fix(content): key the entry memo by (file, slug, locale) to stop fresh-build slug poisoning#29
Merged
Merged
Conversation
…h-build slug poisoning The loadEntry memo was keyed by file path + mtime, but the cached ContentEntry embeds the slug and locale ARGUMENTS, which are scan-dependent. generateContent's bootstrap two-pass scans the same tree twice in one process: pass 1 without the config (regex locale guessing, where a 2-3 letter folder like docs/adr/ matches the BCP-47 shape and becomes a locale bucket whose files get FLAT slugs) and pass 2 with the declared locale set (nested slugs). Same file, same mtime, so pass 2 got pass 1's poisoned entry back and froze flat slugs into app/_content.ts. Only FRESH builds hit this (app/_content.ts must be missing for pass 1 to run), which made it look environment-specific: CI/action builds in a clean dir flattened docs/adr/0001-x to /0001-x while a warm local rebuild of the same tree nested it correctly. Regression tests cover both the scan pair and the frozen emit; verified end to end on a fresh build (control on the published package flattens, patched build nests and rewrites in-content links).
There was a problem hiding this comment.
Pull request overview
This PR addresses a fresh-build-only slug flattening bug in the June content pipeline, caused by memoization returning a cached ContentEntry whose embedded slug/locale came from an earlier “guessed locale” scan.
Changes:
- Adds regression coverage to ensure a guessed-locale scan (pass 1) cannot poison a declared-locale rescan (pass 2).
- Adds a
generateContentModuleassertion to ensure the frozen_content.tsoutput contains nested slugs and no phantom locale artifacts. - Publishes a patch changeset for
@junejs/serverdescribing the fix and its impact.
Reviewed changes
Copilot reviewed 2 out of 3 changed files in this pull request and generated 1 comment.
| File | Description |
|---|---|
| packages/june/test/content.test.ts | Adds a focused regression suite covering the two-pass “slug poisoning” scenario and verifying frozen module output. |
| .changeset/content-memo-slug-poisoning.md | Declares a patch release and documents the fresh-build slug flattening fix. |
💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.
Comment on lines
+351
to
+356
| test("the frozen _content.ts of the rescan carries the nested slugs", () => { | ||
| generateContentModule(dir, undefined); // pass 1 (throwaway) | ||
| const { code } = generateContentModule(dir, []); // pass 2 — what lands in app/_content.ts | ||
| expect(code).toMatch(/"slug": "adr\/0001-x"/); | ||
| expect(code).not.toMatch(/"slug": "0001-x"/); | ||
| expect(code).not.toMatch(/"locale": "adr"/); |
Contributor
Author
There was a problem hiding this comment.
Great catch, fixed in b3bd4fb. The separator is now the \u0000 escape sequence in source (runtime key unchanged: NUL still separates, and it cannot appear in file paths or slugs), so the file is plain text to grep/diff tooling again. Verified: 0 raw NUL bytes, file(1) reports UTF-8 text, suite still 35/35.
The (file, slug, locale) memo key used literal NUL bytes as separators, which made content.ts a binary file to grep/ripgrep/diff tooling. Write the separator as the \u0000 escape instead: the runtime key is byte-identical (NUL still cannot appear in paths or slugs), but the source file is plain text again.
linyiru
added a commit
to linyiru/rustbgpd
that referenced
this pull request
Jul 3, 2026
…ttening) The ADRs were dropped because fresh CI builds flattened docs/adr/0001-x to /0001-x (a June content-memo bug, junebuild/june#29). With @junejs/server 0.0.56 the fresh build nests them correctly, so the Internals tab gets its design records back.
This was referenced Jul 3, 2026
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.
What
Fixes a slug-flattening bug that only bites FRESH builds: a docs subfolder whose name matches the BCP-47 shape (
adr/,cli/,api/,sdk/, ...) had its pages frozen with flat slugs (docs/adr/0001-x.mdrendered at/0001-xinstead of/adr/0001-x), while a warm rebuild of the same tree nested them correctly. Found in production on a Kura site built by the kurajs/pages action (always a clean dir): the sidebar and prerendered pages were internally-consistent but wrongly flat.Root cause
Three cooperating steps, pinned by a deterministic repro:
generateContent, intentional): on a fresh checkout the wrapper config imports the not-yet-generatedapp/_content.ts, so pass 1 scans with regex-guessed locales.adrmatchesLOCALE_DIR(/^[a-z]{2,3}(-...)*$/), sodocs/adr/is treated as a locale bucket whose files get FLAT slugs (0001-x,locale: "adr").loadEntry's memo was keyed byfile+ mtime, but the cachedContentEntryembeds theslug/localearguments. Pass 2 (declared locale set, computes the correct nestedadr/0001-x) got pass 1's poisoned entry back (same file, same mtime) and froze the flat slugs intoapp/_content.ts.Fix
Key the memo by
(file, slug, locale)+ mtime. One-line semantic change; dev-edit invalidation via mtime is unchanged, and the map stays bounded (one extra entry per file only during the bootstrap pass pair).Tests
scanCollection: a guessed-locale scan (pass 1) must not poison the declared-locale rescan (pass 2): nested slugs, no phantomlocale: "adr".generateContentModule: the frozen pass-2 emit carries"slug": "adr/0001-x"and no flat/locale remnants.@junejs/server0.0.55) flattens to/0001-x; with this patch the same fresh build nests/adr/0001-xand in-content links rewrite correctly.Changeset:
@junejs/serverpatch.