Non React Coverage Roadmap Phase 3 Execution

Scope

• Execute Tier 1 and Tier 2 from 2026-03-24-non-react-coverage-roadmap-phase-3.md.
• Keep the batch file-first.
• Expand existing honest specs before inventing wrapper smoke junk.

Phase Plan

• [completed] Phase 1: inspect current seams and harnesses
• [completed] Phase 2: Tier 1 (link, markdown)
• [completed] Phase 3: core and autoformat
• [completed] Phase 4: basic-nodes, table, list-classic, suggestion
• [completed] Phase 5: verification and roadmap status update

Learnings Applied

• Restore Bun spies in the same spec file or full-suite order lies. See 2026-03-24-spec-spies-must-be-restored-or-full-suite-order-breaks.md.
• Keep Bun module mock surfaces consistent across related specs. See 2026-03-24-bun-module-mocks-must-export-a-consistent-surface-across-related-specs.md.
• Filtered Turbo typecheck can lie. Use the normal build-first flow, then serialize Turbo if it gets weird. See 2026-03-24-turbo-filtered-typecheck-can-lie-when-package-typecheck-passes.md.
• Markdown fallback tests should cover incomplete tails and void-block preservation. See 2026-03-14-markdown-incomplete-mdx-fallback-drops-void-blocks.md.

Notes

• core has the biggest cluster. Expand the existing HTML and static specs instead of making more tiny one-off files than necessary.
• markdown already has nearby serializer and deserializer specs. Reuse them.
• basic-nodes should stay honest: one plugin contract expansion, not a new wrapper sweep.

Outcome

• Tier 1 and Tier 2 are complete.
• upsertLink.ts and markdownToSlateNodesSafely.ts each had dead branches removed instead of receiving fake tests.
• New direct specs landed for the remaining honest helper seams in core, markdown, table, list-classic, and suggestion.
• This pass exposed and fixed two real product bugs:
  • splitIncompleteMdx.ts treated a closing > at end-of-string as incomplete.
  • getTableCellBorders.ts accepted any parent node as a table ancestor instead of requiring the real table type.

Verification

• Targeted roadmap slice:
  • bun test ... on 19 touched specs -> 105 pass, 0 fail
• Build-first gate:
  • pnpm install
  • pnpm turbo build --filter=./packages/link --filter=./packages/markdown --filter=./packages/core --filter=./packages/autoformat --filter=./packages/basic-nodes --filter=./packages/table --filter=./packages/list-classic --filter=./packages/suggestion
  • pnpm turbo typecheck --concurrency=1 --filter=./packages/link --filter=./packages/markdown --filter=./packages/core --filter=./packages/autoformat --filter=./packages/basic-nodes --filter=./packages/table --filter=./packages/list-classic --filter=./packages/suggestion
  • pnpm lint:fix
• Post-lint confirmation:
  • targeted roadmap slice rerun -> 105 pass, 0 fail

Verification Notes

• A broad bun test packages/link/src packages/markdown/src packages/core/src packages/autoformat/src packages/basic-nodes/src packages/table/src packages/list-classic/src packages/suggestion/src sweep is still dirty outside this roadmap slice. The failures are shared suite-order or mock-surface debt in unrelated areas, not regressions in the touched Phase 3 seams.