems, layers } = await checkRunnerImageLayers({ slim: archive("slim"), full: archive("full") });
|
|
for (const [variant, rows] of Object.entries(layers)) {
|
|
for (const [position, layer] of rows.entries()) {
|
|
process.stdout.write(`${variant} layer ${position + 1}: ${layer.digest} ${String(layer.bytes).padStart(11)} bytes ${String(layer.entries).padStart(6)} entries ${layer.role}\n`);
|
|
}
|
|
}
|
|
if (problems.length > 0) {
|
|
process.stderr.write(`Runner image layer check failed:\n${problems.map((problem) => `- ${problem}`).join("\n")}\n`);
|
|
process.exit(1);
|
|
}
|
|
process.stdout.write("Runner image layer check passed.\n");
|
|
}
|
|
} catch (error) {
|
|
process.stderr.write(`${error instanceof Error ? error.message : String(error)}\n`);
|
|
process.exit(1);
|
|
}
|
|
}
|