Repository navigation
[GEN-E5] Fixed per-operation behaviour moves into hand-written bases #482
Description
Part of #350, the quality-refactoring roadmap (road to 1.0, step 3).
Goal: Fixed per-operation behaviour moves into hand-written bases (a spike first, then a design record). Lands paired with RT-E9.
Lane: Generator + Runtime. Status: todo.
Every PR of this epic starts its title with [GEN-E5]. The finding issues it closes are linked as sub-issues; this issue closes when the epic's last PR merges and master CI is green.
Activity
Blind-Striker commentedon Oct 18, 2026
Clone baseline ownership from #517 (#452): the guard-prelude ignore made a pre-existing duplicated body visible:
tools/OpenCode.Sdk.Tools/Generator/Emission/EnvelopeEmitter.cs:10-20↔ResponseAdapterEmitter.cs:11-21(the Emit preludes). It is baselined with this issue as its owner, since ResponseAdapterEmitter is GEN-E5's (GEN-E4 decision 2). When the hand-written bases replace it, drop the fingerprint fromtools/jscpd/tools.baseline.json(verify-clonesreports it stale).Blind-Striker commentedon Oct 24, 2026
GEN-E4 (#481) takes one item from
ResponseAdapterEmitter, which otherwise stays this epic's: its singleOpenCodeJsonContextregistry literal becomes aRegistryNamescall, byte-identical, in GEN-E4 PR 2. It is a named exception, so that GEN-E4's proof ("noOpenCodeJsonContextliteral outsideRegistryNames") holds across Emission. GEN-E5 should not redo it. The rest ofResponseAdapterEmitter(its size, itsNumber,StatusVerdictandRouteshelper copies, and the per-operation tables) stays here, as GEN-E4's decision 2 assigned.Blind-Striker commentedon Oct 25, 2026
Moved to milestone 1.1 by the 1.0 triage (2026-10-25). The epic's refactor is internal; the one piece that changes documented consumer behaviour stays in milestone 1.0 as its own issue and lands before 1.0.0 as a small PR: #30 (the envelopes' printing), with #177 (the doc-comment wording).