[devin-ai-integration]

What and why

Since #1205 (SARIF/CI formats count columns in UTF-16) every sysml diagnostic carries a UTF16Column, computed by utf16Column as len(utf16.Encode([]rune(string(content[lineStart:offset])))): an encoded copy of the line prefix per diagnostic (and two more per quick fix), thrown away after its length is read. The snippet path also converted the whole snippet source to []byte twice per diagnostic (utf16Column([]byte(sn.src), …), locatedFixes(…, []byte(sn.src), …)). In the 0.11.0 release-gate measurement internal/frontend/repl BenchmarkDiagnostics reads 755 → 1 044 B/diagnostic (attributes=50), 638 → 929 (200), 574 → 829 (800) against 0.10.1; the allocation profile puts the utf16.Encode copies at about half of the new allocations per diagnostic (the rest, and nearly all of the new bytes, is the semantics work of #1276/#1319 — AllSupertypes and the wider type sets — measured separately).

The column is now counted in place: utf16Units walks the prefix with utf8.DecodeRune, two units for an astral rune, one otherwise — the same count utf16.Encode gives, including one unit for an invalid byte (RuneError). The snippet source is converted to bytes once per snippet, beside the line index it already caches per snippet. Columns are unchanged.

How it was verified

• Failing-before test TestUTF16ColumnCountsWithoutEncoding (internal/frontend/repl/utf16_column_test.go): for ASCII, two-byte, astral and invalid-byte lines the column equals len(utf16.Encode(...)) + 1, and a repeated call allocates 0 times — on develop d8b60ed it allocates once per call for every line.

• A/B over six interleaved rounds on one machine (allocations and bytes are exact, so benchstat reports p=0.002 on every row; sec/op was not usable on the loaded machine and is omitted). Columns: pre-feat(cli): SARIF and JUnit reports, a baseline and -warnings-as-errors for sysml -validate #1205 731645777 | develop d8b60ed62 + fix(semantics): derive a declaration's implied generals once per symbol #1350 | this change + fix(semantics): derive a declaration's implied generals once per symbol #1350.

  row	allocs/op	B/diagnostic
  Diagnostics/attributes=50	575 | 691 | 641	762.9 | 1 044.0 | 1 020.0
  Diagnostics/attributes=200	1 483 | 1 903 | 1 703	639.6 | 928.6 | 904.6
  Diagnostics/attributes=800	5 637 | 7 261 | 6 461	574.1 | 829.1 | 805.1
  tests/perf REPLLoadFile	5.407M | 6.872M | 6.872M	—
  tests/perf REPLSubmitSnippet	4.703M | 6.829M | 6.829M	—

  The change removes one allocation per diagnostic (two with a quick fix) and the per-diagnostic snippet copies — half of the row's allocation growth since feat(cli): SARIF and JUnit reports, a baseline and -warnings-as-errors for sysml -validate #1205; REPLLoadFile/REPLSubmitSnippet do not render diagnostics and are unchanged, as expected.

• go build ./..., go vet ./internal/frontend/repl, gofmt -l . empty, go test ./internal/frontend/repl (with the training corpus present), python3 scripts/changelog.py check.

Checklist

• make test and make lint pass locally (repl package run locally; the full suite runs in CI)
• Tests added or updated for the change
• Documentation extended where it already covers the surface (no doc covers this internal)
• Changelog entry added as changes/unreleased/<slug>.<section>.md, not as an edit to CHANGELOG.md
• baselines regenerated and make docs-counts run if a gate count moved (no gate count moved)
• No internal work-item labels (waves, slices, F4, K5) in the body, docs, or changelog

Link to Devin session: https://nasa-jpl-demo.devinenterprise.com/sessions/cf22d7a746814d20b2f0eaf11682fef9
Open in Devin Desktop: https://nasa-jpl-demo.devinenterprise.com/desktop/session/cf22d7a746814d20b2f0eaf11682fef9?variant=devin
Requested by: @HuiJun

[devin-ai-integration]

I'll fix CI failures and address comments from users with write access. I'll skip comments containing "(aside)".

• Disable automatic comment, CI, and merge conflict monitoring

[devin-ai-integration]

✅ Devin Review: No Issues Found

Devin Review analyzed this PR and found no bugs or issues to report.