You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
Design question from the #1096 control corpus: kotlin's branch rule counts return. Every kotlin function contributes its returns to branch mass — the corpus's 3-planted-branch file records 8 (5 returns + if/else/when), and files with zero branch keywords record branch 3 from returns alone.
Question
Assembly counting call/ret as control transfers is defensible bare-metal morphology (ledgered, not filed). Kotlin counting return while java/swift/c/etc do not looks like an inconsistency rather than a decision — it makes kotlin structurally "hotter" than identical java in cross-language comparison.
If deliberate (early-return complexity signal?), a doc note settles it; if not, dropping return from the kotlin branch alternation aligns it with its siblings.
References
#1096 corpus ledger kotlin-return-in-branch; data/kotlin/expected_signals.json.
Batch-4 addendum: return-counts-as-branch is a 5-language family, not a kotlin quirk — apex, objective-c, powershell, and zig (pattern-confirmed: ...|try|catch|orelse|break|continue|return)) all count return, while java/swift/c/scala/dart/groovy do not. Whichever way the design question resolves, the split is currently arbitrary across siblings. Corpus evidence in each language's expected_signals.json.
Both halves merged and re-baselined: engine (#2634 — return removed from branch in kotlin/apex/objective-c/powershell/solidity/zig, relocated to structural_boundaries for the 2 languages that had no other rule tracking it) and corpus (keyword-rosetta#19 — all 6 languages' branch now matches planted intent, apex's state_mutation flux-weighting ripple confirmed and re-baselined). KEYWORD_ROSETTA_REF bumped, rosetta-audit dispatched on main against the new pin. Closing.
Activity
squid-protocol commentedon Aug 31, 2026
Batch-4 addendum: return-counts-as-branch is a 5-language family, not a kotlin quirk — apex, objective-c, powershell, and zig (pattern-confirmed:
...|try|catch|orelse|break|continue|return)) all countreturn, while java/swift/c/scala/dart/groovy do not. Whichever way the design question resolves, the split is currently arbitrary across siblings. Corpus evidence in each language'sexpected_signals.json.fix(standards): stop counting bare `return` as `branch` in 6 languages (
squid-protocol commentedon Sep 1, 2026
Both halves merged and re-baselined: engine (#2634 —
returnremoved frombranchin kotlin/apex/objective-c/powershell/solidity/zig, relocated tostructural_boundariesfor the 2 languages that had no other rule tracking it) and corpus (keyword-rosetta#19 — all 6 languages'branchnow matches planted intent, apex'sstate_mutationflux-weighting ripple confirmed and re-baselined).KEYWORD_ROSETTA_REFbumped,rosetta-auditdispatched on main against the new pin. Closing.fix(standards): assembly + agc_assembly `branch` counts decisions, no…
fix(standards): branch counts decisions -- not handlers, closing word…