You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
Publishes the vendored-bundle detection fix. It was merged to main before this
release, and for a short window the website changelog described it as shipped
while npm still served 0.8.6 without it — the #274 defect with its polarity
reversed, advertising a fix the released artifact lacked rather than an accuracy
figure it no longer achieved.
The website changelog loses its "Unreleased / In progress" section, and the
render branches that produced "Coming next" are removed rather than left unused.
An entry describing work that has only landed on main tells a reader a fix is
available when npm install will not give it to them; the type now documents that
every entry must be an installable version.
Versions move to 0.8.7 across root, core, cli, action, server.json and the
lockfile, with the README Action pin. apps/web keeps 0.0.0 — it is unpublished
and publish.yml validates only its core dependency. The lockfile was patched by
hand; regenerating on Windows drops the @emnapi entries Linux CI needs (9
present, verified).
Co-authored-by: Claude Opus 5 <noreply@anthropic.com>
Copy file name to clipboardExpand all lines: CHANGELOG.md
+26Lines changed: 26 additions & 0 deletions
Display the source diff
Display the rich diff
Original file line number
Diff line number
Diff line change
@@ -6,6 +6,32 @@ Accuracy figures inside a released entry are the numbers measured **at that rele
6
6
left as written. The current numbers live on the [evidence page](https://usefixmap.vercel.app/evidence),
7
7
which is generated from the recorded results rather than transcribed by hand.
8
8
9
+
## 0.8.7 - 2026-08-02
10
+
11
+
### Fixed
12
+
13
+
- A pretty-printed vendored dependency bundle could rank first at high confidence. It escaped content-based bundle detection — 96 characters per line against a 400-character threshold, and one bundler marker against a threshold of two — and it escaped generated-duplicate filtering, which by design only drops a generated path that has a maintained source twin. A vendored dependency has none, so it fell through both. Detection now accepts a single marker when the path is itself conventional generated output, such as a `compiled/` segment (#446).
14
+
15
+
### Evidence
16
+
17
+
- Added a pretty-printed vendored-bundle case to the adversarial suite, which was previously all minified fixtures. The suite is 9/9 with a false-confidence rate of 0.0.
18
+
- Chalk's `source/vendor/supports-color/index.js` still ranks first for a colour-detection task. It is a vendored path whose content genuinely is the only implementation of the behavior, and it is the counterexample any path-based rule has to keep passing.
The fix above was merged to `main` before this release, and for a short window the published changelog described it as shipped while npm still served 0.8.6 without it. That is the #274 defect with its polarity reversed: the site advertising a fix the released artifact lacks, rather than an accuracy figure it no longer achieves. The website changelog no longer has an unreleased state — every entry on it is a version you can install.
24
+
25
+
### Installation
26
+
27
+
```bash
28
+
npm install --global @aryam/fixmap@0.8.7
29
+
fixmap doctor
30
+
fixmap chalk/chalk#624
31
+
```
32
+
33
+
The npm packages, MCP Registry entry, GitHub tag/release, Action tag, and production site must all resolve to 0.8.7 before the release is considered complete.
0 commit comments