Tracking: Effect framework end-to-end compat (post-#309 / #310)

Goal

Get Effect (effect@3.21.2 and later) to compile + link + run end-to-end via perry compile with perry.compilePackages: ["effect"]. This validates Perry on one of the largest pure-TypeScript libraries in the ecosystem (~150 modules, ~30k LOC for Effect.ts alone) and proves Perry can handle the heavy-generics / heavy-closures / fiber-runtime patterns that mid-to-large TS apps use.

Status snapshot (perry 0.5.405)

✅ Module collection — 362 modules collected (#310 closed: namespace re-exports traverse correctly)
✅ Codegen completes — no OOM, codegen runs to completion (#309 closed: was 34 GB RSS / 7m SIGKILL)
🔴 Codegen produces 26 module failures — see sub-issues
🔴 Link fails — 901 missing-symbol references resolving to ~5 distinct root causes

Sub-issues blocking end-to-end

#	Issue	Modules affected	Severity
#318	closure-collector dead arms (`use of undefined value '@perry_closure_*'`)	17	largest — single fix flips many modules
#314	closure-call arity ceiling at 5 args	5	high — easy to bump constant if runtime supports it
#320	ReadableStream / Web Streams stdlib not implemented	needed by Stream module	medium — requires new stdlib work
#319	`string self-append: local N not in scope`	1	low — likely subsumed by #318
#315	`String.startsWith` 2-arg form	1	low
#316	`String.slice()` 0-arg form	1	low
#317	`String.matchAll` codegen	1	low

Reproducible test bed

mkdir effect-compat && cd effect-compat
cat > package.json <<'JSON'
{
  "name": "effect-compat",
  "version": "0.0.0",
  "type": "module",
  "perry": { "compilePackages": ["effect"] }
}
JSON
bun add effect

// test_minimal.ts
import { Effect } from "effect";
const program = Effect.succeed(42);
const result = Effect.runSync(program);
console.log("result:", result);

perry compile test_minimal.ts -o /tmp/out
# Expected eventually: "result: 42"
# Today: link fails (post-#309/#310)

Recommended fix order (by leverage)

perry-codegen: closure-collector misses HIR variants — use of undefined value '@perry_closure_*' on Effect #318 — closure-collector walker fix is the biggest win and prevents this bug class from recurring on any future package. Recommend the v0.5.329 Tier 1.1 pattern (centralize via perry_hir::walker::walk_expr_children).
perry-codegen Phase D.1: closure-call arity ceiling at 5 args blocks Effect #314 — single-line fix if the runtime already supports up to 16 args (it does).
perry-codegen: String.startsWith 2-arg form (with position) unsupported #315 + perry-codegen: String.slice() 0-arg form unsupported #316 + perry-codegen Phase 2: String.matchAll codegen not implemented #317 — small focused fixes in lower_string_method.rs, can land as one PR.
perry-codegen: string self-append: local 441 not in scope on Effect's internal/channel.ts #319 — recheck after perry-codegen: closure-collector misses HIR variants — use of undefined value '@perry_closure_*' on Effect #318; likely subsumed.
perry-stdlib: ReadableStream / Web Streams API not implemented #320 — biggest scope; can be deferred until everything else is done since Effect's Stream is one of many features users want, but other Effect parts (Effect.gen, Layer, Schedule, etc.) work without it.

Definition of done

The minimal repro above produces a binary that prints result: 42 and exits 0. Bonus: Effect.gen + Effect.runPromise of an async program also works.

Backlog beyond the link errors

Even after #314-#320 are fixed, two pre-existing categorical gaps will surface in user code that touches Effect deeply:

arguments magic object — Effect's dual() helper uses it 148 times to implement data-first / data-last API duality. Already documented in docs/src/language/limitations.md. Workaround: keep using "unknown identifier 'arguments'" warnings for now; long-term, support arguments or land a dual() shim.
Intl.* — DateTime / Schedule / Duration formatting. Adjacent to existing perry-runtime i18n work.

Cross-references: #309 (OOM, fixed), #310 (silent-zero, fixed).

fix(codegen): closes #318 — delegate closure-collector descent to cen…

chore(stdlib): closes #320 — pin ReadableStream regression test (link…

fix(codegen): closes #319 — gate string self-append fast path on allo…

fix(codegen): closes #316 — accept String.slice/substring 0-arg form …

All seven sub-issues called out in the tracking table are now closed:

#	Issue	Closed in
#318	closure-collector dead arms	v0.5.408
#314	closure-call arity ceiling at 5 args	v0.5.409
#320	ReadableStream / Web Streams stdlib	v0.5.410
#319	string self-append: local N not in scope	v0.5.411
#315	String.startsWith / endsWith 2-arg form	v0.5.414
#317	String.matchAll codegen	v0.5.415
#316	String.slice() 0-arg form	v0.5.424

The remaining work to close this umbrella is the Definition of Done smoke test — the 4-line Effect.runSync(Effect.succeed(42)) repro from this issue's body should compile, link, and print result: 42. Recommend running that on a fresh checkout of main (≥ v0.5.424) and closing once verified.

Backlog notes in the issue body about arguments magic object (Effect's dual() helper, 148 sites) and Intl.* formatting still apply as separate follow-up work for users compiling Effect deeply.

Status update — perry 0.5.424

All 7 original sub-issues (#314–#320) are closed. Re-running the minimal repro on v0.5.424:

Phase	v0.5.405	v0.5.424
Modules collected	362 ✅	362 ✅
Codegen failures	26	10
Missing-symbol references	901	419
Final state	link fail	still link fail

Roughly 60% reduction in failure surface, but the binary still doesn't link. Three new failure modes have surfaced from the residual tail and are filed as new sub-issues:

#	Issue	Modules affected
#336	`@perry_class_keys_<module>__<ClassName>` collisions on same-named classes in different scopes	7 (Refinement×3, Class×2, SingleShotGen×2)
#337	`@__perry_wrap_perry_unknown_func` referenced in IR but never defined	2
#338	Residual closure-collector miss (post-#318) — undefined `@perry_closure_..._SchemaAST_ts__142`	1

Plus #320 (ReadableStream / Web Streams stdlib) is still the largest stdlib gap.

Updated checklist

End-to-end definition of done (binary prints result: 42 and exits 0) requires #336 + #337 + #338 + #320.

fix(hir,codegen): closes #336 — dedup same-named classes in one module

fix(codegen): closes #337 — emit fallback __perry_wrap_perry_unknown_…

Status update post-v0.5.464:

The smaller sub-issues called out in this umbrella are all closed (#314, #315, #316, #317, #318, #319, #320 — all per the changelog). However, the end-to-end Effect.runSync(Effect.succeed(42)) repro from this issue's "definition of done" still fails at link with 248 unresolved @perry_closure_* symbol references across multiple Effect modules.

This is the same closure-collector walker bug class that #318 closed — but for closure shapes the v0.5.408 fix didn't yet enumerate. Each new shape needs another pass through collectors.rs::collect_closures_in_expr's walker delegation or a new Expr variant added to the centralized walker.

Leaving this open as the umbrella tracking issue. The hot-fix pattern is: rerun the Effect compile, find the closure ID symbol mentioned in the linker error, find its source HIR site, identify which Expr variant the codegen-emit path is descending through but the collector isn't, fix the walker.

Re-tested perry 0.5.495 against Effect 3.21.2 (same minimal repro from this issue). Two new sub-issues blocking the link:

perry-codegen: cross-module class-name collision emits method bodies under the wrong module prefix #431 — perry-codegen: cross-module class-name collision emits method bodies under the wrong module prefix (sibling of perry-codegen: @perry_class_keys_<module>__<ClassName> symbol-name collisions when the same class name appears twice in one module #336, hits 8 Effect modules: Inspectable.ts, Streamable.ts, internal/queue.ts, internal/fiberId.ts, ParseResult.ts, Schema.ts, SchemaAST.ts, plus the secondary one named Class collisions). Has a 2-file minimal repro.
perry-codegen: @js_array_create referenced without declare in fiberRuntime.ts IR (Effect) #432 — perry-codegen: @js_array_create referenced without declare in fiberRuntime.ts IR. Independent of perry-codegen: cross-module class-name collision emits method bodies under the wrong module prefix #431.

Status snapshot at v0.5.495:

✅ Module collection, parsing, HIR-lowering: clean across all 177 source files.
❌ Codegen: 8 modules fail clang -c — all symptoms collapse to perry-codegen: cross-module class-name collision emits method bodies under the wrong module prefix #431 + perry-codegen: @js_array_create referenced without declare in fiberRuntime.ts IR (Effect) #432.
❌ Link: cascading "undefined symbol" cleanup from those 8 missing .os.
(Pre-existing) arguments warnings still fire ~60×; tracked separately as a known limitation.

Once #431 and #432 land, expect the same minimal repro (Effect.runSync(Effect.succeed("hi"))) to be the next thing to test end-to-end. If it links, the next likely runtime layer (generators / Effect.gen / yield* delegation) hasn't been exercised yet.

fix(codegen): #431 — cross-module class-name collision emitted method…

fix(runtime): validate raw string pointer via GcHeader in js_dyn_inde…

Improve Zod compatibility support (#5874)

fix(codegen): resolve class-field index/type via the authoritative in…

perf(runtime,codegen): stop unrelated descriptor installs disabling t…

fix(codegen): #5094/#26/#321 — derive GC raw-f64/pointer masks from t…

fix(runtime): dynamic index of a subnormal number is undefined on Lin…

fix(runtime): numeric-key computed calls must read the element, not d…

fix(runtime): resolve `[Symbol.iterator]` on a class declaration (#6454…

fix(runtime): static-member CALL walks the parent-closure chain (#6475)

fix: static/namespace resolution for effect Tag & Schema — web.ts ser…

fix(hir): fold an imported binding's array-named method only with arr…

fix(runtime): give iterator helpers their own class id (#7576)

fix(compile): guard the transitive class closure against global names…

Navigation Menu

Repository navigation

Tracking: Effect framework end-to-end compat (post-#309 / #310) #321

Goal

Status snapshot (perry 0.5.405)

Sub-issues blocking end-to-end

Reproducible test bed

Recommended fix order (by leverage)

Definition of done

Backlog beyond the link errors

Status update — perry 0.5.424

Updated checklist

79 remaining items

Metadata

Assignees

Labels

Type

Projects

Milestone

Relationships

Development

Tracking: Effect framework end-to-end compat (post-#309 / #310) #321

Description

Goal

Status snapshot (perry 0.5.405)

Sub-issues blocking end-to-end

Reproducible test bed

Recommended fix order (by leverage)

Definition of done

Backlog beyond the link errors

Activity

proggeramlug commented on Apr 30, 2026

proggeramlug commented on Apr 30, 2026

Status update — perry 0.5.424

Updated checklist

proggeramlug commented on May 1, 2026

proggeramlug commented on May 3, 2026

79 remaining items

Metadata

Metadata

Assignees

Labels

Type

Projects

Milestone

Relationships

Development

Issue actions