Repository navigation
ci: enable trusted publishing releases (v3.x) - #229
Merged
Merged
Conversation
Pin the reusable release workflow to the fastify/workflows#243 merge and allow the pre* bump types (published under `three-next`), as on v2.x. Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
mcollina
marked this pull request as ready for review
October 17, 2026 19:06
Eomm
approved these changes
Oct 18, 2026
This was referenced Oct 25, 2026
This was referenced Oct 27, 2026
vharseko
added a commit
to OpenIdentityPlatform/OpenAM
that referenced
this pull request
Oct 29, 2026
…26-110725 fast-uri: Scheme spoofing, host confusion and relative-reference rewriting (3.1.7 -> 3.1.9) (#1204) Bumps `fast-uri` from 3.1.7 to 3.1.9 in both UI modules (`package-lock.json` only), clearing Dependabot alerts #377–#380 (`openam-ui-js-sdk`) and #382–#385 (`openam-ui-ria`). This follows #1118, which took both modules from 3.1.5 to 3.1.7; 3.1.8 and 3.1.9 are the next two security releases of the 3.x line. Supersedes #1203, which bumps `openam-ui-js-sdk` only; its lock-file change is byte-for-byte the one here. ## Advisories `fast-uri` is the RFC 3986 parser behind `ajv` 8 (`parse`, `serialize`, `resolve`, `normalize`, `equal`). In all five advisories its output names a different URI than the one it was given — a different scheme, host or path than RFC 3986 and WHATWG parsers read, or than the caller checked. Each one matters only to an application that makes a security decision on one representation of an attacker-supplied URI and then uses another. **CVE-2026-110306 / GHSA-w74x-v6f9-j3xg** (fixed in 3.1.9) — URI scheme spoofing; encoding error (CWE-172), validate before canonicalize (CWE-180). `parse` and `serialize` ran the scheme through the legacy global `unescape()` before validating it, although RFC 3986 §3.1 allows no percent-encoding in a scheme. `normalize('%6Aavascript:alert(1)')` and `resolve('https://a.example/', '%6Aavascript:alert(1)')` returned `javascript:alert(1)`, the `%u006A` form did the same, and `equal('%68ttp://a/', 'http://a/')` was `true`. An application that screens dangerous schemes on the raw input and then stores or renders the normalized form gets a live `javascript:`, `data:` or `vbscript:` URL. 3.1.9 validates the scheme as written and never decodes it, so such input fails with `URI scheme is malformed.` (fastify/fast-uri@597f254). **CVE-2026-110180 / GHSA-wjr9-j795-jxcr** (fixed in 3.1.9, affected since 3.1.3) — Host confusion; improper input validation (CWE-20), interpretation conflict (CWE-436). For `http`, `https`, `ws`, `wss` and any scheme with `domainHost`, `canonicalizeHost()` handed the host to the WHATWG `URL` parser and adopted its `.hostname` without checking that the parser had read the whole input as a host. fast-uri's host group admits `@`, and the WHATWG parser strips TAB, LF and CR, so `http://a@b@evil.example/` came back with host `evil.example` and no `error`, a host carrying a raw TAB, CR or LF lost those characters, and `equal()` treated the smuggled host as the clean one. 3.1.9 refuses a domain host that contains a character outside the RFC 3986 reg-name set, or that the WHATWG parser does not read as a bare host: `parse` reports `URI host is malformed.` and `resolve` throws (fastify/fast-uri@ddfc6a5). **CVE-2026-110095 / GHSA-xhq7-xmqj-769h** (fixed in 3.1.9) — Improper resolution of path equivalence (CWE-41). `normalize()` and `serialize()` applied `remove_dot_segments` to the path of a relative-path reference, although RFC 3986 §5.2.4 defines it only for the merged target path during resolution. `normalize('a/../b')` returned `/b` and `normalize('../x')` returned `x`, so `resolve(base, normalize(ref))` named a different resource than `resolve(base, ref)` and `equal()` conflated distinct references. 3.1.9 keeps a relative reference relative and keeps its unresolvable leading `..` segments (`removeDotSegmentsRelative`; fastify/fast-uri@b110a35, fastify/fast-uri@9eb8759). **CVE-2026-110725 / GHSA-333x-w4r6-6crp** (fixed in 3.1.9) — Scheme injection; improper resolution of path equivalence (CWE-41), improper encoding or escaping of output (CWE-116). With the non-default option `reference: 'suffix'`, `serialize()`, `normalize()` and `resolve()` emitted a reference without authority whose first path segment kept a bare colon, which RFC 3986 §3.3 forbids in `path-noscheme`: `normalize('javascript:alert(1)', { reference: 'suffix' })` returned `javascript:alert(1)`, which reparses as a live scheme. 3.1.9 escapes that colon in suffix output, as the default serialization already did (fastify/fast-uri@39e6493). Only callers that pass `reference: 'suffix'` are affected. **CVE-2026-86472 / GHSA-hrr3-gc8f-f4qj** (fixed in 3.1.8) — Improper handling of case sensitivity (CWE-178). fast-uri folded the host to lowercase before percent-decoding it, so `%41` decoded to an uppercase `A` that was never folded, and for a scheme-relative reference (`//host`) no scheme-specific canonicalization repaired it: `parse('//%41.com').host` was `A.com` while `//a.com` and `//A.com` gave `a.com`, and `equal('//%41.com', '//a.com')` was `false`. A case-sensitive host allow- or denylist could be evaded, although the evading spelling still reaches the same host. 3.1.8 folds the case after decoding (fastify/fast-uri@c88b59e). Dependabot auto-dismissed the matching alerts (#347 in `openam-ui-js-sdk`, #353 in `openam-ui-ria`) on 2026-09-30, so they are not open, but 3.1.7 is inside the affected range and this bump is what takes both modules past it. | | Severity | Affected (3.x) | Fixed in | |---|---|---|---| | GHSA-w74x-v6f9-j3xg | Medium — CVSS 3.1 6.1 (`AV:N/AC:L/PR:N/UI:R/S:C/C:L/I:L/A:N`) | >= 3.0.0, < 3.1.9 | 3.1.9 | | GHSA-wjr9-j795-jxcr | Medium — CVSS 3.1 4.8 (`AV:N/AC:H/PR:N/UI:N/S:U/C:L/I:L/A:N`) | >= 3.1.3, < 3.1.9 | 3.1.9 | | GHSA-xhq7-xmqj-769h | Low — CVSS 3.1 3.7 (`AV:N/AC:H/PR:N/UI:N/S:U/C:N/I:L/A:N`) | >= 3.0.0, < 3.1.9 | 3.1.9 | | GHSA-333x-w4r6-6crp | Low — CVSS 3.1 3.7 (`AV:N/AC:H/PR:N/UI:N/S:U/C:N/I:L/A:N`) | >= 3.0.0, < 3.1.9 | 3.1.9 | | GHSA-hrr3-gc8f-f4qj | Medium — CVSS 3.1 4.8 (`AV:N/AC:H/PR:N/UI:N/S:U/C:L/I:L/A:N`) | >= 3.0.0, < 3.1.8 | 3.1.8 | The 2.x and 4.x lines received the same fixes in 2.4.8 and 4.2.3 (2.4.7 and 4.1.5 for GHSA-hrr3); neither line is in these lock files. ## Impact on OpenAM None at runtime, none in the build. In both modules `fast-uri` is a transitive development dependency (`dev: true`) that no OpenAM source imports, and it does not end up in any module zip or in the WAR. ### `openam-ui-js-sdk` The single copy in the lock file is required (`^3.0.1`) by the three nested `ajv` 8 copies under `@microsoft/tsdoc-config` (8.18.0), `@rushstack/node-core-library` (8.20.0) and `ajv-formats` (8.20.0), all of which belong to the `@microsoft/api-extractor` 7.58.12 subtree — a direct `devDependency` that `vite-plugin-dts` also uses through `unplugin-dts`. The top-level `ajv` 6.15.0 that `eslint` uses parses URIs with `uri-js` and does not load `fast-uri`. Nothing from the subtree reaches the `app` / `lib` zips (`target/app`, `target/lib`) or the WAR's `extui/`. The Maven build does load it. `npm run build` (compile phase) runs `build:lib` → `vite build --config vite.lib.config.ts`, whose `dts({ bundleTypes: true, … })` plugin calls API Extractor to roll up the declarations: - `ExtractorConfig.prepare` validates the configuration object that `unplugin-dts` builds in memory against API Extractor's bundled `api-extractor.schema.json` (the module has no `api-extractor.json`). - `TSDocConfigFile.loadFile` validates API Extractor's bundled `extends/tsdoc-base.json` (the module has no `tsdoc.json`) against `@microsoft/tsdoc`'s `tsdoc.schema.json`. Ajv calls `fast-uri` only while compiling those schemas — `parse`, `serialize` and `resolve` on their `$id` / `$ref` values and base IDs (`dist/compile/resolve.js`, `dist/compile/index.js`) — and never on the data it validates; the `uri` and `uri-reference` formats of `ajv-formats` are regular expressions. It never calls `normalize()` or `equal()` and never passes `reference: 'suffix'`, so GHSA-333x cannot occur at all. The only strings `fast-uri` sees are identifiers from JSON schemas shipped inside the npm packages — nothing from OpenAM sources, users or the network. `npm run test:run` (`vitest.config.js`) and `npm run dev` (`vite.config.ts`) configure only `@vitejs/plugin-react` and never load API Extractor. ### `openam-ui-ria` The single copy in the lock file is required (`^3.0.1`) by the single `ajv`, 8.17.1. That `ajv` is a direct `devDependency` only so that `overrides.eslint.ajv = "$ajv"` can replace the ajv 4 that `eslint` 3.8.1 would otherwise pull in through `table` 3.8.3 and its `ajv-keywords` 1.5.1 peer; nothing else depends on it. Nothing in the build loads it. `npm run build:production` (`grunt prod`) runs the `eslint` task through `grunt-eslint` 19.0.0. eslint 3.8.1 validates its configuration with `is-my-json-valid` and requires `table` only from its own `table` formatter (`lib/formatters/table.js`), but the Gruntfile passes `require.resolve("eslint-formatter-warning-summary")`, which `CLIEngine.getFormatter` loads by path and which needs only `chalk` and `text-table`. `npm run test` (`grunt karma:build`) has no path to `ajv` either. The module zip (`src/main/assembly/zip.xml`) packs `target/compiled`. Neither flaw is reachable in either module. The bump takes the development toolchain out of the affected range. ## Change The `node_modules/fast-uri` entry in `openam-ui/openam-ui-js-sdk/package-lock.json` and in `openam-ui/openam-ui-ria/package-lock.json`: 3.1.7 → 3.1.9 (`version`, `resolved`, `integrity`). Neither `package.json` changes — every `ajv` that requires `fast-uri` asks for `^3.0.1`, which 3.1.9 satisfies. No code change. Apart from the five fixes, v3.1.7...v3.1.9 contains tests, a README note on suffix-reference authority ambiguity and a release workflow (fastify/fast-uri#229): 3.1.9 is the first `fast-uri` release published from GitHub Actions through npm trusted publishing, which is the new releaser Dependabot reports on #1203. Verified: - both lock files carry an `integrity` that matches `npm view fast-uri@3.1.9 dist.integrity` and the SHA-512 of the downloaded tarball; the `openam-ui-js-sdk` lock file is identical to the one on #1203 - the published 3.1.9 carries the fixes: `index.js` validates the scheme without `unescape()` (`validateScheme`), rejects malformed domain hosts (`hasForbiddenDomainHost`, `isHostOnlyURL`), keeps relative references relative (`removeDotSegmentsRelative`), escapes the first-segment colon under `reference: 'suffix'`, and folds host case after percent-decoding; the package still has no dependencies - the registry entry was published by the `GitHub Actions` trusted publisher with an SLSA provenance attestation, and its `gitHead` is fastify/fast-uri@0def703 (`Bumped v3.1.9`) - `node_modules/fast-uri` is the only `fast-uri` entry in each lock file; the `openam-ui-api` lock file does not resolve it - the GitHub Advisory Database lists no advisory affecting `fast-uri@3.1.9` ## References - GHSA-w74x-v6f9-j3xg - GHSA-wjr9-j795-jxcr - GHSA-xhq7-xmqj-769h - GHSA-333x-w4r6-6crp - GHSA-hrr3-gc8f-f4qj - https://nvd.nist.gov/vuln/detail/CVE-2026-110306 - https://nvd.nist.gov/vuln/detail/CVE-2026-110180 - https://nvd.nist.gov/vuln/detail/CVE-2026-110095 - https://nvd.nist.gov/vuln/detail/CVE-2026-110725 - https://nvd.nist.gov/vuln/detail/CVE-2026-86472 - fastify/fast-uri@597f254 - fastify/fast-uri@ddfc6a5 - fastify/fast-uri@b110a35 - fastify/fast-uri@9eb8759 - fastify/fast-uri@39e6493 - fastify/fast-uri@c88b59e - https://github.com/fastify/fast-uri/releases/tag/v3.1.8 - https://github.com/fastify/fast-uri/releases/tag/v3.1.9 - fastify/fast-uri@v3.1.7...v3.1.9
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Backport of 42e2318 (trusted publishing via the
fastify/workflowsreusable release) tov3.x.Differences from
main:threedist-tag (the existing npm dist-tag for v3), solateststays on v4;make-latest: false, so the GitHub release isn't marked as latest;pre*bump types with an optionalpreid; prereleases go tothree-next.Pinned to the fastify/workflows#243 merge commit (
b0a2ab2), which adds thedist-tag,make-latestandpreidinputs.Releases must be dispatched from the Actions tab with the
v3.xbranch selected.🤖 Generated with Claude Code