test: validate saved markup against Gutenberg's own deprecation chain

[dknauss]

Closes the largest untested risk in the plugin.

The gap

Borges is a static-save block: its markup is baked into post_content, and the editor re-validates that markup against save() every time a post is opened. When save() changes, previously saved posts keep working only because an entry in src/deprecated.js still reproduces the old markup exactly. Break one and every affected post shows "Attempt Block Recovery" — the worst regression this plugin can ship, and one that surfaces in readers' content rather than in CI.

There are 6 deprecations. src/save.test.js and src/deprecated.test.js exercise those functions directly, but nothing round-tripped real saved markup through Gutenberg's actual validator — the thing that decides whether a post breaks.

What this adds

The test registers the block against the real @wordpress/blocks registry (already a dependency — no new packages) with its shipping save plus the full deprecated chain, then parses 7 committed fixtures, one per shipped save() shape, and asserts each still validates.

It asserts block.name before isValid. That ordering matters: an unregistered block parses as core/missing and reports valid vacuously, so checking validity alone would be a false green.

Why the fixtures are frozen on disk

Minting them at test time from the same deprecated array under test would be circular — editing a deprecation would rewrite its own expectation and the test would still pass, which is precisely the regression it exists to catch. Regeneration is deliberate and env-gated:

BORGES_WRITE_DEPRECATION_FIXTURES=1 npm test -- deprecation-fixtures

When adding a deprecation, the pre-existing fixtures must come back unchanged. A diff in an existing fixture means history was altered, not appended to.

A gap found in review, and closed

The first version of this test contained zero <a href> anchors across all 7 fixtures. linkVisibleUrls splits the citation's formattedText on literal URL substrings — it never reads the CSL URL field, which is where the URL had been placed. The three linkVisibleUrls: true deprecations and the three false ones therefore serialized identically on that dimension, while the test's own doc comment claimed the path was covered.

Fixed by putting a literal URL in formattedText. The split is now real — anchors in v00, v04, v05, v06 (the true entries plus current), none in v01, v02, v03 — and a dedicated test asserts both sides still differ, so dropping the URL later fails loudly rather than quietly shrinking coverage.

Verification

Proven non-vacuous by sabotage in two independent dimensions, each failing exactly the affected fixture and no others:

Sabotage	Result
deprecation 1: headingTag h2 → h4	only deprecation 1 (index 5) fails
deprecation 4: linkVisibleUrls true → false	only deprecation 4 (index 2) fails

Both restored, git diff clean, suite green. The regenerated fixtures were byte-diffed against reconstructed pre-fix versions: every file changed by exactly two additive insertions (the JSON attribute and the rendered text/anchor), with no incidental reordering or drift.

Full suite: 42 suites / 669 tests (was 41/660). Lint and verify-metrics clean.

One config change worth noting

Jest needed transformIgnorePatterns for uuid. The repo pins uuid: ^14 in overrides, which is ESM-only with no CJS build, and @wordpress/blocks requires it transitively. The pattern only adds a transformable path and preserves Jest's default .pnp. entry; suite runtime is unchanged (~6s). The alternative — unpinning uuid — would undo a deliberate pin.

Risk

Low. Test-only plus one Jest config line. No plugin runtime code is touched, and nothing ships in the distributed package.

🤖 Generated with Claude Code

[dknauss]

@codex review

[codecov]

Codecov Report

✅ All modified and coverable lines are covered by tests.
✅ Project coverage is 81.04%. Comparing base (1140ad7) to head (d70911c).

Additional details and impacted files

@@            Coverage Diff             @@
##             main      #74      +/-   ##
==========================================
+ Coverage   81.01%   81.04%   +0.03%     
==========================================
  Files          43       43              
  Lines        3113     3113              
  Branches      522      522              
==========================================
+ Hits         2522     2523       +1     
+ Misses        185      184       -1     
  Partials      406      406              

☔ 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.
• 📦 JS Bundle Analysis: Save yourself from yourself by tracking and limiting bundle sizes in JS merges.

[chatgpt-codex-connector]

💡 Codex Review

Here are some automated review suggestions for this pull request.

Reviewed commit: 4058d16da4

ℹ️ About Codex in GitHub

Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you

• Open a pull request for review
• Mark a draft as ready
• Comment "@codex review".

If Codex has suggestions, it will comment; otherwise it will react with 👍.

Codex can also answer questions or update the PR. Try commenting "@codex address that feedback".