Repository navigation
Remove dead code from C++ bindings, bindgen glue, ast, and orphaned scripts - #39581
Merged
Jarred-Sumner merged 1 commit intoAug 23, 2026
Merged
Conversation
Contributor
|
Warning Review limit reachedYour included review limit has been reached. You’re in a promotional period — use the checkbox below to run this review for free:
On-demand reviews are free for the next 31 days. After that, they cost $0.25 per reviewed file. How can I continue?Run this review now using the option above, or comment You can also wait for the limit to reset (next review available in 56 seconds), then comment An organization admin can change what happens after included review limits in Billing. How do review limits work?CodeRabbit enforces per-developer PR review limits within each organization. For paid Pro and Pro+ reviews, CodeRabbit uses a developer's included PR review attempts over the past 7 days to set the current hourly allowance. At typical activity levels, the full plan allowance applies. Higher sustained activity can lower the allowance until earlier attempts leave the 7-day window. Please refer docs for additional details. Review details⚙️ Run configurationConfiguration used: Path: .coderabbit.yaml Review profile: ASSERTIVE Plan: Pro Run ID: 📒 Files selected for processing (55)
Comment |
![claude[bot]](https://avatars.githubusercontent.com/in/1236702?s=60&v=4){kind=small}
Contributor
There was a problem hiding this comment.
I reviewed this PR and didn't find any bugs. Because it's a broad deletion sweep across 63 files — including ZigGlobalObject, 33 ncrypto methods plus the DataPointer::secure_ field, the bindgen adapter layer, and 9 scripts whose supersession claims reference workflow history — a maintainer look to confirm scope would still be worthwhile.
What was reviewed:
- Grepped every removed C++/Rust symbol (
m_worldIsNormal/offsetOfWorldIsNormal,m_JSCryptoKey,performMicrotaskVariadic,m_cachedGlobalProxyStructure,io_darwin_close_machport,BindgenOptionalRepr/BindgenOptionalCustom/BindgenExternalShared/ExternalSharedOptional,reserveEstimated,Binding::init,ImportsIteratorEntry.value) — zero remaining references in src/ or codegen. - Deleted headers/.cpp files (DOMJIT*, MessagePortChannel*, JSDOMBuiltinConstructorBase, BroadcastChannelRegistry.h, JSDOMConvertSerializedScriptValue.h) have no remaining
#includesites and no CMake/.cmake listings; theBunBroadcastChannelRegistryhits are a separate class. DataPointer::operator=uses placement-new via the move ctor, so droppingsecure_there is complete;reset()no longer branches on it.- The
#include "webcrypto/JSCryptoKey.h"in ZigGlobalObject.cpp was flagged as potentially orphaned but ruled out.
Extended reasoning...
Overview
This PR is a reopen of #37149: pure dead-code deletion across 63 files with no behavioral changes intended. It removes ~1,800 lines spanning C++ JSC bindings (ncrypto, ZigGlobalObject, DOMJIT IDL headers, 13 empty stub files), Rust bindgen/ast/ptr glue, and 9 orphaned scripts. The only additions are a direct #include <JavaScriptCore/JSPromise.h> in WorkerMessagingProxy.cpp (previously transitive via a deleted unified-source neighbor) and <JavaScriptCore/FrameTracers.h> at the three sites that used to get it via DOMJITHelpers.h.
Security risks
None introduced. The ncrypto deletions remove uncalled wrappers (FIPS toggles, secure-heap, unused cipher factories) but do not touch any live crypto path; DataPointer::reset() still OPENSSL_clear_frees. No auth, permissions, or input-validation code is weakened.
Level of scrutiny
High, for scope rather than correctness-per-line. Each deletion is compiler-verified (a stale reference would fail the debug build the author ran), and I spot-checked the highest-risk items — the OBJECT_OFFSETOF(m_worldIsNormal) (potential JIT consumer), the bindgen adapter types (potential codegen emitter), and the deleted .cpp files (potential CMake listing) — all confirmed dead. But the decision that each of 33 ncrypto methods, 19 ZigGlobalObject accessors, and 9 scripts is intentionally-abandoned rather than kept-for-upcoming-work is a maintainer judgment call this review can't make. ZigGlobalObject.h's own header warns it is included nearly everywhere.
Other factors
The PR reopens #37149 "as requested there," implying prior maintainer engagement on scope; that context isn't visible here. The ncrypto file is a port from Node's ncrypto and stripping it diverges from upstream — reasonable, but worth a maintainer nod. The nine deleted scripts reference workflow/PR history (#36778, #24201, ci:slowest, bun run node:test) that a human familiar with the tooling can confirm in seconds. No tests are added, which is standard for dead-code removal, and no existing tests are weakened.
This was referenced Aug 19, 2026
Jarred-Sumner
pushed a commit
that referenced
this pull request
Aug 21, 2026
) ### Problem - 92 hand-written trait impls (`Default`, `From`, `Clone`, `PartialEq`) across 29 crates have no caller. rustc's `dead_code` lint exempts trait impls, so the workspace lints never report them. - A few items existed only for those impls: two tables in `src/resolver/options.rs`, `Flags::DEFAULTS` in `src/md/types.rs`, and five error variants that only a removed `From` impl constructed. ### Fix - Delete the impls and the items that became unused. 80 files, 1064 lines removed. The one added line is a doc comment with its last sentence cut. - The candidates come from the linker: a relink of the debug build with `--gc-sections --print-gc-sections` lists the trait methods dropped from every codegen unit. Impls that a bound or a supertrait still needs (`Deref` under `DerefMut`, `PartialEq` under `Ord`, `From<RGBA> for SRGB`) do not compile without the impl, so they stay. - Verified: `bun run rust:check-all` (11 targets), `bun run rust:clippy`, `cargo build --tests` for the 21 touched crates with unit tests, `cargo miri test` for the touched miri crates, `bun bd`, and 2112 tests across css, md, shell, env, resolve, node:fs, console.table, archive, http2, build errors, install, and compile sourcemaps. ### Background - To rustc a trait impl is always reachable: generic code could call it through the trait. Only a whole-program view shows which impls nothing calls. - The debug build uses `-ffunction-sections` at opt-level 0, so a function that nothing references is its own section and is not inlined away. `--gc-sections` drops exactly those sections. A method that survives in any codegen unit (an `#[inline]` copy, for example) counts as live. - The removed impls are listed below. <details><summary>Removed impls by crate</summary> - `src/runtime` (19): `Default for Chmod`; `Default for ExecCfg`; `Default for FChmod`; `Default for FetchOptions`; `Default for FrameHeader`; `Default for Framework`; `Default for FromJSOptions`; `Default for GzipOptions`; `Default for ID`; `Default for InitOptions<'a>`; `Default for IntoArray`; `Default for JSS3Error`; `Default for MkdirTemp`; `Default for Open`; `Default for ReactFastRefresh`; `Default for RmDir`; `Default for Stat`; `From<CreatePtyError> for crate::Error`; `From<InitError> for crate::Error` - `src/react_compiler` (10): `Clone for GlobalRegistry`; `Clone for ShapeRegistry`; `Default for Environment`; `Default for ExhaustiveEffectDepsMode`; `Default for ScopeBlockTraversal`; `From<&str> for JsString`; `From<String> for JsString`; `From<f64> for FloatValue`; `PartialEq<&str> for JsString`; `PartialEq<str> for JsString` - `src/ast` (8): `Default for ClassStaticBlock`; `Default for CommonJSNamedExport`; `Default for Dependency`; `Default for MetadataResolve`; `Default for NamedImport`; `Default for Query`; `From<SetError> for crate::Error`; `PartialEq for Data` - `src/jsc` (6): `Default for BuildMessage`; `Default for CPUProfilerConfig`; `Default for Column`; `Default for ResolveMessage`; `Default for RuntimeTranspilerCache`; `Default for Store` - `src/resolver` (6): `Clone for DependencyMap`; `Default for BundleOptions`; `Default for EntryCache`; `Default for ExtensionOrder`; `Default for ExtensionOrderGroup`; `Default for Package<'_>` - `src/bun_core` (4): `Clone for SmolStr`; `Default for CtorOptions`; `Default for DeserOpts`; `From<crate::Error> for JsError` - `src/install` (4): `Default for Entry`; `Default for MoreInstructions<'_>`; `Default for Scratch`; `PartialEq<crate::Error> for ForManifestError` - `src/css` (3): `PartialEq for CustomPropertyName`; `PartialEq for KeyframesName`; `PartialEq for LayerName` - `src/md` (3): `Default for BlockHeader`; `Default for Flags`; `Default for Theme<'a>` - `src/parsers` (3): `Default for JSONOptions`; `From<AddToLogError> for crate::Error`; `From<bun_alloc::AllocError> for PErr` - `src/uws_sys` (3): `Default for ListenConfig`; `Default for VTable`; `Default for us_socket_stream_buffer_t` - `src/bun_alloc` (2): `Default for MaxHeapAllocator`; `Default for Mutex` - `src/bundler` (2): `Clone for OutputFile`; `Default for BunLogOptions` - `src/install_types` (2): `Default for TarballInfo`; `From<CreateMatcherError> for FromExprError` - `src/js_parser` (2): `Default for ReactFastRefresh`; `Default for ThenCatchChain` - `src/sourcemap` (2): `Default for Chunk`; `Default for Mapping` - `src/base64` (1): `Default for VLQ` - `src/collections` (1): `Default for StringMap` - `src/dns` (1): `Default for Backend` - `src/dotenv` (1): `Default for Map` - `src/io` (1): `Default for FilePoll` - `src/paths` (1): `From<Error> for crate::Error` - `src/picohttp` (1): `Default for Response<'a>` - `src/ptr` (1): `Default for WeakPtrData` - `src/resolve_builtins` (1): `Default for Alias` - `src/s3_signing` (1): `Default for SignQueryOptions` - `src/sql_jsc` (1): `From<FlushQueueError> for crate::Error` - `src/sys` (1): `Default for Address` - `src/threading` (1): `Default for ThreadPool` Follow-up removals that became unused with them: - `src/resolver/options.rs`: the `EXTENSION_ORDER` / `MODULE_EXTENSION_ORDER` copies and the `TargetMainFields` / `DEFAULT_MAIN_FIELDS_*` tables (only `Default for BundleOptions` read them; the bundler projects these options itself). - `src/md/types.rs`: `Flags::DEFAULTS` (only `Default for Flags` read it; `root.rs` builds `Flags` field by field). - `src/parsers/xml.rs`: the `PErr::Oom` variant and its match arm (only the removed `From<AllocError>` produced it). - `src/css/rules/keyframes.rs`, `src/css/rules/layer.rs`: the `Hash` and `Eq` impls that went with the removed `PartialEq` impls, and the two comments that described them. The `ShapeRegistry` and `GlobalRegistry` docs in `src/react_compiler/hir/` lose the sentence about cloning for the same reason. Neither type is used as a map key; `LayerName` comparisons go through the inherent `eql`. - `src/ast/stmt.rs`: `impl Eq for Data`. - Error variants whose only constructor was a removed `From` impl, with their `name()` arms: `bun_runtime::Error::TerminalInit`, `bun_ast::Error::Clobber`, `bun_paths::Error::MaxPathExceeded`, `bun_sql_jsc::Error::AuthenticationFailed`. The inner error types (`InitError`, `SetError`, `path_options::Error`, `FlushQueueError`) are still live on their own. </details> <details><summary>Notes</summary> - `bun_core::strings::rsplit_once` has no caller, but `clippy.toml` names it as the replacement for `str::rsplit_once`, `str::rsplitn` and `bstr::ByteSlice::rsplit_once_str`, so it stays. Policy files are one kind of reference the linker does not see. No other removed item is named in `clippy.toml`, `hawk.toml`, or the docs. - Three impls came back during verification and are not in the diff: `Semaphore: Default` (used by the macOS `fs_events.rs`, found by `rust:check-all`), `clap::Help: Default` and `braces::Token: PartialEq` (used only by unit tests, found by `cargo build --tests`). The cross-target and test builds are needed for this kind of sweep: the linux binary alone does not see them. Scope of this sweep and what came back clean, so the next run can skip it: - Seven dead-code PRs are open right now (#37181, #37659, #39581, #39582, #39618, #39697, #39732). This PR deletes no line that any of them deletes (checked by diffing the removed lines). `src/jsc/headergen/sizegen.cpp` is dead too, but #39618 already removes it. - Flipping every `pub` to `pub(crate)` in `bun_runtime` (349k lines) and letting rustc report leaves 53 items, nearly all Windows-only or FFI enum tags. The same two-round flip over the other 95 crates found only macro-generated tables and items that are used from other platforms or from macro bodies. - The gc-sections list for C++ is almost entirely covered by the open PRs. What is left is small: 15 `JSFoo::toWrapped` stubs, a few `ncrypto` helpers, a handful of unused overloads. - `src/js/internal/**` (111 files) and the remaining `src/js/node`, `builtins`, `bun`, `thirdparty` files (86 files) were scanned export by export: nothing dead beyond a few single lines (`WASI.prototype.getState/setState`, three unused WASI constants). - No orphan files: every `.rs` is mounted, every `.cpp` is compiled, every `.h` is included. No stale `cfg` names (the build emits no `unexpected_cfgs` warnings). - Left alone on purpose: `src/runtime/api/bun/h2/` (in-progress rewrite, `#![allow(dead_code)]` by design), `src/react_compiler/compile_result.rs` (marked as not yet wired), the MySQL `CharacterSet` collation table (222 unused constants, a code table like the ones `hawk.toml` keeps), `Sys`/`Alloc` error variants that follow the crate `Error` convention, and the `JsPoster` clone vtable slot. - Possible larger follow-up for a maintainer decision: `PerformanceResourceTiming::create` and the `ResourceTiming` / `NetworkLoadMetrics` / `PerformanceServerTiming` constructors are unreferenced, so no such entry can ever exist. The JS classes stay reachable as globals, which is why about 1.8k lines of `src/jsc/bindings/webcore/*Timing*` survive. Removing them changes the global surface, so it is not in this PR. </details> --------- Co-authored-by: autofix-ci[bot] <114827586+autofix-ci[bot]@users.noreply.github.com>
robobun
force-pushed
the
claude/farm/99b45785/dead-code-cpp-scripts-ast
branch
from
August 21, 2026 01:16
f349ece to
1c66d5a
Compare
Collaborator
Author
|
Updated 11:53 PM PT - Aug 20th, 2026
❌ @robobun, your commit ed53499 has 2 failures in
🧪 To try this PR locally: bunx bun-pr 39581That installs a local version of the PR into your bun-39581 --bun |
![github-actions[bot]](https://avatars.githubusercontent.com/in/15368?s=60&v=4){kind=small}
Collaborator
Author
|
Current head is ed53499, rebased onto main, full debug build passes. The red CI on the Windows 2019 lane comes from two failures that reproduce on unmodified main and are reported for triage: the bake/deinitialization.test.ts segfault (4/10 on a main debug build on the same Windows version) and bun-server.test.ts (tagged pre-existing by the CI scraper). No failure in the run touches this diff. |
robobun
force-pushed
the
claude/farm/99b45785/dead-code-cpp-scripts-ast
branch
2 times, most recently
from
August 21, 2026 05:30
f365cfa to
09db9f8
Compare
Contributor
There was a problem hiding this comment.
Thanks for addressing the JSGlobalProxyInlines.h include — confirmed removed. Correcting my earlier sibling claim: webcrypto/JSCryptoKey.h is not orphaned (line ~1182 WEBCORE_GENERATED_CONSTRUCTOR_GETTER(CryptoKey) still consumes it), so it should stay. The remaining inline nit on JSBuffer.cpp is the only new item this pass.
Extended reasoning...
The prior review flagged two orphaned includes in ZigGlobalObject.cpp. Re-checking against the current diff: JSGlobalProxyInlines.h is now deleted (addressed), and JSCryptoKey.h turns out to still have a consumer — WEBCORE_GENERATED_CONSTRUCTOR_GETTER(CryptoKey) — so that half of the earlier comment was incorrect and the include should remain. Verifier agents independently reached the same conclusion this run. Noting this so the author does not act on the stale suggestion.
…cripts Every deleted symbol was verified to have zero references across src/, scripts/, packages/, and freshly regenerated build/debug/codegen/ output, then validated with a full debug build. C++ (src/jsc/bindings): - ncrypto.cpp/h: 33 uncalled methods from the ncrypto port (bytesToKey, FIPS toggles, secure-heap helpers, X509 time/usages getters, unused AES CTR/GCM/KW factories, BIO/Bignum constructors), the 5 named Digest factories orphaned by the bytesToKey removal (only Digest::FromName has callers), and the DataPointer secure_ field plus constructor parameters that SecureAlloc was the sole producer for - DOMJIT IDL helper headers (DOMJITIDLTypeFilter.h, DOMJITIDLType.h, DOMJITIDLConvert.h, DOMJITHelpers.h/.cpp): zero instantiations of any template they define; include sites now pull FrameTracers.h directly where JITOperationPrologueCallFrameTracer is used - 13 empty stub files (MessagePortChannel*, MessagePortIdentifier.h, BroadcastChannelRegistry.h, JSDOMConvertSerializedScriptValue.h, JSDOMBuiltinConstructorBase.*) kept by an earlier cleanup only for its own verification mechanics - ZigGlobalObject: 19 accessors with zero callers (8 *SinkPrototype() getters, worldIsNormal, bunStdin/bunStderr/bunStdout, KeyObjectStructure, JSBufferStructure, JSCryptoKeyStructure, builtinInternalFunctions, performMicrotaskVariadicFunction, globalProxyStructure), the backing fields and initLater closures this orphaned, and jsFunctionPerformMicrotaskVariadic - BunString isCrossThreadShareable, NodeValidator validateArrayBufferView, JSDOMConvertBase convertResult, JSDOMConvertSequences reserveEstimated, EventNames isGamepadEventType (declaration with no definition), ExceptionDetails EnumTraits, ResourceTiming overrideInitiatorType, WebSocket hasNativeCallbacks, PerformanceObserverCallback hasCallback (never-invoked virtual pair), ScriptExecutionContext isDocument/ isWorkerGlobalScope - io_darwin.cpp: io_darwin_close_machport (both cfg arms); unlike its siblings it has no Rust-side declaration and no caller anywhere Rust: - jsc/bindgen.rs: BindgenOptionalRepr trait, BindgenOptionalCustom, BindgenExternalShared and their impls; the bindgen codegen emitters never name them. Also the Optional::adopt helper and the ExternalSharedOptional type (bun_core/external_shared.rs plus its two re-exports) that this orphaned. - ast/binding.rs: BindingInit trait and Binding::init; all construction goes through Binding::alloc or literal struct init. Two doc comments that pointed at Binding::init are updated. - ast/runtime.rs: write-only ImportsIteratorEntry.value field and two commented-out leftover fields - runtime/webcore.rs: two unused flat re-exports - ini/lib.rs: commented-out Zig code from the port Scripts (invoker .github/workflows/labeled.yml.disabled was deleted in #36778, or superseded): - label-issue.ts, read-issue.ts, handle-crash-patterns.ts, is-outdated.ts, associate-issue-with-sentry.ts - nav2readme.ts (imports docs/nav, deleted in #24201; cannot run) - buildkite-slow-tests.js (superseded by ci:slowest and update-test-durations) - check-node.sh, check-node-all.sh (superseded by bun run node:test)
robobun
force-pushed
the
claude/farm/99b45785/dead-code-cpp-scripts-ast
branch
from
August 21, 2026 06:02
09db9f8 to
ed53499
Compare
Contributor
There was a problem hiding this comment.
Both earlier inline notes (orphaned JSGlobalProxyInlines.h include, unnecessary FrameTracers.h in JSBuffer.cpp) are now addressed, and this pass found no further issues. Given the breadth — 55 files including ZigGlobalObject.{h,cpp} lazy-property tables, the DataPointer constructor signature change in ncrypto, and bindgen trait removal — a maintainer sign-off is still worthwhile.
What was reviewed:
- Confirmed
webcrypto/JSCryptoKey.hinclude is still live (WEBCORE_GENERATED_CONSTRUCTOR_GETTER(CryptoKey)), so leaving it was correct. - Confirmed
kMinCrossThreadShareableLengthremains used bytoCrossThreadShareableafterisCrossThreadShareableremoval. - Checked that each removed
FOR_EACH_GLOBALOBJECT_GC_MEMBERentry has its matching field, accessor, and initLater table entry removed together.
Extended reasoning...
Overview
This PR deletes dead code across 55 files: ~33 uncalled ncrypto methods and the DataPointer secure_ field/ctor param, 5 DOMJIT IDL helper headers, 13 empty stub .cpp files, ~19 unused ZigGlobalObject accessors plus 3 backing lazy-property fields (m_cachedGlobalProxyStructure, m_JSCryptoKey, m_performMicrotaskVariadicFunction) and jsFunctionPerformMicrotaskVariadic, assorted single-function removals in webcore headers, Rust bindgen adapter traits (BindgenOptionalRepr/BindgenOptionalCustom/BindgenExternalShared) and the ExternalSharedOptional type they orphaned, an unused AST BindingInit trait, a write-only iterator field, and 9 orphaned scripts whose invoking workflow was deleted.
Security risks
None identified. The ncrypto deletions remove uncalled wrapper methods (FIPS toggles, secure-heap helpers, unused cipher factories) — no live crypto path changes behavior. DataPointer::reset now unconditionally calls OPENSSL_clear_free since the only producer of secure_=true (SecureAlloc) is deleted; under BoringSSL both branches were already equivalent.
Level of scrutiny
High, because of surface area rather than logic complexity. Each hunk is a mechanical deletion whose correctness reduces to "zero callers on current main" plus "still builds", but the deletions span ZigGlobalObject's GC-visited member table, the class-codegen include block, and FFI-adjacent Rust traits — files where a missed reference fails at link time or worse. The author states a full debug build passes and repo-wide greps return only deletion sites; CI build #102376 is the ground truth.
Other factors
I left two inline nits on earlier revisions; both are addressed in the current diff (JSGlobalProxyInlines.h include dropped; FrameTracers.h no longer added to JSBuffer.cpp). One candidate this run — JSCryptoKey.h orphaned — was verified false: WEBCORE_GENERATED_CONSTRUCTOR_GETTER(CryptoKey) at ZigGlobalObject.cpp:1182 still consumes it. The comment-cop bot flags on binding.rs/p.rs/bindgen.rs were resolved as false positives (section-banner comments being shortened, not workaround justifications). No tests are added by design; this is deletion-only and the source-lints doc forbids dead-symbol pin tests. Given the file count and that this touches the GlobalObject GC-member macro, deferring to a maintainer rather than auto-approving.
This was referenced Aug 21, 2026
Jarred-Sumner
pushed a commit
that referenced
this pull request
Aug 23, 2026
…JS, the bake dev server assets, the shell, and the SQL crates (#40066) Scheduled dead-code sweep. Net -3275 lines (78 files, +112 / -3387; the source side is +32 / -3387, the rest is the test pin). No behavior change: every item below has zero references on `main` in `src/`, `packages/`, `scripts/`, `test/` and the regenerated `build/debug/codegen/`, and the build passes without it. About 2,350 of these lines carry over the removals from #37659, which was closed only for merge conflicts. Each of those was re-verified against current `main` (nothing gained a caller), and the diff applies to current `main` with no conflicts. The rest comes from fresh scans of the shell, SQL, bake and built-in JS areas, which no open dead-code PR covers. ### bun-uws (`packages/bun-uws/src`, -1168) Bun reaches uWS only through the C shim in `src/uws_sys/` and eight bindings files, so those includers are the complete set of callers. Checked against every one of them: * Whole headers nothing includes: `ClientApp.h`, `HttpError.h` (superseded by `HttpErrors.h`), `Multipart.h` and `MessageParser.h` (its only includer), `ProxyParser.h` (only reachable under `UWS_WITH_PROXY`, which no build defines; `grep UWS_ build/debug/compile_commands.json` is empty), plus every `#ifdef UWS_WITH_PROXY` block and the `reserved` / `proxyParser` parameters that only carried the proxy parser into `getHeaders()`. * `WebSocketBehavior::subscription` and `WebSocketContextData::subscriptionHandler`: the only place a `WebSocketBehavior` is built (`uws_ws()` in the shim) never sets it, so every dispatch site in `WebSocket::end/subscribe/unsubscribe` and `WebSocketContext::onClose` was unreachable. `TopicTree::unsubscribe` drops the `newCount` only those sites read. * `HttpRequest`: the per-route `std::map` of parameter offsets (`setParameterOffsets`, `getParameter(name)`; every caller uses the index overload), `getQuery()` with no key, `isShortRead`, `notFieldNameWord` / `hasMore` / `hasBetween`. * `HttpContext::HTTP_IDLE_TIMEOUT_S`, `HttpFlags::isAuthorized` (the live flag is the per-socket one in `AsyncSocketData`), `HttpResponse::getHttpResponseDataS`, `HttpResponse::endWithoutBody`, `HttpResponse::overrideWriteOffset` (the shim calls `setWriteOffset`). * `App::publish(unsigned char)`, two `App::listen` overloads, the `UWS_NO_ZLIB` block; two `H3App::listen` overloads; `Http3Request::isAncient` / `getCaseSensitiveMethod`; nine `Http3Response` members whose `uws_h3_res_*` shims are stubs; the write-only `Http3ResponseData::totalSize`; `Loop::integrate`, `Loop::setSilent` and `LoopData::noMark`; the `UWS_NO_ZLIB` / `UWS_MOCK_ZLIB` mock streams and `UWS_USE_LIBDEFLATE` guards in `PerMessageDeflate.h` (the macro is defined unconditionally three lines above its first use). * `HttpContext::layoutAssert()` and `WebSocketContext::layoutAssert()` were never called, and as static members of class templates their `static_assert`s never ran. The `HttpContext` one guarded nothing. The `WebSocketContext` one guards the layout `WebSocket::getContextData()` depends on (`group` first, `data` right after), so it becomes a `RELEASE_ASSERT` in `create()`: `offsetof` cannot express it because the struct is not standard-layout, and the layout can differ per ABI, so the check runs once per context in every build. * `misc/` (upstream README assets, demo `cert.pem` / `key.pem`), and the `.gitattributes`, `.dockerignore`, `.cursorignore` entries for a `fuzzing/` directory that no longer exists. ### Native `JSBufferList` (`src/jsc/bindings`, -641) `JSBufferList.cpp` / `.h`, the lazy class structure on `GlobalObject`, its three accessors, and the `DOMIsoSubspaces` / `DOMClientIsoSubspaces` slots. The JS streams implementation replaced it; the only remaining mention of "BufferList" in the tree is a comment in `internal/streams/readable.ts`. The sources are globbed, so no build-script change is needed. ### Built-in JS (`src/js`, -508) * `internal/util/inspect.d.ts` (ambient types for `node-inspect-extracted`, which nothing imports) and `src/js/.gitignore` (ignores `src/js/out`, which codegen has not written since the build moved under `build/`). * `private.d.ts`: the `BunFS` / `BunFSWatcher` types (`Bun.fs()` no longer exists), `Bun.TOML`, `Bun.tty`. * `QuicStream.prototype[kSendHeaders]` and its symbol, the unused `get [kVerifyPeer]` (the setter stays; `session[kVerifyPeer] = …` writes it), `colors.ts` `yellow` / `gray` / `clear` / `reset`, the runtime `SSLMode` export of `internal/sql/shared.ts` (only imported as a type), the `REPL_MODE_*` re-exports of `internal/repl/utils.js` (`repl.js` takes them from `internal/repl/mode`), `Dequeue#clear()`, the write-only `listeningId` in `net.ts`, and the `kGetNativeReadableProto` symbol in `internal/shared.ts` (its last reader went in 032713c). ### Bake (`src/runtime/bake`, -830) * `incremental_visualizer.html` and `memory_visualizer.html`: #36184 removed the `runtime_embed_file!` sites and the `/_bun/incremental_visualizer` route that served them, so the two pages are orphaned assets. The no-op `emit_*_visualizer_*` methods that #36184 left behind are still reachable and are not touched here. * `client/websocket.ts` `getMainWebSocket()` (exported, never imported) and a commented-out Zig `writeJsValue` in `serialized_failure.rs`. ### Shell (`src/runtime/shell`, `src/shell_parser`, -51) * `Base::end_scope()`, an empty function, and its 12 call sites in `states/*.rs`. `Script::deinit_from_interpreter` only called it and goes too. * `ShellErr::InvalidArguments`: never constructed (only matched), with its four match arms. * `bun_shell_parser` dependencies `scopeguard`, `const_format`, `enum-map`, `enumset`, `typed-arena`, `bun_collections`: no `use` or path reference in the crate. ### SQL (`src/sql`, `src/sql_jsc`, -22) and `hawk.toml` (-104) * `bun_sql` dependencies `const_format`, `enum-map`, `enumset`, `libc`; `bun_sql_jsc` dependencies `const_format`, `enum-map`, `enumset`, `libc`, `bun_uws_sys`, and a commented-out `bun_runtime` edge that would be a dependency cycle (`bun_runtime` depends on `bun_sql_jsc`). * `pub use` re-exports nothing imports: `ArrayBuffer`, `JSCell`, `host_fn` in `sql_jsc/jsc.rs` (the proc-macro is used by its absolute path), `mysql::{MySQLConnection, MySQLQuery, MySQLStatement}` (every consumer imports the leaf module), `postgres::SASL`. * `hawk.toml`: 13 overrides for `bun_sql::mysql::status_flags::StatusFlag::*` variants that #37229 removed; `StatusFlag` has one variant left. ### Rust (-18) `zig__ModuleInfo__destroy` and `zig_log` in `src/bundler/analyze_transpiled_module.rs`: `extern "C"` exports that no C++ declares or calls (`BunAnalyzeTranspiledModule.cpp` frees through `zig__ModuleInfoDeserialized__deinit`), with no hit in `vendor/WebKit` either. ### Verification * `rg` for every symbol across `src/`, `packages/`, `scripts/`, `test/`, `docs/` and `build/debug/codegen/`; `.classes.ts` files, `src/codegen/`, string-named lookups (`$getByIdDirectPrivate(this, "…")`, `$zig` / `$cpp`) and token-paste macros (`name##PublicName`) were checked by hand. Two candidates failed those checks during the run and stay: `BunBuiltinNames` `mockedFunction` (reached through `BUN_COMMON_STRINGS_EACH_NAME`) and `writer` (reached through `$getByIdDirectPrivate(this, "writer")` in `ConsoleObject.ts`). * `bun bd` builds. `cargo check --workspace` passes on all 12 target triples. `cargo fmt --check`, prettier, oxlint and clang-format are clean. `tsc -p src/js` reports the same errors before and after (only line numbers move). * `test/js/bun/websocket/websocket-server.test.ts` gains a pin for the pub/sub paths this PR edits: with two sockets on shared topics, `unsubscribe()` of a topic the socket never joined (or that does not exist) reports false and changes nothing, leaving the last topic frees the subscriber and a later `subscribe()` counts again, and `close()` drops the socket from every topic without touching the other socket. Because this PR only deletes unreachable code, the pin passes before and after by design. * `bun bd test`: `bun/websocket/websocket-server.test.ts`, `web/websocket/websocket-permessage-deflate.test.ts`, `bun/http/bun-serve-routes.test.ts`, `bun/http/serve-http3.test.ts`, `node/http/node-http.test.ts`, `node/quic/quic-stream.test.ts`, `node/net/node-net-server.test.ts`, `bun/util/inspect.test.js`, six `bun/shell/*.test.ts` files, `sql/sql-mysql*.test.ts`, `sql/sql-mariadb-json.test.ts`, `sql/postgres-binary-numeric.test.ts`, `sql/sql-connect-error-reporting.test.ts`, `node/stream/node-stream.test.js`, `node/console/console-table-iterators.test.ts`, `web/console/console-log.test.ts`, `internal/fifo.test.ts`, `bake/dev/esm.test.ts`, `bake/dev/hot.test.ts`, `bun/repl/repl.test.ts` (mode tests), `node/test/parallel/test-repl-definecommand.js`. Details on the local failures that also fail on the released bun are under Notes. ### Looked at and left alone * `src/runtime/api/bun/h2/`: the outbound half (`Connection::send_header_block` / `send_data` / `send_push_promise`, `hpack::Coder::encode`, `SendWindow::available`) is only exercised by its unit tests. The module doc describes it as the engine that will replace `h2_frame_parser.rs`, so it is in progress, not dead. * The bake visualizer protocol residue: `MessageId::Visualizer`, the `IncrementalVisualizer` topic, and the empty `emit_*_visualizer_*` methods and their timer. All reachable; removing them is a protocol change that also touches the generated `generated.ts` ids. * `ColumnDefinition41::{fixed_length_fields_length, decimals}` are decoded and never read, but they mirror the MySQL wire struct; left in place. * `bun_sql::mysql::StatusFlags`'s `Display` impl prints nothing, but a debug log in `MySQLConnection.rs` still formats it. <details><summary>Notes</summary> How the candidates were found: the dead-code PRs that were closed for merge conflicts were re-applied against current `main` and every removal re-verified (#37659 applied with no conflicts; its `assertion_error.ts` and `util.ts` hunks had already landed via #39924 and #39628 and are not in this PR). Read-only scans of `src/runtime/shell` + `src/shell_parser`, `src/sql` + `src/sql_jsc` + `src/runtime/node`, `src/runtime/bake` + `src/runtime/api` + `src/runtime/image`, and `src/js` + `src/jsc/modules` produced the fresh items. `src/runtime/node`, `src/runtime/api` (outside `h2/`), `src/runtime/image` and `src/jsc/modules` came back clean. Local test failures, all reproduced with `USE_SYSTEM_BUN=1` (the released bun) in the same container, so none is caused by this diff: * `websocket-server.test.ts`: eight `it.concurrent` cases that overlap the 300,000-message `(benchmark)` test time out under the debug ASAN build in this container (the whole file takes 64 s here against a 6.4 s CI budget); each passes when run with `-t`. * `node-http.test.ts` "request via http proxy": `ECONNREFUSED` from `localhost` resolution in this container. * `bunshell.test.ts` `-c`, `-f character device`, and the seven "stdin redirect from a zero-length buffer" cases: this container's `/dev/null` is a regular file, not a character device. * `sql-mysql-binary-null-indexed.test.ts`: `Access denied for user 'root'@'localhost'` against the local MariaDB. * `fifo.test.ts` "pushing and shifting a lot of items": a 10 s perf budget exceeded under ASAN; the only `fifo.ts` change is the removal of the unused `clear()` method. Files shared with open PRs, different hunks: `src/jsc/bindings/ZigGlobalObject.cpp` / `.h` (#39581 removes other accessors; its `HTMLRewriterSinkPrototype()` hunk sits one blank line above the `JSBufferList` accessors removed here), `Cargo.lock` (#39618 edits other package blocks). </details> <!-- robobun:evidence:begin --> --- **no test proof** · iteration 4 · platform-specific test(s) that do not run on this machine, deferring to CI, which covers all platforms: test/js/bun/websocket/websocket-server.test.ts <!-- robobun:evidence:end -->
Jarred-Sumner
deleted the
claude/farm/99b45785/dead-code-cpp-scripts-ast
branch
August 23, 2026 06:42
robobun
added a commit
that referenced
this pull request
Aug 23, 2026
robobun
added a commit
that referenced
this pull request
Aug 23, 2026
binding.rs: #39581 removed Binding::init, which nothing called here either; Binding::alloc keeps taking impl Into<Option<Loc>>.
This was referenced Aug 23, 2026
robobun
added a commit
that referenced
this pull request
Aug 31, 2026
Both slow-test parsers (scripts/ci-slowest-tests.ts and scripts/update-test-durations.mjs) measured a file as the gap between its `[N/M] <path>` header and the next header they recognized. Every other header runner.node.mjs emits (`--- napi prebuild:`, the `--- [A-B/M] K files in parallel` bucket, `--- Running N parallel-safe`, `--- End`, `--- Summary`, `--- Received <signal>, exiting`) was invisible, so the serial test that happened to run before one of them was charged the whole phase that followed. - Those phase headers close the open span. The check is an allowlist of the titles the runner emits, not any `--- ` line, because streamed test output can still contain `--- a/<file>` or `--- ps ---`. - `[N/M] <path>` matches with or without the `--- ` prefix. Bucket summary lines `[N/M] <path> (X.XXs)` use the inline timing. Concurrent dispatch spans are clamped to 500 ms. - Retry/error labels (`<path> - code 1`) are a boundary that opens no span, so the retry backoff lands on neither attempt. - A truncated log (no `--- End`) charges the open span to the last timestamp seen instead of dropping it. Both scripts export parseLog behind an entrypoint guard, and test/internal/ci-slowest-tests.test.ts exercises them against the same header fixtures. scripts/buildkite-slow-tests.js from #35418 is gone from main (#39581) and is not restored. Regenerated test/expected-durations.json with the combined parser from builds 108915, 108911, 108910, 108909, 108906.
Jarred-Sumner
pushed a commit
that referenced
this pull request
Sep 1, 2026
…sers, regenerate the table (#41070) ### Problem - `windows-aarch64-11-test-bun` packs its 8 shards with the timings of the windows x64 lane. Its serial files run about 2x slower and not uniformly (`bundler/bundler_compile.test.ts` is 222s there, 48s in the x64 column). One shard carries 922s of work against a 610s mean. This lane finished last in 46 of the 47 completed builds since #40993 merged, so it sets the build time: the median build went from 1342s to 1386s while the linux lanes got faster (x64 glibc slowest shard 233s to 178s). - The slow-test log parsers (`scripts/update-test-durations.mjs`, `scripts/ci-slowest-tests.ts`) only recognize `[N/M] <path>` headers. Every other header the runner emits (`--- napi prebuild:`, the `--- [A-B/M] K files in parallel` bucket, `--- Running N parallel-safe`) is invisible, so the serial file that runs right before one is charged the whole phase that follows. In source build 108642 this hit 19 of 20 linux shards and 8 of 8 windows shards (`js/valkey/unit/basic-operations.test.ts` recorded as 111.6s on windows, real 0.1s). 11 of the phantom entries are 10s or more in the default column, so `--skip-slower-than=10000` dropped those files from the darwin PR lanes. ### Fix - Add a `windows-aarch64` column, measured from `windows-aarch64-11-test-bun`. `runner.node.mjs` picks it for the windows-aarch64 step and falls back to `windows`, then the other columns, for a file that lacks one. - Fold in the parser fixes from #35418 (this PR supersedes it): the runner's phase headers close the open span (an allowlist of the titles the runner emits, since test output can contain `--- a/<file>`), bucket summary lines use their inline `(X.XXs)` timing, retry labels open no span, and a truncated log keeps its last open span. Both scripts export `parseLog` behind an entrypoint guard, and `test/internal/ci-slowest-tests.test.ts` runs them against the same fixtures. `scripts/buildkite-slow-tests.js` from #35418 was deleted from main in #39581 and is not restored. - Regenerate `test/expected-durations.json` with the combined parser from builds 108915, 108911, 108910, 108909, 108906 (`node scripts/update-test-durations.mjs --builds 5`, 5906 entries). - Verified: `bun bd test test/internal/ci-slowest-tests.test.ts` (18 pass). Replaying the LPT packing against per-file costs measured on three other windows-aarch64 runs (builds 108785, 108786, 108801) gives a slowest shard of 673s with the new column, against 922s with the table on main (610s is the floor). Column selection checked for every step key. ### Background - `runner.node.mjs` bin-packs test files across `--max-shards` with LPT (longest file first into the least loaded bin) using `test/expected-durations.json`. Every shard computes the same assignment from the same input, so the shards need no coordination. - The table has one column per lane with distinct timing: `default` (linux x64 debian), `asan`, `musl`, `windows`. Every other lane uses the nearest column (darwin uses `default`). - Per-file cost comes from the `_bk;t=<ms>` timestamps Buildkite prefixes to each log line: the gap between one `[N/M] <path>` group header and the next header. The parallel bucket runs one `bun test --parallel` for many files and prints `[N/M] <path> (X.XXs)` per file afterwards. - The untiered darwin PR lanes pass `--skip-slower-than=10000`, which drops a file when its column cost is 10s or more. - `.buildkite/update-test-durations.yml` regenerates the file on a schedule and uploads it as an artifact. A human commits it. <details><summary>Notes</summary> Measurement: Buildkite REST job timings for 95 builds created after the merge of #40993 (09:17 UTC, 2026-08-31). 64 builds include the new table (GitHub compare status `ahead`), 31 PR builds still ran on the old one. Same fleet, same hours. Median of the slowest shard per lane, old table to new table: | lane | old | new | |---|---|---| | linux-x64-debian-13 | 233s | 178s | | linux-x64-asan | 437s | 387s | | linux-aarch64-debian-13 | 255s | 230s | | linux-x64-ubuntu-2504 | 232s | 178s | | musl (both arches) | 271-282s | 281-294s | | windows-x64-2019 | 491s | 468s | | windows-aarch64-11 | 829s | 887s | | darwin-aarch64-any | 534s | 581s (2 shards, balanced per shard: 506s and 518s) | | build start to last test finished | 1342s | 1386s | windows-aarch64 per-shard medians with the old table: `456 523 594 572 477 375 823 526` (shard 6 slowest in 177 of 188 builds). With the table from #40993: `447 404 575 892 521 651 416 453` (shard 3 slowest in 49 of 49). The runner estimated that shard at 429s. Largest windows-aarch64 costs in the new column (s): bundler_compile 219, serve-body-leak 154, expo 114, v8 97, test-fs-read-stream-pos 90, napi 85. Bucket header bug, source build 108642, default column (parsed vs real): websocket-proxy-tunnel-client-leak 17.6 vs 0.0, valkey/integration/complex-operations 32.8 vs 0.1, regression/issue/24850 15.7 vs 0.1, sql-mysql-bigint-out-of-range 24.4 vs 0.3. Windows column: valkey/unit/basic-operations 122.5 vs 0.1, regression/issue/28632 76.3 vs 0.1. In the logs checked, the `--- napi prebuild:` header carries the same timestamp as the next file header, so it misattributes nothing today; the allowlist still covers it. From #35418, against build 86086 logs: `test/js/bun/test/fake-timers/sinonjs/issue-2086.test.ts` parsed at 22554 ms before, 57 ms after (bun's own timer: 41 ms). `test/regression/issue/24850.test.ts` 60575 ms before, 61 ms after. Unique files found in one windows shard: 63 before, 829 after (the old `ci-slowest-tests.ts` regex required a literal `[90m` SGR and the `--- ` prefix). Side effect: `scripts/update-parallel-allowlist.mjs` takes the slowest column per file. With the new column, 10 files that are under 15s on every other lane are over 15s on windows-aarch64 and will leave the parallel allowlist at its next regeneration (for example `js/bun/spawn/spawn-noread-leak.test.ts` at 38.7s). That is the stated intent of that check. The regenerated default column has 44 entries at 10s or more against 55 on main. Files that the darwin PR lanes skipped because of phantom entries and now run again: websocket-proxy.test.ts, websocket-proxy-tunnel-client-leak, websocket-proxy-tunnel-upgrade-leak, regression/issue/24850, regression/issue/patch-bounds-check, cli/install/bun-install-proxy. </details> <!-- robobun:evidence:begin --> --- **[stamp-90s]** gate passed · iteration 2 · 6 files touched <details><summary>passes on PR (with fix)</summary> ```console Test-only change. Debug/ASAN (expected pass): $ bun bd test 'test/internal/ci-slowest-tests.test.ts' $ BUN_DEBUG_QUIET_LOGS=1 bun scripts/build.ts --profile=debug --quiet test test/internal/ci-slowest-tests.test.ts bun test v1.4.1 (a6c4cc2) test/internal/ci-slowest-tests.test.ts: (pass) scripts/ci-slowest-tests.ts parseLog > does not charge the parallel-safe phase to the last serial test [28.64ms] (pass) scripts/ci-slowest-tests.ts parseLog > sums retry attempts and normalizes Windows path separators [6.28ms] (pass) scripts/ci-slowest-tests.ts parseLog > closes the last serial test at the parallel-phase group header when no parallel headers follow [11.47ms] (pass) scripts/ci-slowest-tests.ts parseLog > does not charge the parallel-bucket phase to the preceding serial test [12.70ms] (pass) scripts/ci-slowest-tests.ts parseLog > ignores stray `--- ` lines that are test output, not group headers [7.17ms] (pass) phase-header boundary > true --- napi prebuild: 3 addon(s), 23.9s [1.37ms] (pass) phase-header boundary > true --- [52-257/829] 206 files in parallel (3×) [0.32ms] (pass) phase-header boundary > true --- Running 444 parallel-safe tests (3-wide) [0.19ms] (pass) phase-header boundary > true --- End [0.18ms] (pass) phase-header boundary > true --- Summary [0.18ms] (pass) phase-header boundary > true --- Received SIGTERM, exiting... [0.18ms] (pass) phase-header boundary > false --- a/index.js [0.20ms] (pass) phase-header boundary > false --- ps --- [0.17ms] (pass) phase-header boundary > false --- logs --- [0.17ms] (pass) phase-header boundary > false ------ [0.18ms] (pass) phase-header boundary > false --- [0.27ms] (pass) phase-header boundary > false --- [52/829] test/a.test.ts [0.19ms] (pass) scripts/update-test-durations.mjs parseLog > does not charge napi prebuild or the parallel-bucket phase to the preceding serial test [17.04ms] 18 pass 0 fail 24 expect() calls Ran 18 tests across 1 file. [2.22s] Exit: 0 ``` </details> <details><summary>diff hotspot</summary> ``` scripts/ci-log-phase.mjs | 32 + scripts/ci-slowest-tests.ts | 282 +- scripts/runner.node.mjs | 18 +- scripts/update-test-durations.mjs | 116 +- test/expected-durations.json | 49849 +++++++++++++++++-------------- test/internal/ci-slowest-tests.test.ts | 196 + 6 files changed, 28355 insertions(+), 22138 deletions(-) ``` </details> **gate history** · 3 passed · 0 rejected · iteration 2 <details><summary>evidence per changed file</summary> ``` file reads edits tests scripts/ci-log-phase.mjs 0 1 0 scripts/ci-slowest-tests.ts 1 1 0 scripts/runner.node.mjs 4 1 0 scripts/update-test-durations.mjs 4 6 0 test/expected-durations.json 0 0 0 test/internal/ci-slowest-tests.test.ts 3 4 0 ``` </details> <!-- robobun:evidence:end -->
robobun
added a commit
that referenced
this pull request
Oct 1, 2026
The watcher thread now frees the Watcher when a dev server stops. MultiArrayList's drop is slab-only, so the owned file_path of every entry that is still in the watchlist leaked. Since #39197, CI's ASAN lane applies its exit-time leak check to test/bake/dev/css.test.ts, and that check reports the leak from append_directory_assume_capacity::<true>. impl Drop for Watcher drops the elements of the watchlist, as #30644 does. KEventWatcher.rs: main deleted io_darwin_close_machport and its non-Darwin stub in #39581. The rebase restores the Darwin definition only, so the comment on the extern block no longer claims a stub for it.
This was referenced Oct 1, 2026
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
What does this PR do?
Reopens #37149 against current main, as requested there. It deletes dead code whose last references are already gone. Every removed symbol was verified again on current main: zero references across src/, scripts/, packages/, and freshly regenerated codegen output.
C++ (
src/jsc/bindings):Digest::FromNamehas callers), and theDataPointersecure_field plus constructor parameters thatSecureAllocwas the sole producer forJITOperationPrologueCallFrameTraceris used*SinkPrototype()getters,worldIsNormal,bunStdin/bunStderr/bunStdout,KeyObjectStructure,JSBufferStructure,JSCryptoKeyStructure,builtinInternalFunctions,performMicrotaskVariadicFunction,globalProxyStructure), the backing fields and initLater closures this orphaned, andjsFunctionPerformMicrotaskVariadicisCrossThreadShareable, NodeValidatorvalidateArrayBufferView, JSDOMConvertBaseconvertResult, JSDOMConvertSequencesreserveEstimated, EventNamesisGamepadEventType(declaration with no definition), ExceptionDetailsEnumTraits, ResourceTimingoverrideInitiatorType, WebSockethasNativeCallbacks, PerformanceObserverCallbackhasCallback(never-invoked virtual pair), ScriptExecutionContextisDocument/isWorkerGlobalScopeio_darwin_close_machport(both cfg arms); unlike its siblings it has no Rust-side declaration and no caller anywhere#include <JavaScriptCore/JSPromise.h>. It got the header transitively through a unified-source neighbor that this PR deletes.Rust:
BindgenOptionalReprtrait,BindgenOptionalCustom,BindgenExternalSharedand their impls; the bindgen codegen emitters never name them. Also theOptional::adopthelper andExternalSharedOptional<T>(bun_core/external_shared.rs plus its two re-exports) that this orphaned.BindingInittrait andBinding::init; all construction goes throughBinding::allocor literal struct init. Two doc comments that pointed atBinding::initare updated.ImportsIteratorEntry.valuefield and two commented-out leftover fieldsScripts (invoker
.github/workflows/labeled.yml.disabledwas deleted in #36778, or superseded):ci:slowestand update-test-durations)bun run node:test)Differences from #37149: the port resolves one conflict in ZigGlobalObject.h (main replaced
JSDOMStructureMap m_structureswith them_domStructuresarray;m_worldIsNormalis still dead and stays deleted), adds the WorkerMessagingProxy include, and drops the dead-symbols source-lint test per test/internal/source-lints/CLAUDE.md.Rebase notes (ab55966): main moved the ZigGlobalObject lazy-property registration from per-field
initLatercalls intoOBJECT_OFFSETOFtables. The rebase re-applies the deletions in that shape: the table entries form_cachedGlobalProxyStructure,m_JSCryptoKey, andm_performMicrotaskVariadicFunctionare removed with their fields. One deletion from #37149 is dropped:m_bunStdin/m_bunStderr/m_bunStdoutare live again (BunObject.cpp reads the fields directly), so the fields and their initializers stay. Only their unused accessor methods remain deleted.Rebase notes (72ec6e2): parallel dead-code sweeps on main deleted neighbors of lines this PR deletes. Resolved by checking each symbol on current main:
BIOPointer::NewFileandX509View::enumUsagesstill have zero callers, so their removal stands, next to main's removal ofNewFp,ifRsa, andifEc. TheperformMicrotaskVariadicFunction()accessor removal stands next to main's removal ofperformanceObject().How did you verify your code works?
bun bd) on this branch at current main (6948a12)bun bd test test/js/node/crypto/crypto.key-objects.test.ts: 95 pass, 0 failbun bd test test/js/web/workers/worker.test.ts: 4 debug-ASAN timeout failures, identical on a main debug build, release bun passes 37/37This PR adds no test files on purpose. Every hunk deletes unreferenced code or keeps behavior identical (the ini hunk keeps the same expression and drops commented-out code, the iterator hunk drops a write-only field, the codegen hunk swaps deleted includes for FrameTracers.h). No test can distinguish the tree before and after this change, and test/internal/source-lints/CLAUDE.md forbids dead-symbol pin tests. The build is the check.