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
{{ message }}
Repository navigation
Remove dead code from JSDOMWrapper.h and ncrypto.cpp: two preprocessor blocks that never compile - #48067
JSDOMWrapper.h has an #ifndef RENAMED_JSDOM_GLOBAL_OBJECT block that no translation unit can compile. The header first includes ZigGlobalObject.h, which defines the macro and the same alias. The block also holds the closing brace of namespace WebCore: when it is active, clang stops with extraneous closing brace ('}').
ncrypto.cpp defines EVP_PKEY_CTX_set_dsa_paramgen_q_bits for OpenSSL older than 1.1.1e. BoringSSL reports 1.1.1g and declares that function.
Fix
Delete both blocks (5 and 11 lines). ZigGlobalObject.h loses the 3 guard lines around its alias: no other file names the macro.
Correct because the compiler sees the same text. The preprocessed output of ncrypto.cpp and of 40 units that read JSDOMWrapper.h (654 MB) differs only in the __LINE__ of assertions.
Verified: bun bd, then 457 tests in 6 files (node-crypto, web-globals, globals, abort, message-channel, broadcast-channel).
Behaviour change: none
Background
JSDOMGlobalObject is the WebCore name for the global object. In Bun it is an alias of Zig::GlobalObject. The macro is a guard from 2022 that let either header declare the alias first.
ncrypto is the OpenSSL wrapper of node:crypto. Bun links BoringSSL on every platform.
Considered a larger cut: 101 more dead lines in bun_libuv_sys and bun_uws_sys. Remove libuv on Windows #47776 deletes that code, so they stay out.
Downsides
None found for users. The macro has no hit in generated code, the code generators or the WebKit headers.
A failed assertion in one of the three files reports a lower line number.
JSDOMWrapper.h, probe: the header with the block as the preprocessor leaves it when the macro is not defined, compiled with the flags of BunProcess.cpp. Clang reports 6 errors: 'NodeConstants' is not a class, namespace, or enumeration (2), unknown type name 'JSDOMGlobalObject' (2), one mismatch of ScriptExecutionContext * types, and extraneous closing brace ('}').
Every translation unit of the Linux debug build was preprocessed with clang -E: 1,140 hand-written sources, each one alone, and 157 generated ones. None has an active line in JSDOMWrapper.h:33-37 or in ncrypto.cpp:25-34.
551 of the 1,297 units read JSDOMWrapper.h. The 40 compared ones are spread over them (every 13th). With the integers masked on the lines that name one of the three files, the text before and after is byte-identical. Without the mask, Undici.cpp differs in one line: WTFReportAssertionFailure(".../JSDOMWrapper.h", 64, ...) becomes 59.
git grep RENAMED_JSDOM_GLOBAL_OBJECT on main finds the two headers only. git blame dates both blocks to March 2022.
vendor/boringssl/include/openssl/base.h defines OPENSSL_VERSION_NUMBER 0x1010107f, and evp.h declares EVP_PKEY_CTX_set_dsa_paramgen_q_bits. The fallback tests < 0x1010105fL.
clang-format reports nothing. bun test test/internal/source-lints/: 194 pass. The three test-crypto-keygen*dsa* tests of the Node suite skip themselves with BoringSSL, as before.
Open pull requests
6,109 were open at 03:44 UTC on 2026-11-07. 112 change one of the two headers and 13 change ncrypto.cpp: 124 in all.
Trial merges (git merge-tree) of the 124:
24 merge with main and with this branch.
99 conflict with main already and gain no conflicted file.
No open pull request deletes these lines. The diffs of the 88 open dead-code pull requests were compared.
How this run looked for dead code
Rust.cargo hawk, the cross-crate analyzer that hawk.toml is written for (upstream main, built on the repository's nightly), on x86_64-unknown-linux-gnu, x86_64-pc-windows-msvc and aarch64-apple-darwin. Each run covers the product graph and --workspace --all-targets. An item is dead when every target that compiles it reports hawk::dead_public. Result: 842 items.
694 are struct fields and enum variants: FFI layouts (most are in bun_windows_sys) and code tables. Kept.
47 more are in the list below. The last 12 are generated code, foreign declarations, or items without a source range.
C and C++. Each #if branch with no active line in any of the 1,297 preprocessed units was listed with its condition: 2,441 branches. 1,098 depend on the platform or the build flavour, and most of the rest are in third-party sources such as the SQLite amalgamation. Outside third-party code 442 lines remain. Apart from the two blocks here, each is a branch for another build (no AOT, malloc debugging, NAPI_VERBOSE), generated code, or a line that an open pull request deletes or edits.
Also:ENABLE(), USE() and HAVE() conditions on a macro that nothing defines (4), #define names that nothing reads (13), and commented-out code of three lines or more (55 blocks).
Found dead in this run, and left alone
Remove libuv on Windows #47776 deletes the code. The mirrors of uv_tcp_t, uv_fs_poll_t and uv_work_t in src/libuv_sys/libuv.rs, with their helper types and layout assertions (96 lines), and WindowsLoop::wake (5 lines). Both Windows targets compile without them.
Not a deletion. The OpenSSL 3 switch in PrintGeneralName (ncrypto.cpp:742-762) never compiles. Without it, prefix is always null and the else branch below it is unreachable, so the cut is a rewrite of that function.
Kept on purpose.
The 23 unused E::E* aliases in windows_errno.rs. The comment above them says why they exist.
Rows of constant tables: UV_ENOTSUP, UV_ENODATA, ARES_EDESTRUCTION, Tag::fchownat, LIBUS_QUEUED_INPUT_NONE.
WindowsLoop::unref and windows_impl::umask. Each has a POSIX twin of the same name.
The empty freebsd stub module of bun_sys.
Not dead, though hawk reports them.
Constants that only an array length or a const generic argument uses: PROPERTY_BITSET_BITS, PIPE_READ_BUFFER_SIZE, MAX_CID_LEN, Kind::ABS, Kind::REL, DiagnosticClass::COUNT.
bun_zstd::inflate_embedded and inflate_embedded_nul. A macro calls them under cfg(bun_codegen_embed).
Modules that only their own file names: EntryPoint in LinkerContext.rs, jsc_abi.
Small, not taken. Four pub use names that no crate imports (bun_aot::diagnostics::log_internal_error, is_colon_token and is_not_semicolon_token in bun_lint, bun_spawn::Result), and WindowsLoop::wait (the comment above it covers two aliases).
Needs a decision. The wasm32 branches in 15 Rust files. No build targets wasm.
no test proof · iteration 0 · the description declares no behaviour change, so there is no failing test to prove; the existing suite in CI is the check
JSDOMWrapper.h: the #ifndef RENAMED_JSDOM_GLOBAL_OBJECT block. The
header includes ZigGlobalObject.h, which defines the macro and the
alias, before the block. The block also holds the closing brace of
namespace WebCore, so a translation unit where it is active does not
compile. ZigGlobalObject.h keeps the alias and loses the guard that
only this block made necessary.
ncrypto.cpp: the EVP_PKEY_CTX_set_dsa_paramgen_q_bits fallback for
OpenSSL older than 1.1.1e. BoringSSL reports 1.1.1g and declares the
function.
This pull request has no repro. It deletes 19 lines in 3 files that the preprocessor never keeps: the #ifndef RENAMED_JSDOM_GLOBAL_OBJECT block of JSDOMWrapper.h with its guard in ZigGlobalObject.h, and the EVP_PKEY_CTX_set_dsa_paramgen_q_bits fallback of ncrypto.cpp for OpenSSL older than 1.1.1e.
Checks that were run: bun bd, 457 tests in 6 files, a comparison of the preprocessed text before and after (654 MB, the same apart from the __LINE__ of assertions), clang-format, the source lints, and trial merges with the 124 open pull requests that change one of the three files. The description has the details.
Self-review: not run as a separate pass. The diff is 19 deleted lines with no runtime effect.
CI: build 138153 finished. All 14 build-bun jobs passed (Linux, macOS, Windows, FreeBSD and Android), so the three files compile on every platform. The build is red only on tests that are red without this diff: 27 files under test/aot/ and test/bundler/compile-aot-size.test.ts. Each one also fails on the final builds of the last merged pull requests, and this diff changes no code that runs.
No actionable comments were generated in the recent review.
Walkthrough
The changes make the WebCore::JSDOMGlobalObject alias unconditional in ZigGlobalObject.h and remove its guarded declaration from JSDOMWrapper.h. They also remove a pre-1.1.1e OpenSSL fallback macro from ncrypto.cpp.
Changes
JSDOM global object alias
Layer / File(s)
Summary
JSDOM global object alias declaration src/jsc/bindings/JSDOMWrapper.h, src/jsc/bindings/ZigGlobalObject.h
The guarded alias is removed from JSDOMWrapper.h. ZigGlobalObject.h now declares the alias unconditionally.
The pre-1.1.1e fallback mapping for EVP_PKEY_CTX_set_dsa_paramgen_q_bits to EVP_PKEY_CTX_ctrl is removed.
Suggested reviewers: dylan-conway
Merge Risk: Minimal · up to 4b7c9
The cleanup has no demonstrated behavior change in the configured build. No actionable merge-blocking issue remains; normal build checks should complete before merging.
Pre-merge checks: 4
Check name
Status
Explanation
Title check
The title names the two dead-code removals in JSDOMWrapper.h and ncrypto.cpp, which match the changeset. It is long at 96 characters, and the colon clause repeats detail, but it still conveys the prim…
Description check
The description follows the template's two sections. 'What does this PR do?' is covered by the Problem and Fix sections, and 'How did you verify your code works?' is covered by the test run, clang-for…
Linked Issues check
Check skipped because no linked issues were found for this pull request.
Out of Scope Changes check
Check skipped because no linked issues were found for this pull request.
Full details: Title check
Explanation
The title names the two dead-code removals in JSDOMWrapper.h and ncrypto.cpp, which match the changeset. It is long at 96 characters, and the colon clause repeats detail, but it still conveys the primary change clearly.
Full details: Description check
Explanation
The description follows the template's two sections. 'What does this PR do?' is covered by the Problem and Fix sections, and 'How did you verify your code works?' is covered by the test run, clang-format, and source-lint results. It is long and includes extensive evidence in collapsed notes, but the required information is present and complete.
Autofix · Keep fixing CodeRabbit findings and required CI, and resolving merge conflicts
Thanks for using CodeRabbit! It's free for OSS, and your support helps us grow. If you like it, consider giving us a shout-out.
Reviewing files that changed from the base of the PR and between 7ca0042 and 4b7c970.
Files selected for processing: 3
src/jsc/bindings/JSDOMWrapper.h
src/jsc/bindings/ZigGlobalObject.h
src/jsc/bindings/ncrypto.cpp
Included review availability: This review used your included allowance. Your plan provides up to 10 included reviews per hour; 4 remain after this review.
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
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.
![claude[bot]](https://avatars.githubusercontent.com/in/1236702?s=60&v=4){kind=small}
Problem
JSDOMWrapper.hhas an#ifndef RENAMED_JSDOM_GLOBAL_OBJECTblock that no translation unit can compile. The header first includesZigGlobalObject.h, which defines the macro and the same alias. The block also holds the closing brace ofnamespace WebCore: when it is active, clang stops withextraneous closing brace ('}').ncrypto.cppdefinesEVP_PKEY_CTX_set_dsa_paramgen_q_bitsfor OpenSSL older than 1.1.1e. BoringSSL reports 1.1.1g and declares that function.Fix
ZigGlobalObject.hloses the 3 guard lines around its alias: no other file names the macro.ncrypto.cppand of 40 units that readJSDOMWrapper.h(654 MB) differs only in the__LINE__of assertions.bun bd, then 457 tests in 6 files (node-crypto,web-globals,globals,abort,message-channel,broadcast-channel).Behaviour change: none
Background
JSDOMGlobalObjectis the WebCore name for the global object. In Bun it is an alias ofZig::GlobalObject. The macro is a guard from 2022 that let either header declare the alias first.ncryptois the OpenSSL wrapper ofnode:crypto. Bun links BoringSSL on every platform.bun_libuv_sysandbun_uws_sys. Remove libuv on Windows #47776 deletes that code, so they stay out.Downsides
mainin 1,612 files.Notes
Evidence
JSDOMWrapper.h, probe: the header with the block as the preprocessor leaves it when the macro is not defined, compiled with the flags ofBunProcess.cpp. Clang reports 6 errors:'NodeConstants' is not a class, namespace, or enumeration(2),unknown type name 'JSDOMGlobalObject'(2), one mismatch ofScriptExecutionContext *types, andextraneous closing brace ('}').clang -E: 1,140 hand-written sources, each one alone, and 157 generated ones. None has an active line inJSDOMWrapper.h:33-37or inncrypto.cpp:25-34.JSDOMWrapper.h. The 40 compared ones are spread over them (every 13th). With the integers masked on the lines that name one of the three files, the text before and after is byte-identical. Without the mask,Undici.cppdiffers in one line:WTFReportAssertionFailure(".../JSDOMWrapper.h", 64, ...)becomes59.git grep RENAMED_JSDOM_GLOBAL_OBJECTonmainfinds the two headers only.git blamedates both blocks to March 2022.vendor/boringssl/include/openssl/base.hdefinesOPENSSL_VERSION_NUMBER 0x1010107f, andevp.hdeclaresEVP_PKEY_CTX_set_dsa_paramgen_q_bits. The fallback tests< 0x1010105fL.clang-formatreports nothing.bun test test/internal/source-lints/: 194 pass. The threetest-crypto-keygen*dsa*tests of the Node suite skip themselves with BoringSSL, as before.Open pull requests
ncrypto.cpp: 124 in all.git merge-tree) of the 124:mainand with this branch.mainalready and gain no conflicted file.JSDOMWrapper.h, because it renamesZig::GlobalObjecton the alias line of the block.How this run looked for dead code
cargo hawk, the cross-crate analyzer thathawk.tomlis written for (upstreammain, built on the repository's nightly), onx86_64-unknown-linux-gnu,x86_64-pc-windows-msvcandaarch64-apple-darwin. Each run covers the product graph and--workspace --all-targets. An item is dead when every target that compiles it reportshawk::dead_public. Result: 842 items.bun_windows_sys) and code tables. Kept.#ifbranch with no active line in any of the 1,297 preprocessed units was listed with its condition: 2,441 branches. 1,098 depend on the platform or the build flavour, and most of the rest are in third-party sources such as the SQLite amalgamation. Outside third-party code 442 lines remain. Apart from the two blocks here, each is a branch for another build (no AOT, malloc debugging,NAPI_VERBOSE), generated code, or a line that an open pull request deletes or edits.ENABLE(),USE()andHAVE()conditions on a macro that nothing defines (4),#definenames that nothing reads (13), and commented-out code of three lines or more (55 blocks).Found dead in this run, and left alone
uv_tcp_t,uv_fs_poll_tanduv_work_tinsrc/libuv_sys/libuv.rs, with their helper types and layout assertions (96 lines), andWindowsLoop::wake(5 lines). Both Windows targets compile without them.ENABLE(MALLOC_BREAKDOWN)branches ofScriptExecutionContext(Remove dead code from the inspector protocol package, bun_jsc, bun_install, and the Windows named pipe modules #40557). Nothing defines the macro.ENABLE(BINDING_INTEGRITY)blocks (Remove dead code from the JSC, WebCore, and node bindings, uWS/uSockets, SecretsLinux, and seven Rust crates #42431, Remove dead code from the WebCore bindings and 20 Rust crates #40525, Remove dead code from libtcc1.c, four C++ preprocessor conditions, JSPropertyIterator, and hawk.toml #44126).ncrypto.cpp(Remove dead code from SerializedScriptValue, ncrypto, and the WebCore IDL converters #43644).#if OPENSSL_VERSION_MAJOR >= 3around#include <openssl/provider.h>inncrypto.cpp(3 lines). crypto: port OpenSSL's prime search for generatePrime with add #37963 adds two includes directly above it.switchinPrintGeneralName(ncrypto.cpp:742-762) never compiles. Without it,prefixis always null and theelsebranch below it is unreachable, so the cut is a rewrite of that function.E::E*aliases inwindows_errno.rs. The comment above them says why they exist.UV_ENOTSUP,UV_ENODATA,ARES_EDESTRUCTION,Tag::fchownat,LIBUS_QUEUED_INPUT_NONE.WindowsLoop::unrefandwindows_impl::umask. Each has a POSIX twin of the same name.freebsdstub module ofbun_sys.PROPERTY_BITSET_BITS,PIPE_READ_BUFFER_SIZE,MAX_CID_LEN,Kind::ABS,Kind::REL,DiagnosticClass::COUNT.bun_zstd::inflate_embeddedandinflate_embedded_nul. A macro calls them undercfg(bun_codegen_embed).EntryPointinLinkerContext.rs,jsc_abi.pub usenames that no crate imports (bun_aot::diagnostics::log_internal_error,is_colon_tokenandis_not_semicolon_tokeninbun_lint,bun_spawn::Result), andWindowsLoop::wait(the comment above it covers two aliases).wasm32branches in 15 Rust files. No build targets wasm.no test proof · iteration 0 · the description declares no behaviour change, so there is no failing test to prove; the existing suite in CI is the check