test: add mutation testing over the pure logic

[skjnldsv]

Coverage tells you a line ran. It does not tell you anything checked what the line did. My SVG sanitiser test in #55 was the clean example: it passed with the sanitiser deleted, because a browser never runs script in an <img>-loaded SVG. I only caught that because I happened to try.

Stryker does that systematically. It changes the code in small ways, flips a comparison, deletes a call, swaps a string, and reruns the tests. A mutant that survives is a change to the library that no test noticed.

Scoped to lib/utils and lib/composables, which takes about three minutes and is where a survivor usually means a real gap rather than a change nothing could observe. npm run test:mutation.

Not wired into CI and no threshold set. Chasing a percentage produces padded assertions; the report is for reading.

What the first run found

useRotation.ts came out at 69.61%, twenty-nine survivors. That is code I wrote two days ago and gave twenty tests to, and was pleased with.

Three of those survivors were worth fixing:

The direction was never checked. My stub for rotateOrientation ignored which way it was asked to turn, so changing 'left' to 'right' in the composable passed the whole suite. For a button whose entire job is to rotate left. The end-to-end test would have caught it, but nothing at unit level did.

saving is returned by the composable and no test ever read it, so three mutants lived in it. Observing it needs the request held open, because advancing fake timers also drains the promise chain and the write is over before there is anything to see.

An etag comes back quoted either literally or as ", and only the literal form was exercised.

Four tests later: 78.43%, twenty survivors. Most of the rest cannot be killed while axios is mocked, for instance the responseType, and I have left them rather than writing assertions that pretend otherwise.

plyrTranslations.ts is excluded. It is a table of strings and scored 1.43%, which measures nothing useful.

Where the rest of the code stands

file	score	survivors
previewUtils.ts	97.5%	1
livePhotoUtils.ts	90.6%	2
canDownload.ts	86.7%	6
history.ts	73.3%	31
useViewerActions.ts	74.3%	15
usePlyrPlayer.ts	43.0%	44

history.ts and usePlyrPlayer.ts are the obvious next places to look. I have not touched them here; this PR is the tool plus the one file I used to find out whether it was worth having.

It runs with inPlace, because the sandbox copy breaks vite's filesystem allowlist on the ?raw icon imports. Stryker backs the files up and restores them, and git status is clean after a run.

382 unit tests.

👾 This pull request was assisted by Claude Code, commits carry an Assisted-by trailer.

[codecov]

Codecov Report

✅ All modified and coverable lines are covered by tests.
✅ Project coverage is 88.62%. Comparing base (e2333e0) to head (70c298c).

Additional details and impacted files

@@           Coverage Diff           @@
##             main      #58   +/-   ##
=======================================
  Coverage   88.62%   88.62%           
=======================================
  Files          39       39           
  Lines        2743     2743           
  Branches      561      562    +1     
=======================================
  Hits         2431     2431           
  Misses        305      305           
  Partials        7        7           

☔ View full report in Codecov by Harness.
📢 Have feedback on the report? Share it here.

🚀 New features to boost your workflow:

• ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.