fix(yaml): quoted scalars spanning lines were truncated to their first line on the stpa-yaml read path (REQ-364)

[avrabe]

Fixes #969. Stacked on #970 (REQ-363), now merged; rebased onto main.

What

A single- or double-quoted scalar that spans lines was truncated by the rowan read path at its first line, opening quote kept and the rest undecoded, with no diagnostic. rowan is the production reader for stpa-yaml sources, and PyYAML's default safe_dump wraps any long quoted string at 80 columns — so STPA files written by Python tooling lose descriptions silently.

Reproduced end to end on 0.37.0 with a PyYAML-written hazards file: rivet validate → PASS, exit 0, while

rivet get H-1 → description: "'Controller issues: a ''quoted'' brake command with a colon: and a"

On this branch the same file returns the whole sentence. One-line quoted scalars were always fine.

Latent on this repository: PyYAML's composer finds 0 multi-line quoted scalars across all 118 tracked YAML files, and serde_yaml (what rivet writes with) keeps long quoted strings on one line. The exposure is files rivet didn't write. It is also a precondition of REQ-346, like REQ-361 and REQ-363.

Fix, in two layers

• yaml_cst: scan_quoted_scalar continues a quoted token past a line break only when the next non-empty line is indented strictly deeper than the line the quote opened on. Otherwise the scalar is unclosed and falls back to a plain scalar, as before. The lexer also tokenizes block scalar bodies, so this rule is what stops an apostrophe in prose (Rivet's) from swallowing later keys. It also bounds an older runaway: a backslash at a line end, or "C:\", carried a double-quoted token past its line with no limit.
• yaml_hir: fold_flow_break implements YAML 1.2 §7.3.1/§7.3.2 folding for both decode sites (titles/descriptions and field values): white space before an unescaped break dropped, one break → a space, each empty line → a newline, indentation dropped, an escaped break joins lines but keeps empty lines, and white space produced by an escape is protected from the trim.

Evidence (expected values from PyYAML)

• tests/differential_scalars.rs, a new rowan-vs-serde_yaml gate over five families: one-line quoted scalars (also checked against the intended value), literal blocks, 1024 generated multi-line quoted scalars (escapes, doubled quotes, escaped and trailing white space, empty lines, escaped breaks, CRLF), quotes and unclosed quotes inside block bodies followed by a quoted key a runaway scan would close on, and a named survey. Before the fix, 5 of its 8 multi-line styles diverged.
• 11 unit tests in yaml_cst.rs / yaml_hir.rs, where the --lib mutation gate can see them.

Negative controls (each restored; all redden the lib tests alone)

Mutation	Caught by
continuation rule removed	sibling-key test and parse_actual_hazards_file — an unbounded scan corrupts this repo's own STPA hazards
rule loosened to deeper-or-equal	only an_unclosed_quote_does_not_swallow_a_sibling_key
lexer back to one line	8 tests
no trim / escaped break as space / empty lines as space / no escape protection / CRLF not joined / single quotes not folded	the test named for each

Local gates on this rebased head

fmt 0 · clippy --all-targets -D warnings (1.97.0) 0 · cargo test --workspace 0 (2461 passed, 0 failed, corpus differential at 0 divergences) · cli_commands 218/218 · rivet validate 0 · rivet docs check 0 · yamllint 0.

REQ-364 is verified, release v0.38.0.

🤖 Generated with Claude Code

https://claude.ai/code/session_015HMQUV3u86jN2hmCtXNTc9

[github-actions]

📐 Rivet artifact delta

Change	Count
Added	1
Removed	0
Modified	0
Downstream impacted (depth ≤ 5)	0

Graph

graph LR
  REQ_364["REQ-364"]:::added
  classDef added fill:#d4edda,stroke:#28a745,color:#155724
  classDef removed fill:#f8d7da,stroke:#dc3545,color:#721c24
  classDef modified fill:#fff3cd,stroke:#ffc107,color:#856404
  classDef overflow fill:#e2e3e5,stroke:#6c757d,color:#495057,stroke-dasharray: 3 3

Loading

Added

• REQ-364

📎 Full HTML dashboard attached as workflow artifact rivet-delta-pr-976 — download from the workflow run.

_{Posted by rivet-delta workflow. The graph shows only changed artifacts; open the HTML dashboard (above) for full context.}

[github-actions]

⚠️ Performance Alert ⚠️

Possible performance regression was detected for benchmark 'Rivet Criterion Benchmarks'.
Benchmark result of this commit is worse than the previous benchmark result exceeding threshold 1.20.

Benchmark suite	Current: 96f6f17	Previous: f16245a	Ratio
store_insert/10000	17098039 ns/iter (± 1525897)	12844170 ns/iter (± 331034)	1.33
store_lookup/100	1947 ns/iter (± 9)	1606 ns/iter (± 10)	1.21
store_lookup/1000	24240 ns/iter (± 101)	19373 ns/iter (± 101)	1.25
store_lookup/10000	358394 ns/iter (± 12120)	289329 ns/iter (± 1813)	1.24
store_by_type/100	95 ns/iter (± 0)	76 ns/iter (± 0)	1.25
store_by_type/1000	95 ns/iter (± 0)	76 ns/iter (± 0)	1.25
store_by_type/10000	95 ns/iter (± 2)	76 ns/iter (± 0)	1.25
diff/1000	719128 ns/iter (± 8050)	581112 ns/iter (± 7534)	1.24
diff/10000	11082434 ns/iter (± 476317)	7291375 ns/iter (± 149125)	1.52
query/100	1176 ns/iter (± 5)	910 ns/iter (± 14)	1.29
query/1000	14799 ns/iter (± 37)	11672 ns/iter (± 131)	1.27

This comment was automatically generated by workflow using github-action-benchmark.

[codecov]

Codecov Report

❌ Patch coverage is 97.29730% with 6 lines in your changes missing coverage. Please review.

Files with missing lines	Patch %	Lines
rivet-core/src/yaml_cst.rs	94.69%	6 Missing ⚠️

📢 Thoughts on this report? Let us know!