Remove dead code from bun init, bun create, bun_md, bun_install, and the socket and Postgres code by robobun · Pull Request #44102 · oven-sh/bun

robobun · 2026-09-27T09:02:27Z

Behaviour change: none

Problem

10 groups of code in 13 files cannot run, or hold values that nothing reads. No lint reports them, because each still has a textual user.
Example: bun init returns early for every React template (src/runtime/cli/init_command.rs:615), so the React dependency tables after that return never run.

Fix

Delete the 10 groups: 550 lines removed, 20 added. The Notes give each group with its proof.
Correct because each proof is local: a constant, a variant that nothing constructs, a field that nothing reads, or an exit before the code.
Verified: bun bd, cargo check for 12 targets and each custom cfg, and the tests of each touched area.
Self-reviewed: 12 concerns raised, 11 addressed. Rejected: to unwrap the second match template in init_command.rs, which re-indents 88 lines in a file that 17 open PRs touch.

Background

The workspace denies dead_code. But rustc counts a store to a field, and a call in a branch that cannot run, as uses.
Three groups are follow-ups that earlier PRs named and left: Ast.runtime_imports (Remove dead code from js_printer, bundler, js_parser, patch, shell, http #35978), Subcommand::Pack (Remove dead code: write-only fields and never-set options in 11 crates #43745), JSCInlines.h (Remove dead code from sys/windows, cli, shell, test_runner, v8 shim #36237, closed as stale).
The self-review weighed 16 groups. Six are out: an open PR uses the code, or a merged PR kept it on purpose.

Downsides

None found. Checked: no removed name is left in the tree or in the generated code. No open PR adds a user that git does not report as a conflict (2793 searched). bun create and bun pm pack print the same bytes as the released bun.

Notes

Removed, one group each (net lines)

src/runtime/cli/init_command.rs (254): DependencyGroup with REACT, TAILWIND, SHADCN, Template::{dependencies, is_react, write_to_package_json, scripts}, and the code that wrote a dependencies object. exec returns in match template for the three React variants and does not assign template after that. So only Blank and TypescriptLibrary reach the later code, and BLANK.dependencies was the empty slice. The React templates take their dependencies from their embedded package.json. The one list that is left is now const DEV_DEPENDENCIES.
src/runtime/cli/create_command.rs (86): CreateListExamplesCommand. bun_create exits with the help text when there is no positional, and CreateCommand::extract_info does the same, both before CreateCommand::exec tests positionals.is_empty(). With it gone, both callers of Example::print pass a name and both callers of fetch_all_local_and_remote pass no progress node. So Example::print takes &[u8], the static APP_NAME_BUF goes, and fetch_all_local_and_remote and fetch_all lose the node parameter. fetch_all only stored that node in client.progress_node, whose initial value is None (src/http/AsyncHTTP.rs:174). bun create with no template prints a usage text since Progress for Next.js #4468 (2023-09-07), which made bun create <template> run bunx create-<template>.
src/runtime/cli/create_command.rs (14): is_nextjs, is_create_react_app, create_react_app_entry_point_path. Immutable locals bound to false, false and b"", with the if that printed bun bun hints.
src/jsc/bindings/JSCInlines.h (57, whole file). All 147 includes in the tree name JavaScriptCore/JSCInlines.h, and src/jsc/bindings has no JavaScriptCore directory. WebKit headers that include "JSCInlines.h" with no prefix reach WebKit's file, because a quoted include searches the directory of the including file first. Bun's file was a stale variant: it included Operations.h where WebKit's header includes OperationsInlines.h. Remove dead code from sys/windows, cli, shell, test_runner, v8 shim #36237 deleted the same file and its C++ compiled on every CI lane (build 87733).
src/md/blocks.rs (26): the first ATX trim block in analyze_line. *line = Line::default() sets line.end to 0, and the first store to line.end is after the loop. So both guards (line.end > line.beg) are false. The block after line.end = off does the trimming.
src/md/containers.rs, src/md/autolinks.rs (16): is_container_compatible is now existing.ch == new.ch. One branch returned false right before the function returned false. Another tested for two > right before the test for equal characters. The helper is_list_bullet had no other user.
src/runtime/socket/socket_body.rs, Listener.rs (15): Flags::OWNED_PROTOS. Its only readers guarded protos.set(None) right before protos.set(new). Its writers were the Default impl and the raw upgradeTLS socket. Flags now derives Default, which is the empty set. No code reads the flags word as a whole.
src/sql_jsc/postgres/PostgresSQLConnection.rs (12): the map backend_parameters. The ParameterStatus message is still decoded, so the stream advances. The OutOfMemory error of the insert is gone with the insert. The map has had no reader since the first Postgres client added it in Add unusable postgres client behind a canary-or-debug-only feature flag #11920 (2024-07-07).
src/install/PackageManager/CommandLineArguments.rs, PackageManager.rs (41): Subcommand::Pack and PACK_PARAMS. bun pm pack runs as Subcommand::Pm. Nothing casts Subcommand to an integer. bun pm --help lists the pack flags.
src/ast/ast_result.rs, src/js_parser/p.rs (9): Ast.runtime_imports. Written in to_ast, never read. The parser keeps its own table.

Verification

All on the final commit, rebased on a4f1429.

bun bd passes. That is a full C++ compile against the new WebKit, without JSCInlines.h.
cargo check --workspace passes for 12 targets (scripts/rust-check-all.ts), with --all-targets, and on the host with --cfg=bun_codegen_embed and with --cfg=socket_fault_injection --cfg=bun_debug --cfg=bun_asan. cargo clippy, cargo fmt --check and bun run rust:mordant (3 targets, fresh records) exit 0.
Tests with the debug build, no failure: test/cli/init/init.test.ts (17, asserts the dependencies of all three React templates), test/js/bun/md/ (1098), bun-pack.test.ts, bun-pm.test.ts and bun-create.test.ts (129), transpiler.test.js, runtime-transpiler.test.ts and bundler_cjs2esm.test.ts (365), socket-retention.test.ts (7).
Postgres, 31 files (postgres-*.test.ts, sql.test.ts and 3 more) with a 120 s timeout: 260 pass, 0 fail. An earlier run with the default timeout of 5 s had 6 timeouts, while the load average of the machine was above 500. sql.test.ts runs only the tests that need no docker.
Sockets and TLS, 6 files (socket.test.ts, tcp-server.test.ts, tls-reject-before-client-cert.test.ts, node-tls-connect.test.ts, node-tls-server.test.ts, node-http2-upgrade.test.mts): 432 pass, 1 fail. The failure is should not call drain before handshake, which needs DNS for www.example.com. It fails the same way with the released bun.
test/internal/source-lints/: 169 pass, 1 fail. primordials-exports.test.ts stopped at the 5 s timeout. It passes in 0.7 s when it runs alone.
test/cli/create/create-jsx.test.ts: 5 pass, 8 fail, the same 8 with the released bun. They need ui.shadcn.com, which this machine cannot resolve.
Probes, same bytes as the released bun: bun create with no argument (help text, exit 1), bun create oven-sh/<missing repository> ./out (the path through fetch_all_local_and_remote and Example::print), bun pm pack --help, and bun pm pack --destination --gzip-level (same tarball).
Not measured: the timing of a debug build of the base commit. Not built here: macOS, Windows and FreeBSD. For those, cargo check covers the Rust code and CI covers the C++ code.

Self-review

The self-review read a cut of 16 groups. I checked each of its claims in the PR diffs and in the code before I acted. The finishing edits in groups 1, 2 and 6 came after it. The compile checks, the probes and the tests above cover them. The 12 concerns:

The allocated parameter of MimeType::init: node:fs: openAsBlob pins the file, and a read through the Blob verifies it #43896 adds a call with three arguments. git reports no conflict, and the build breaks after the second merge. Group removed.
JSGlobalObjectSqlExt::sql_vm and ArgumentsSlice::_vm: Remove dead code: 17 log scopes nothing logs to and four fields nothing reads, in 10 crates #44050 and socket, timer, sql, redis: replace unsafe with typed ownership #40139 remove the field. Group removed.
TransformOutstream: bun build --no-bundle: honor --sourcemap #41939 edits the same functions, and Remove dead code from the JSC, WebCore, and node bindings, uWS/uSockets, SecretsLinux, and seven Rust crates #42431 left it for a change that also handles BundleOptions::output_dir_handle. Group removed.
Six BOOL_FIELD_SETTERS entries: feat(mdx): add native compilation, loading, and direct serving #27047 reads the table with no filter, and its test sets permissiveAutolinks. Group removed.
Bun__allocUint8ArrayForCopy: Remove dead code from react_compiler, the WebCore/IDL bindings, install_jsc, node-fallbacks, and misc crates #39420 kept it so that ArrayBuffer::alloc mirrors ArrayBuffer::create. The removal also made a Uint8Array call panic at run time. Group removed.
help_command::Reason::InvalidCommand: the removal leaves an enum with one variant and a printer macro with one use. That is a refactor of the help printer. Group removed.
Example::print kept an Option that no caller sets to None. Finished in group 2.
fetch_all_local_and_remote and fetch_all kept a node parameter that no caller sets. Finished in group 2.
is_container_compatible kept a branch that the next branch covers. Finished in group 6.
The second match template in init_command.rs has an arm that cannot run. Rejected, see Fix.
The description did not name the earlier PRs that left Subcommand::Pack and JSCInlines.h. Added.
The description did not name the open PRs that conflict. Added below.

Open PRs

I listed the files of all 5725 open PRs. 2793 of them touch a directory of this PR, a C or C++ file, or more than 100 files. I searched the added lines of those 2793 for every removed name and read each of the 118 hits.
Suggest the closest command when bun run or bun pm finds none #40855 defines its own enum PmSubcommand with a Pack variant. It does not use bun_install's Subcommand.
sql: implement Query.cancel() for postgres #33370, Rewrite the uSockets core in Rust #34037, Replace boolean flag parameters with two-variant enums #37180, Make the allocation-only collection APIs infallible and drop the OOM wrappers around them #38888 and sql(postgres): move the allocate-and-dial block of createConnection into PostgresSQLConnection::open #43942 move the line that sets backend_parameters. Each conflicts with this PR in that file, so git reports it.
A trial merge shows new conflicts with Run bun create hooks through the shell and stop on failure #41598 (create_command.rs), sql(postgres): move the allocate-and-dial block of createConnection into PostgresSQLConnection::open #43942 (PostgresSQLConnection.rs), init: exit 1 when the nested bun install fails #38474 and Atomic writes: convert bun init and test a write that fails #39701 (init_command.rs). cli: remove unsafe from bun run / bun test / --filter and the other CLI commands #40242 and bun init: add react + @types/react to the blank scaffold so .tsx works #36607 conflict with main already. There is no conflict with socket, timer, sql, redis: replace unsafe with typed ownership #40139, node:fs: openAsBlob pins the file, and a read through the Blob verifies it #43896, bun build --no-bundle: honor --sourcemap #41939, Remove dead code: 17 log scopes nothing logs to and four fields nothing reads, in 10 crates #44050 or feat(mdx): add native compilation, loading, and direct serving #27047.
No open dead-code PR removes a line that this PR removes.

One deletion was wrong and is not in this PR

I first deleted StoredTrace::from (src/bun_core/Global.rs). bun bd failed: src/install/PackageInstall.rs:224 calls it under #[cfg(bun_debug)]. Plain cargo check and mordant do not set that cfg, so both see no caller. After the revert I searched the whole tree for every other removed name and compiled with each custom cfg.

Found and left in place

Needs the open PR to land first: the allocated parameter of MimeType::init with content_type_needs_free in RequestContext.rs (node:fs: openAsBlob pins the file, and a read through the Blob verifies it #43896), TransformOutstream with BundleOptions::output_dir_handle (bun build --no-bundle: honor --sourcemap #41939), sql_vm and the VM parameter of ArgumentsSlice::init (Remove dead code: 17 log scopes nothing logs to and four fields nothing reads, in 10 crates #44050, socket, timer, sql, redis: replace unsafe with typed ownership #40139), the format path of pretty_format.rs (Remove dead code from the web streams bindings, the test runner formatter, and the bindgen codegen #41445), cssInternals.attrTest with StyleAttribute (Remove dead code from bun_core, bun_jsc, bun_css, react_compiler, and two C++ bindings #40367), MaxHeapAllocator (Remove dead code from bun_jsc, bun_core, the bindgenv2 codegen, 15 error types, and the C++ bindings #40915).
Dead on main, used by an open PR: six BOOL_FIELD_SETTERS entries that parse_options in MarkdownObject.rs skips (feat(mdx): add native compilation, loading, and direct serving #27047).
Needs a refactor: help_command::Reason::InvalidCommand. The second match template in init_command.rs.
Looks like a bug, not a leftover: bun init --react tailwind cannot match because the argument must start with -. --dump-environment-variables is parsed and ignored. publish always sends workspaces/false.
Protocol tables and handlers: the MySQL CharacterSet table, the Postgres binary int2, oid and timetz parsers, the MySQL Time encode path.
Needs a maintainer decision: the httplog! macro in src/runtime/server/mod.rs builds its arguments and prints nothing. server_body.rs has a working one.
Read as dead by a second reader, not read by me: Request::detach_lifetime in picohttp, WebSocket::clear_send_buffers(free), InternalState::content_encoding_i.

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

…the socket and Postgres code Each group cannot run, or holds a value that nothing reads. No lint reports them, because each still has a textual user: an unreachable call, a store, or a parameter that every caller sets to one value. - bun init: the React dependency tables and the code that wrote a `dependencies` object. `exec` returns early for the React templates. - bun create: `CreateListExamplesCommand` (the help text exits first), the `Option` of `Example::print`, the progress node parameter of `fetch_all_local_and_remote` and `fetch_all`, and three constant locals with the hints they guarded. - bun_md: the first ATX trim block in `analyze_line` (`line.end` is 0 there) and two branches of `is_container_compatible` that do not change its result. - bun_install: `Subcommand::Pack` and `PACK_PARAMS`. `bun pm pack` runs as `Subcommand::Pm`. - Sockets: `Flags::OWNED_PROTOS`, which only guarded a store of `None` right before a store of the new value. - Postgres: the map `backend_parameters`, which nothing reads. - bun_ast, bun_js_parser: `Ast.runtime_imports`, which nothing reads. - `src/jsc/bindings/JSCInlines.h`, which no include reaches.

coderabbitai · 2026-09-27T09:09:16Z

Navigate logical layers of code changes, visualize relationships, and explore their blast radius.

No actionable comments were generated in the recent review. 🎉

ℹ️ Recent review info

⚙️ Run configuration

Configuration used: Repository: oven-sh/bun/.coderabbit.yaml

Review profile: ASSERTIVE

Plan: Essentials

Run ID: 5c7194a2-caef-4925-b8e5-6563a0b61e76

📥 Commits

Reviewing files that changed from the base of the PR and between a4f1429 and ee4ef39.

📒 Files selected for processing (13)

src/ast/ast_result.rs
src/install/PackageManager.rs
src/install/PackageManager/CommandLineArguments.rs
src/js_parser/p.rs
src/jsc/bindings/JSCInlines.h
src/md/autolinks.rs
src/md/blocks.rs
src/md/containers.rs
src/runtime/cli/create_command.rs
src/runtime/cli/init_command.rs
src/runtime/socket/Listener.rs
src/runtime/socket/socket_body.rs
src/sql_jsc/postgres/PostgresSQLConnection.rs

💤 Files with no reviewable changes (7)

src/install/PackageManager.rs
src/md/blocks.rs
src/jsc/bindings/JSCInlines.h
src/runtime/socket/Listener.rs
src/md/autolinks.rs
src/js_parser/p.rs
src/ast/ast_result.rs

Included review availability: This review used your included allowance. Your plan provides up to 10 included reviews per hour; 7 remain after this review.

Walkthrough

The changes remove package-manager and parser state, CLI template behaviors, a JavaScriptCore header, socket protocol-flag handling, and PostgreSQL parameter storage. They also adjust Markdown container compatibility and ATX header trimming.

Changes

AST parser state

Layer / File(s)	Summary
Remove runtime imports from AST `src/ast/ast_result.rs`, `src/js_parser/p.rs`	The AST no longer stores `runtime_imports`. Parser construction still uses the parser’s runtime imports to select `require_ref`.

Package manager Pack subcommand

Layer / File(s)	Summary
Remove Pack command and options `src/install/PackageManager.rs`, `src/install/PackageManager/CommandLineArguments.rs`	Removes the Pack subcommand, its option table and help output, and its argument-parsing paths.

JavaScriptCore bindings header

Layer / File(s)	Summary
Remove inline-header umbrella `src/jsc/bindings/JSCInlines.h`	Deletes the umbrella header, its include guidance, and its JavaScriptCore inline-header includes.

Markdown parsing

Layer / File(s)	Summary
Container marker compatibility `src/md/autolinks.rs`, `src/md/containers.rs`	Removes `is_list_bullet`; container compatibility now depends on marker-character equality.
ATX header trimming `src/md/blocks.rs`	Removes the earlier trimming of trailing whitespace and closing hash marks. The later shared trimming step remains.

CLI template commands

Layer / File(s)	Summary
Create command and example fetching `src/runtime/cli/create_command.rs`	Removes the examples-list command and progress-node handling. `Example::print` now requires an app name, which template-not-found paths provide from the requested directory name.
Init dependencies and package metadata `src/runtime/cli/init_command.rs`	Init retains the shared `@types/bun` dev dependency and TypeScript peer-dependency check. Template-specific runtime dependency handling and React package metadata generation are removed.

Socket protocol flags

Layer / File(s)	Summary
Remove owned-protocol flag handling `src/runtime/socket/Listener.rs`, `src/runtime/socket/socket_body.rs`	Removes `OWNED_PROTOS` and conditional protocol clearing during reconnect. `Flags` now derives `Default`, and the raw TLS-upgrade half no longer sets that flag.

PostgreSQL backend parameters

Layer / File(s)	Summary
Remove backend parameter storage `src/sql_jsc/postgres/PostgresSQLConnection.rs`	Removes the connection’s `backend_parameters` field and the insertion of decoded `ParameterStatus` values.

Suggested reviewers: jarred-sumner

Priority: ➖ Normal

Merge Risk: ⚪ Minimal · up to ee4ef

The documented pack command remains available, and no behavior change requiring a fix was established. The PR is mergeable after normal checks.

🚥 Pre-merge checks | ✅ 4

✅ Passed checks (4 passed)

Check name	Status	Explanation
Linked Issues check	✅ Passed	Check skipped because no linked issues were found for this pull request.
Out of Scope Changes check	✅ Passed	Check skipped because no linked issues were found for this pull request.
Title check	✅ Passed	The title clearly summarizes the primary change: removing dead code across the listed Bun initialization, creation, Markdown, package-manager, socket, and PostgreSQL areas.
Description check	✅ Passed	The description explains the problem, identifies the removed code groups, states that behavior should not change, and provides detailed build, test, probe, limitation, and self-review information. It …

_{Comment @coderabbitai help to get the list of available commands.}

robobun · 2026-09-27T09:12:12Z

Status: ready for review. CI is green. This PR only deletes code, so it has no new test.

CI on commit ee4ef39, which is on the current head of main (a4f1429):

Buildkite build 121175 passed: 182 jobs, 0 failed. That includes the C++ and Rust builds for macOS, Windows, FreeBSD and Linux.
The GitHub checks pass, except claude-find-issues. That check fails on every PR opened since 2026-09-27 08:00 UTC, so this PR is not the cause.

How it was checked before the push:

bun bd passes, which includes a full C++ compile without JSCInlines.h.
cargo check --workspace passes for 12 targets, with --all-targets, and with each custom cfg of the build (bun_debug, bun_asan, bun_codegen_embed, socket_fault_injection). cargo clippy, cargo fmt --check and bun run rust:mordant exit 0.
The tests of each touched area pass with the debug build. The Notes in the description give the numbers and explain the failures that the test machine also shows with the released bun.
bun create and bun pm pack print the same bytes as the released bun.

This PR conflicts with four open PRs, which need a rebase after the merge: #41598, #43942, #38474, #39701.

claude

I reviewed this PR and didn't find any bugs. Because it deletes 550 lines across ten unrelated subsystems (sockets/TLS, Postgres, the parser, the markdown parser, three CLI commands) and some of the removed code was a latent product surface (the bun create examples listing, Postgres backend_parameters), a maintainer should still confirm these are wanted deletions rather than features to revive.

What was reviewed:

Independently confirmed each deadness proof: React templates return at init_command.rs:618 before the dependency tables; bun_create in cli/mod.rs exits on missing positionals before CreateCommand::exec; JsCell::set already drops the previous protos Box so the OWNED_PROTOS guard was a no-op; line.end is 0 from Line::default() until blocks.rs:564, so the removed ATX trim block never ran; the old is_container_compatible truth table equals existing.ch == new.ch.
Grepped for surviving references: no reader of Ast.runtime_imports or backend_parameters, no constructor of Subcommand::Pack (enum derives only IntoStaticStr, no integer casts), no unprefixed "JSCInlines.h" include anywhere in src/ or cmake/.
Checked all remaining SocketFlags::default() users differ only in the removed bit, and Example::print/fetch_all* callers all match the new signatures.

Extended reasoning...

The change is a single commit removing dead code from 13 files (+20/-550) spanning the JS parser AST, package-manager CLI arg parsing, a vendored JSC umbrella header, the markdown block/container parser, bun init/bun create, the TLS socket ALPN protos ownership flag, and Postgres ParameterStatus handling. It touches the TLS socket and Postgres wire-parsing surfaces, but the edits there only remove a redundant drop guard and an unread map; no authentication, validation, or crypto logic changes. Every deletion I checked is provably unreachable or unread at HEAD, no CODEOWNERS entry covers the changed paths, and the bug hunt ran dry. The breadth across unrelated subsystems and the fact that two removed groups were dormant user-facing features (examples listing, backend parameter capture) are what tip this to defer rather than approve.

robobun · 2026-09-27T09:29:24Z

History of the two groups that the review names as possible features:

The bun create examples listing: bun create with no template prints a usage text since Progress for Next.js #4468 (2023-09-07). That change made bun create <template> run bunx create-<template>. The listing that still runs, after a template is not found, is not changed.
The Postgres map backend_parameters: the first Postgres client added it in Add unusable postgres client behind a canary-or-debug-only feature flag #11920 (2024-07-07). It has had no reader since then. I found no issue and no open PR that asks for the server parameters.

Both facts are now in the Notes of the description. If a maintainer wants to keep one of these groups, I will remove it from this PR.

robobun · 2026-09-28T19:55:34Z

@robobun wake up!!

Jarred-Sumner · 2026-09-28T20:40:18Z

@robobun wake up!!

robobun · 2026-09-28T21:58:07Z

I am here. The state of this PR:

CI is green. Buildkite build 121175 passed on all lanes (182 jobs, 0 failed). The GitHub checks pass, except claude-find-issues, which fails on every PR opened since 2026-09-27 08:00 UTC.
The branch is on the current head of main (a4f1429). There is no conflict.
No review thread is open. The automated review found no bug. It asks a maintainer to confirm two deletions: the bun create examples listing and the Postgres map backend_parameters. The history of both is in the comment above.

The PR waits for your review. If you want a change, tell me which one. For example, I can remove one group from the PR.

) Behaviour change: none One error path differs. See Downsides. ### Problem - `bun build --dump-environment-variables` does nothing. `BUILD_ONLY_PARAMS` declares the flag (`src/runtime/cli/Arguments.rs:538`), but no code reads it. - `DebugOptions.dump_environment_variables` is always `false`, so the branch at `src/runtime/cli/build_command.rs:604` and `Transpiler::dump_environment_variables` cannot run. - The completions still offer the flag, with `--dump-limits` and `--disable-bun-js`, whose fields #36184 deleted. ### Fix - Delete the param, the field, the branch and the function. Remove the three names from the completions. - Correct on current main: the parser skips an unknown long flag in silence (`src/clap/streaming.rs:157`). The bare flag gives the same bundle as before. - Verified: `bun bd`, `test/bundler/cli.test.ts`, `bundler_env.test.ts`, `test/cli/bun.test.ts`, `test/cli/run/env.test.ts`. No new test: `REVIEW.md` says "Do not add tests to check dead code stays dead". - Self-reviewed: 14 concerns raised, 13 addressed. Rejected: to drop the first line. It stays because the PR adds no test. ### Background - The flag printed the environment map as JSON and stopped. `bun build` printed valid JSON up to bun 0.8.0, a malformed stream up to 1.0.11, and nothing since 1.0.12 (#6395). - #43745, #44102 and #44343 ask to connect the flag or remove it. Close this PR if a maintainer wants it connected. - Considered connecting it: it prints every process variable, and no `.env` value without `--env`. The docs give `bun --print process.env`. ### Downsides - `bun build --dump-environment-variables=value` exited 1 with "does not take a value". It is now skipped. - The bare flag still builds with no message. With #40560 (open) it exits 1, like every unknown flag. <details><summary>Notes</summary> #### Deleted, one line each - `parse_param!("--dump-environment-variables")` in `BUILD_ONLY_PARAMS` (`src/runtime/cli/Arguments.rs`). - `DebugOptions.dump_environment_variables` and its default (`src/options_types/context.rs`). - The `if ctx.debug.dump_environment_variables { ... }` branch in `BuildCommand::exec`, and the stale comment above it (`src/runtime/cli/build_command.rs`). - `Transpiler::dump_environment_variables` (`src/bundler/transpiler.rs`). `write_json_string` and the iterator of `bun_dotenv::Map` have other callers. - `completions/bun.zsh`: the `--dump-environment-variables` and `--dump-limits` specs in `_bun_run_completion`. - `completions/bun.bash`: `--dump-environment-variables --dump-limits --disable-bun-js` in `GLOBAL_OPTIONS`. #### Behaviour, measured Released bun 1.4.3 against a debug build of this branch. "Same bundle" means the same bytes on stdout as `bun build x.js`. | Input | Before | After | |---|---|---| | `bun build --dump-environment-variables x.js` | exit 0, same bundle | exit 0, same bundle | | `bun build --dump-environment-variables=1 x.js` | exit 1, "The argument '--dump-environment-variables' does not take a value." | exit 0, same bundle | | `bun build --dump-environment-variables= x.js` | exit 1, same error | exit 0, same bundle | | `BUN_OPTIONS=--dump-environment-variables=1 bun build x.js` | exit 1, same error | exit 0, same bundle | | `bun build --no-such-flag=1 x.js` | exit 0, same bundle | exit 0, same bundle | `bun build --help` never listed the flag. `WARN_ON_UNRECOGNIZED_FLAG` (`src/clap/streaming.rs:10`) is never set to true, so the skip prints nothing. This PR does not touch it. #40560 removes it. #### History - `38f83c50c4` and `5691bf385b` (October 2021) added the flag for the old `bun bun` and `bun dev` commands. - Runs of release binaries with `bun build --dump-environment-variables x.js`: 0.8.0 prints valid JSON. 1.0.1 and 1.0.11 print one JSON string per fragment (`"{",` then `"\n ",` then `"PATH",`), which is not valid JSON. The cause is `19aa9d93de` (#4233, first in 0.8.1): `Map.jsonStringify` changed from `writer.writeAll` to `writer.write` on the JSON stream. - Up to 1.0.11 the flag was declared one time, with help text, in `debug_params`, and `src/cli.zig:593` read it for every command. #6395 deleted that list and added a new declaration, with no help text, to `build_only_params` only. - In the branch of #6395, `5bf8ad8fb9` commented out the line that reads the flag, under a source comment that says the two dump flags were "only used in bun dev". `fbbb5e7e38` ("Remove comments") deleted the commented lines. The merged commit does not contain that comment. `bun build` had its own consumer (`src/cli/build_command.zig:220`), and it stayed. - #5712 is the only issue that names the flag. It asked for `bun run` on bun 1.0.2, and it ended with "This flag appears to have been removed". - No maintainer has ruled on the flag. #36184 deleted `DebugOptions.dump_limits` and `DebugOptions.fallback_only`, the fields of the other two debug flags. #41699 removed the dead Solid JSX runtime in the same way. #### If the flag is wanted back The one-line reader is not enough. `bun build --app` returns into `bake::production::build_command` before the branch. The branch runs after the compile target lookup, so it must move to directly after `configure_defines()`. `--watch` needs a decision. The flag needs help text and docs. #### Completions: what stays - The three names are the completion entries of the three old `DebugOptions` debug flags. After this PR none of their fields exists. - `completions/bun.bash` `GLOBAL_OPTIONS` still has nine names that no param table declares: `--use --bunfile --server-bunfile --disable-react-fast-refresh --disable-hmr --jsx-production --platform --public-dir --inject`. #35443 generates that file from `bun-cli.json`. #42275 rewrites the same line. - `_bun_run_completion` in `completions/bun.zsh` still has eleven specs that `bun run` does not declare: `--no-summary --version -v --revision --external --packages --minify --minify-syntax --minify-whitespace --minify-identifiers --target`. Other commands declare them. #### `DebugOptions` fields with no reader that stay - `output_file`: #44050 removes it. - `editor`: #38059 connects it. - `package_bundle_map`: `[bundle.packages]` in `bunfig.toml` fills it (`src/bunfig/bunfig.rs:910`) and nothing reads it. No open PR covers it. #### Other params with no reader 207 long flags are declared in `Arguments.rs`. 13 have no `b"--name"` reader in `src/**/*.rs`: - `--tls-min-v1.0`, `--tls-min-v1.1`, `--tls-min-v1.2`, `--tls-max-v1.3`: `src/js/node/tls.ts` reads them from `process.execArgv`. - `--trace-events-enabled` and six more `--trace-*` flags: declared on purpose, see the comment above them. - `--grep`: an alias of `--test-name-pattern`. - `--dump-environment-variables`: the only one with no reader and no stated reason. #### Checks - `test/bundler/cli.test.ts`: 38 pass, 0 fail. `test/bundler/bundler_env.test.ts`: 7 pass, 0 fail. `test/cli/bun.test.ts`: 39 pass, 0 fail. `test/cli/run/env.test.ts`: 107 pass, 0 fail. All with `--timeout 120000`, because the machine was under heavy load. - `bash -n completions/bun.bash` passes. zsh was not available, so `completions/bun.zsh` was checked by reading. The new last line of the `_arguments` call ends with `' &&`, like the other calls in the file. - The debug binary has no `dump-environment-variables` string. The completion scripts are embedded with `include_bytes!`, so `bun completions` installs the edited scripts. - Release binary size in CI against `main`: no target grows. Four of the twelve targets are 2 to 4 KB smaller (FreeBSD x64 and aarch64, Windows x64 and aarch64). - Substitutes, run on bun 1.4.3: `bun --print process.env` prints the process variables and the `.env` values. `logLevel = "debug"` in `bunfig.toml` makes `bun build --env inline` print the `.env` files that it loads. </details>  --- **no test proof** · iteration 1 · the description declares no behaviour change, so there is no failing test to prove; the existing suite in CI is the check

robobun assigned Jarred-Sumner Sep 27, 2026

github-actions Bot added the claude label Sep 27, 2026

claude Bot reviewed Sep 27, 2026

View reviewed changes

robobun mentioned this pull request Oct 1, 2026

Remove the dead --dump-environment-variables flag from bun build #44370

Merged

robobun mentioned this pull request Oct 2, 2026

init: never write over a package.json that was not loaded #41807

Open

This was referenced Oct 11, 2026

init: do not set type or private in a package.json that was loaded #44992

Open

init: do not add @types/bun or typescript that dependencies or optionalDependencies lists #45836

Open

Jarred-Sumner mentioned this pull request Oct 19, 2026

CLI: fixes for bun run, flags, bunfig.toml, .env, --filter, bun init and completions #46007

Open

This was referenced Oct 20, 2026

Remove dead code from JSWebSocket, ArrayBuffer::alloc, and Subprocess #46140

Open

install: read a .npmrc in linear time with hash tables past 32 keys #46932

Open

Navigation Menu

Repository navigation

Remove dead code from bun init, bun create, bun_md, bun_install, and the socket and Postgres code - #44102

Remove dead code from bun init, bun create, bun_md, bun_install, and the socket and Postgres code#44102
robobun wants to merge 1 commit into
mainfrom
robobun/c5fd30c2/dead-code-sweep

robobun commented Sep 27, 2026 •

edited

Loading

Uh oh!

coderabbitai Bot commented Sep 27, 2026 •

edited

Loading

Uh oh!

robobun commented Sep 27, 2026 •

edited

Loading

Uh oh!

claude Bot left a comment

Uh oh!

robobun commented Sep 27, 2026

Uh oh!

robobun commented Sep 28, 2026

Uh oh!

Jarred-Sumner commented Sep 28, 2026

Uh oh!

robobun commented Sep 28, 2026

Uh oh!

This branch has not been deployed

Reviewers

Uh oh!

Assignees

Labels

Projects

Milestone

Development

Uh oh!

2 participants

Conversation

robobun commented Sep 27, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Problem

Fix

Background

Downsides

Removed, one group each (net lines)

Verification

Self-review

Open PRs

One deletion was wrong and is not in this PR

Found and left in place

Uh oh!

coderabbitai Bot commented Sep 27, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Walkthrough

Changes

Uh oh!

robobun commented Sep 27, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

claude Bot left a comment

Choose a reason for hiding this comment

Uh oh!

robobun commented Sep 27, 2026

Uh oh!

robobun commented Sep 28, 2026

Uh oh!

Jarred-Sumner commented Sep 28, 2026

Uh oh!

robobun commented Sep 28, 2026

Uh oh!

This branch has not been deployed

Reviewers

Uh oh!

Assignees

Labels

Projects

Milestone

Development

Uh oh!

2 participants

robobun commented Sep 27, 2026 •

edited

Loading

coderabbitai Bot commented Sep 27, 2026 •

edited

Loading

robobun commented Sep 27, 2026 •

edited

Loading