Merge pull request #2489 from deepseek-harness/fix/publish-retry-idempotence

fix(release): make publication retry, space out, and skip what already landed
This commit is contained in:
imccyu
2026-08-13 15:42:39 +08:00
committed by GitHub
11 changed files with 269 additions and 25 deletions
@@ -2,5 +2,5 @@
# side as of the last confirmed-consistent state. Both languages carry equal authority; # side as of the last confirmed-consistent state. Both languages carry equal authority;
# after editing either side, bring the other along and re-record with: # after editing either side, bring the other along and re-record with:
# pnpm run verify-translation-pairing --write .agents/notes/implemented/process/2026-08-10-npm-release-sequences.md # pnpm run verify-translation-pairing --write .agents/notes/implemented/process/2026-08-10-npm-release-sequences.md
2026-08-10-npm-release-sequences.md: e8138aef923e201cc0883232f48ce6921452ca00 2026-08-10-npm-release-sequences.md: e74a4ac8f2aadd8665ec0db198c6a317a0c201bc
2026-08-10-npm-release-sequences.zh.md: 7ce9fa01dbe10cbdae5585f64392d0ca6a9fb868 2026-08-10-npm-release-sequences.zh.md: e152163976f945224f2524fccd7f831ba98e8161
@@ -70,6 +70,10 @@ Publication runs only from GitHub Actions; there is no local publication path. P
The third state catches code that changed without a version bump. The first two provide idempotence — re-running publish over one artifact republishes nothing and needs no manual selection of packages. The same rule resolves the tension between one vendor release carrying several tags and a workflow that can only run from one ref: the workflow never infers which packages to publish from the tag it ran from. The third state catches code that changed without a version bump. The first two provide idempotence — re-running publish over one artifact republishes nothing and needs no manual selection of packages. The same rule resolves the tension between one vendor release carrying several tags and a workflow that can only run from one ref: the workflow never infers which packages to publish from the tag it ran from.
All three sequences decide this way, including the native one: it publishes through its own script rather than a shell loop, because a loop of bare `npm publish` calls cannot be retried — the registry answers a repeat of an existing version permanently, so one failure partway through left no way forward.
Two registry behaviours shape how a publish is attempted. Writes are spaced by at least two seconds and retried with a backoff, because publishing several packages back to back outruns the registry's own processing and earns `E409 Failed to save packument`. And every retry re-reads the registry first: a reported failure can answer a write that landed anyway, so a version that now exists with this tarball's integrity counts as published rather than as a version to place again.
### Workspace-internal references use the `workspace:` protocol ### Workspace-internal references use the `workspace:` protocol
Every reference to a workspace member uses `workspace:^`, so `pnpm pack` substitutes a range matching the target version: sibling `peerDependencies` follow the family version, and a reference to a vendored package follows that package's own line. The Landlock platform packages keep `workspace:*`, which publishes the exact version, because a platform package and its entry must agree exactly. Every reference to a workspace member uses `workspace:^`, so `pnpm pack` substitutes a range matching the target version: sibling `peerDependencies` follow the family version, and a reference to a vendored package follows that package's own line. The Landlock platform packages keep `workspace:*`, which publishes the exact version, because a platform package and its entry must agree exactly.
@@ -70,6 +70,10 @@ tag 只是 commit 指针,不是发布成功的证明。bump 会向 registry
第三态拦住「改了代码却没 bump 版本」。前两态给出幂等——同一个 artifact 重跑 publish 不会重复发布,也不需要人工挑拣包。同一条规则还解决了「一次 vendor 发布携带多个 tag,而 workflow 只能从一个 ref 触发」的矛盾:workflow 从不从触发它的 tag 去推断该发哪些包。 第三态拦住「改了代码却没 bump 版本」。前两态给出幂等——同一个 artifact 重跑 publish 不会重复发布,也不需要人工挑拣包。同一条规则还解决了「一次 vendor 发布携带多个 tag,而 workflow 只能从一个 ref 触发」的矛盾:workflow 从不从触发它的 tag 去推断该发哪些包。
三条序列都按这套判定,native 也在内:它通过自己的脚本发布,而不是 shell 循环——一串裸 `npm publish` 无法重试,registry 对「重发已存在的版本」的回答是永久失败,因此中途失败一次就没有前路了。
registry 的两个行为决定了「怎么尝试一次发布」。写入之间至少间隔两秒并带退避重试,因为连续背靠背发多个包会超出 registry 自身的处理速度,换来 `E409 Failed to save packument`。而每次重试都先重查 registry:报出来的失败可能对应一次其实已经落地的写入,所以「该版本现在存在且 integrity 与本 tarball 相同」算作已发布,而不是又一个待放置的版本。
### workspace 内部引用走 `workspace:` 协议 ### workspace 内部引用走 `workspace:` 协议
所有指向 workspace 成员的引用都用 `workspace:^`,由 `pnpm pack` 替换成匹配目标版本的范围:兄弟包的 `peerDependencies` 跟随族版本,指向 vendored 包的引用跟随那个包自己的版本线。Landlock 平台包保留 `workspace:*`(发布成精确版本),因为平台包与它的入口必须版本完全一致。 所有指向 workspace 成员的引用都用 `workspace:^`,由 `pnpm pack` 替换成匹配目标版本的范围:兄弟包的 `peerDependencies` 跟随族版本,指向 vendored 包的引用跟随那个包自己的版本线。Landlock 平台包保留 `workspace:*`(发布成精确版本),因为平台包与它的入口必须版本完全一致。
+11 -9
View File
@@ -147,6 +147,12 @@ jobs:
contents: read contents: read
id-token: write id-token: write
steps: steps:
# The publish script is the only repository file this job needs, and it
# imports nothing outside Node's builtins, so there is no install step.
- uses: actions/checkout@v4
with:
persist-credentials: false
- uses: actions/setup-node@v4 - uses: actions/setup-node@v4
with: with:
node-version: 24 node-version: 24
@@ -167,12 +173,8 @@ jobs:
fi fi
- name: Publish tarballs - name: Publish tarballs
run: | # Publication is decided per package against the registry, so re-running
version="${GITHUB_REF#refs/tags/landlock-run-v}" # this job over the same artifact skips what already landed instead of
tag_args=() # failing on it. A bare `npm publish` loop could not be retried: the
case "$version" in *-*) tag_args=(--tag next);; esac # registry answers a repeat of an existing version permanently.
while IFS= read -r tarball; do run: node ./scripts/publish-release.mjs dist/npm
# No --access: publishConfig.access in each manifest decides, and a
# command-line flag would override it.
npm publish "dist/npm/${tarball}" "${tag_args[@]}"
done < dist/npm/publish-order.txt
+1
View File
@@ -9,6 +9,7 @@
"musl-gcc", "musl-gcc",
"python3", "python3",
"sandbox-exec", "sandbox-exec",
"tar",
"taskkill", "taskkill",
"where.exe" "where.exe"
], ],
+2 -1
View File
@@ -1,6 +1,6 @@
{ {
"name": "@deepseek-ai/node-addon-landlock-run-workspace", "name": "@deepseek-ai/node-addon-landlock-run-workspace",
"version": "0.1.0", "version": "0.1.1",
"private": true, "private": true,
"type": "module", "type": "module",
"license": "BSD-3-Clause", "license": "BSD-3-Clause",
@@ -19,6 +19,7 @@
"release:assemble-prebuilds": "node ./scripts/assemble-prebuilds.mjs", "release:assemble-prebuilds": "node ./scripts/assemble-prebuilds.mjs",
"release:verify": "node ./scripts/verify-release.mjs", "release:verify": "node ./scripts/verify-release.mjs",
"release:pack": "node ./scripts/pack-release.mjs", "release:pack": "node ./scripts/pack-release.mjs",
"release:publish": "node ./scripts/publish-release.mjs",
"release:verify-packed-install": "node ./scripts/verify-packed-install.mjs" "release:verify-packed-install": "node ./scripts/verify-packed-install.mjs"
}, },
"devDependencies": { "devDependencies": {
@@ -1,6 +1,6 @@
{ {
"name": "@deepseek-ai/node-addon-landlock-run", "name": "@deepseek-ai/node-addon-landlock-run",
"version": "0.1.0", "version": "0.1.1",
"type": "module", "type": "module",
"description": "Landlock self-restrict-then-exec launcher for sandboxing subprocesses on Linux: per-platform prebuilt static binaries plus the JS seam that resolves, probes, and speaks their CLI contract", "description": "Landlock self-restrict-then-exec launcher for sandboxing subprocesses on Linux: per-platform prebuilt static binaries plus the JS seam that resolves, probes, and speaks their CLI contract",
"repository": { "repository": {
@@ -1,6 +1,6 @@
{ {
"name": "@deepseek-ai/node-addon-landlock-run-linux-arm64", "name": "@deepseek-ai/node-addon-landlock-run-linux-arm64",
"version": "0.1.0", "version": "0.1.1",
"description": "Prebuilt landlock-run Landlock launcher binary for linux-arm64 (static musl) — resolved as a file path by @deepseek-ai/node-addon-landlock-run, never imported", "description": "Prebuilt landlock-run Landlock launcher binary for linux-arm64 (static musl) — resolved as a file path by @deepseek-ai/node-addon-landlock-run, never imported",
"repository": { "repository": {
"type": "git", "type": "git",
@@ -1,6 +1,6 @@
{ {
"name": "@deepseek-ai/node-addon-landlock-run-linux-x64", "name": "@deepseek-ai/node-addon-landlock-run-linux-x64",
"version": "0.1.0", "version": "0.1.1",
"description": "Prebuilt landlock-run Landlock launcher binary for linux-x64 (static musl) — resolved as a file path by @deepseek-ai/node-addon-landlock-run, never imported", "description": "Prebuilt landlock-run Landlock launcher binary for linux-x64 (static musl) — resolved as a file path by @deepseek-ai/node-addon-landlock-run, never imported",
"repository": { "repository": {
"type": "git", "type": "git",
@@ -0,0 +1,165 @@
#!/usr/bin/env node
/**
* Publish the packed launcher family from the tarballs `pack-release.mjs`
* produced, in `publish-order.txt` order.
*
* What goes out is decided per package against the registry, never from the
* order file alone: a version the registry lacks is published, a version whose
* published tarball has the same integrity is skipped, and a version whose
* published tarball differs fails the run — that last case means the content
* changed without a version bump. Skipping on identical integrity is what makes
* re-running the publish step over the same artifact safe, which matters here
* because a partial publication used to leave no way forward: republishing an
* existing version fails permanently.
*
* Usage: `node scripts/publish-release.mjs [packed dir]`.
*/
import fs from 'node:fs';
import path from 'node:path';
import crypto from 'node:crypto';
import { spawnSync } from 'node:child_process';
import { setTimeout as sleep } from 'node:timers/promises';
import { root } from './repo.mjs';
/**
* Registry codes that answer a write which did not settle, rather than a
* rejection of what was sent. `E409 Failed to save packument` is the one this
* sequence actually hits: publishing the platform packages and the entry back
* to back can outrun the registry's own processing. A rejected payload (`E403`
* over an existing version, a malformed manifest) never clears on a retry.
*/
const TRANSIENT_PUBLISH_CODES = ['E409', 'E429', 'E500', 'E502', 'E503', 'E504', 'ETIMEDOUT', 'ECONNRESET', 'EAI_AGAIN'];
/** How many times one tarball's publish is attempted before the run fails. */
const PUBLISH_ATTEMPTS = 4;
/**
* Shortest gap between two publishes, and the first retry backoff. The registry
* needs a moment to commit a packument before the next write; back to back
* publishes are what produce `E409`.
*/
const PUBLISH_SPACING_MS = 2_000;
const destination = path.resolve(process.argv.slice(2).find((arg) => !arg.startsWith('--')) || path.join(root, 'dist', 'npm'));
/**
* @param {string} output Combined npm output.
* @returns {boolean} True when the registry reported a write it did not commit.
*/
function isTransientFailure(output) {
return TRANSIENT_PUBLISH_CODES.some((code) => output.includes(`code ${code}`));
}
/**
* @param {string} tarball Absolute tarball path.
* @returns {string} The `sha512-<base64>` integrity npm records for it.
*/
function integrityOf(tarball) {
return `sha512-${crypto.createHash('sha512').update(fs.readFileSync(tarball)).digest('base64')}`;
}
/**
* @param {string} tarball Absolute tarball path.
* @returns {{name: string, version: string}} What the packed manifest declares.
*/
function packedIdentity(tarball) {
const result = spawnSync('tar', ['-xOzf', tarball, 'package/package.json'], { encoding: 'utf8' });
if (result.status !== 0) throw new Error(`cannot read the manifest inside ${tarball}:\n${result.stderr}`);
const manifest = JSON.parse(result.stdout);
if (typeof manifest.name !== 'string' || typeof manifest.version !== 'string') {
throw new Error(`${tarball} manifest lacks name/version`);
}
return { name: manifest.name, version: manifest.version };
}
/**
* Ask the registry whether a version exists, and with what integrity.
* @param {string} name Package name.
* @param {string} version Package version.
* @returns {{kind: 'absent'} | {kind: 'present', integrity: string}} Registry state.
*/
function registryState(name, version) {
const result = spawnSync('npm', ['view', `${name}@${version}`, 'dist.integrity', '--json'], { encoding: 'utf8' });
if (result.status !== 0) {
const output = `${result.stdout}${result.stderr}`;
if (output.includes('E404') || output.includes('404 Not Found')) return { kind: 'absent' };
throw new Error(`npm view ${name}@${version} failed:\n${output}`);
}
const parsed = JSON.parse(result.stdout);
if (typeof parsed !== 'string' || parsed === '') {
throw new Error(`registry reported no dist.integrity for ${name}@${version}`);
}
return { kind: 'present', integrity: parsed };
}
/**
* Publish one tarball, retrying a registry write that did not settle.
*
* Every retry re-reads the registry first, because `E409` can answer a write
* that landed anyway: republishing a version that now exists fails permanently,
* so the same integrity appearing under the failed attempt counts as success.
* @param {string} tarball Absolute tarball path.
* @param {string} name Package name the tarball declares.
* @param {string} version Package version the tarball declares.
*/
async function publishTarball(tarball, name, version) {
// A prerelease version never takes the latest dist-tag.
const tagArgs = version.includes('-') ? ['--tag', 'next'] : [];
for (let tries = 1; tries <= PUBLISH_ATTEMPTS; tries += 1) {
// No --access: publishConfig.access in each manifest decides, and a
// command-line flag would override it.
const result = spawnSync('npm', ['publish', tarball, ...tagArgs], { encoding: 'utf8' });
const output = `${result.stdout}${result.stderr}`;
if (result.status === 0) return;
const settled = registryState(name, version);
if (settled.kind === 'present' && settled.integrity === integrityOf(tarball)) {
console.log(`landlock publish: ${name}@${version} landed despite a reported failure, continuing`);
return;
}
if (tries === PUBLISH_ATTEMPTS || !isTransientFailure(output)) {
throw new Error(`npm publish ${name}@${version} failed:\n${output}`);
}
const backoff = PUBLISH_SPACING_MS * 2 ** (tries - 1);
console.log(
`landlock publish: ${name}@${version} hit a transient registry failure`
+ ` (attempt ${tries} of ${PUBLISH_ATTEMPTS}), retrying in ${backoff}ms`,
);
await sleep(backoff);
}
}
const order = fs
.readFileSync(path.join(destination, 'publish-order.txt'), 'utf8')
.split('\n')
.filter((line) => line !== '');
let published = 0;
let skipped = 0;
for (const filename of order) {
const tarball = path.join(destination, filename);
const { name, version } = packedIdentity(tarball);
const state = registryState(name, version);
if (state.kind === 'present') {
const local = integrityOf(tarball);
if (state.integrity !== local) {
throw new Error(
`${name}@${version} is already published with different content`
+ `\n registry: ${state.integrity}\n packed: ${local}`
+ '\nBump the version, or investigate why the build is not reproducible.',
);
}
console.log(`landlock publish: ${name}@${version} already published, skipping`);
skipped += 1;
continue;
}
// Space out the writes: the gap belongs between publishes, so a run that only
// skips does not wait at all.
if (published > 0) await sleep(PUBLISH_SPACING_MS);
await publishTarball(tarball, name, version);
console.log(`landlock publish: ${name}@${version} published`);
published += 1;
}
console.log(`landlock publish: ${published} published, ${skipped} already present`);
+77 -10
View File
@@ -15,16 +15,46 @@
import { createHash } from 'node:crypto' import { createHash } from 'node:crypto'
import { readFileSync } from 'node:fs' import { readFileSync } from 'node:fs'
import { join, resolve } from 'node:path' import { join, resolve } from 'node:path'
import { setTimeout as sleep } from 'node:timers/promises'
import { parseArgs } from 'node:util' import { parseArgs } from 'node:util'
import { releaseFamily } from './families.ts' import { releaseFamily } from './families.ts'
import { attempt, isEntry, run } from './process.ts' import { attempt, isEntry } from './process.ts'
import { packedIdentity, readPublishOrder } from './tarball.ts' import { packedIdentity, readPublishOrder } from './tarball.ts'
/**
* Registry codes that answer a write which did not settle, rather than a
* rejection of what was sent. `E409 Failed to save packument` is the one this
* sequence actually hits: publishing several packages in a row can outrun the
* registry's own processing. A rejected payload (`E403` over an existing
* version, a malformed manifest) never clears on a retry and must surface.
*/
const TRANSIENT_PUBLISH_CODES = ['E409', 'E429', 'E500', 'E502', 'E503', 'E504', 'ETIMEDOUT', 'ECONNRESET', 'EAI_AGAIN'] as const
/** How many times one tarball's publish is attempted before the run fails. */
const PUBLISH_ATTEMPTS = 4
/**
* Shortest gap between two publishes, and the first retry backoff.
*
* The registry needs a moment to commit a packument before the next write; back
* to back publishes are what produce `E409`.
*/
const PUBLISH_SPACING_MS = 2_000
/** What the registry knows about one version. */ /** What the registry knows about one version. */
type RegistryState = type RegistryState =
| { readonly kind: 'absent' } | { readonly kind: 'absent' }
| { readonly kind: 'present'; readonly integrity: string } | { readonly kind: 'present'; readonly integrity: string }
/**
* Whether a failed publish is worth another attempt.
* @param output - combined npm output.
* @returns True when the registry reported a write it did not commit.
*/
function isTransientFailure(output: string): boolean {
return TRANSIENT_PUBLISH_CODES.some(code => output.includes(`code ${code}`))
}
/** /**
* The subresource integrity string npm records for a tarball. * The subresource integrity string npm records for a tarball.
* @param tarball - absolute tarball path. * @param tarball - absolute tarball path.
@@ -54,8 +84,47 @@ function registryState(name: string, version: string): RegistryState {
return { kind: 'present', integrity: parsed } return { kind: 'present', integrity: parsed }
} }
/**
* Publish one tarball, retrying a registry write that did not settle.
*
* Every retry re-reads the registry first, because `E409` can answer a write
* that landed anyway: republishing a version that now exists fails permanently,
* so the same integrity appearing under the failed attempt counts as success.
* @param tarball - absolute tarball path.
* @param name - package name the tarball declares.
* @param version - package version the tarball declares.
*/
async function publishTarball(tarball: string, name: string, version: string): Promise<void> {
// A prerelease version never takes the latest dist-tag.
const tagArgs = version.includes('-') ? ['--tag', 'next'] : []
for (let tries = 1; tries <= PUBLISH_ATTEMPTS; tries += 1) {
// No --access: the sequences do not share one access level, so a
// command-line flag could not serve both and would override the manifest
// that does. Each packed manifest decides, and
// check-workspace-constraints holds every manifest to its sequence's level.
const result = attempt('npm', ['publish', tarball, ...tagArgs])
const output = `${result.stdout}${result.stderr}`
if (result.status === 0) return
const settled = registryState(name, version)
if (settled.kind === 'present' && settled.integrity === integrityOf(tarball)) {
console.log(`release publish: ${name}@${version} landed despite a reported failure, continuing`)
return
}
if (tries === PUBLISH_ATTEMPTS || !isTransientFailure(output)) {
throw new Error(`npm publish ${name}@${version} failed:\n${output}`)
}
const backoff = PUBLISH_SPACING_MS * 2 ** (tries - 1)
console.log(
`release publish: ${name}@${version} hit a transient registry failure`
+ ` (attempt ${String(tries)} of ${String(PUBLISH_ATTEMPTS)}), retrying in ${String(backoff)}ms`,
)
await sleep(backoff)
}
}
/** Publish the family named by `--family` from the directory named by `--from`. */ /** Publish the family named by `--family` from the directory named by `--from`. */
function main(): void { async function main(): Promise<void> {
const { values } = parseArgs({ const { values } = parseArgs({
options: { family: { type: 'string' }, from: { type: 'string' } }, options: { family: { type: 'string' }, from: { type: 'string' } },
allowPositionals: false, allowPositionals: false,
@@ -86,17 +155,15 @@ function main(): void {
skipped += 1 skipped += 1
continue continue
} }
// A prerelease version never takes the latest dist-tag. // Space out the writes: the gap belongs between publishes, so a run that
const tagArgs = version.includes('-') ? ['--tag', 'next'] : [] // only skips does not wait at all.
// No --access: the sequences do not share one access level, so a if (published > 0) await sleep(PUBLISH_SPACING_MS)
// command-line flag could not serve both and would override the manifest await publishTarball(tarball, name, version)
// that does. Each packed manifest decides, and console.log(`release publish: ${name}@${version} published`)
// check-workspace-constraints holds every manifest to its sequence's level.
run('npm', ['publish', tarball, ...tagArgs])
published += 1 published += 1
} }
console.log(`release publish: family ${family.id}, ${String(published)} published, ${String(skipped)} already present`) console.log(`release publish: family ${family.id}, ${String(published)} published, ${String(skipped)} already present`)
} }
if (isEntry(import.meta.url)) main() if (isEntry(import.meta.url)) await main()