By the numbers

Data collected on 2026-04-30 from commit f53b654a888 (default branch main).

A quantitative snapshot of the codebase. For qualitative architecture see Architecture; for history see Lore.

Size

The repository is dominated by the test suite and the compiler crates:

Area	Rust source files	Notes
tests/	~25,400	UI, codegen, MIR, incremental, run-make, rustdoc, debuginfo, …
src/tools/	~7,100	rustdoc, miri, clippy, rustfmt, rust-analyzer, tidy, compiletest, …
compiler/	~2,100	~70 rustc_* crates
library/	~2,000	std, core, alloc, proc_macro, test, plus stdarch / portable-simd subtrees

xychart-beta
    title "Rust source files by area"
    x-axis ["tests", "src/tools", "compiler", "library"]
    y-axis "Files" 0 --> 30000
    bar [25400, 7100, 2100, 2000]

A few size landmarks:

• compiler/rustc_driver_impl/src/lib.rs — ~65 KB (the driver entry, command-line dispatch, and most of the top-level orchestration)
• bootstrap.example.toml — ~50 KB of documented build options
• RELEASES.md — ~895 KB of release notes going back to 1.0
• src/bootstrap/bootstrap.py — ~54 KB pre-rustc Python runner
• triagebot.toml — ~53 KB of automation rules (assignments, ping groups, mentions)
• Cargo.lock — ~150 KB workspace lock

Activity

Commit volume on the default branch:

Window	Commits
Total (since project start)	hundreds of thousands
Since 2024-01-01	~76,600
Since 2025-04-01	~34,100
Last week	hundreds

(Counts include merge commits from the bors queue, which inflate "commits" relative to "PRs".)

The repository is in continuous flux: at any given moment there are typically dozens of PRs in the merge queue, hundreds open, and several rolling release branches.

Churn hotspots

Areas that change most frequently (over the last several months, by file edit count):

Area	Reason
compiler/rustc_middle/src/ty/	Type system core; touched by nearly every type system change
compiler/rustc_borrowck/src/	Active development on diagnostics + Polonius
compiler/rustc_hir_typeck/src/	Hot path for type-inference fixes
compiler/rustc_trait_selection/src/	Trait solver evolution
library/std/src/	Standard library API additions and bugfixes
tests/ui/	New tests added with most language-affecting PRs

Bot-attributed commits

The rust-lang/rust repository is heavily automated:

• Most commits on main are bors merge commits authored by bors@rust-lang.org. The actual change author appears in the commit body, not the author field.
• Other automation accounts that show up in history:
  • triagebot[bot] — assignments, label maintenance
  • renovate[bot] — npm/JS dep bumps for tooling (configured in .github/renovate.json5)
  • dependabot[bot] — security advisories on tooling deps
  • github-actions[bot] — release-train rollups, generated artifacts

Inline AI-assisted authoring (Copilot, Cursor, etc.) is not visible in git history; this section is a lower bound on automation in commits.

Complexity

The compiler is the most complex area. A few signals:

• rustc_middle — the architectural hub. Re-exports almost every important type and is imported by most other rustc_* crates.
• rustc_driver_impl — the largest crate by entry-file size; orchestrates compilation phases and command-line modes.
• rustc_codegen_llvm + rustc_codegen_ssa — the LLVM backend; ~5,000 lines of Rust calling into FFI bindings under rustc_llvm (which itself wraps the in-tree LLVM at src/llvm-project/).
• rustc_borrowck + rustc_mir_dataflow — the borrow checker and the dataflow framework it builds on.
• rustc_trait_selection — the canonical trait solver; being progressively replaced by rustc_next_trait_solver (the "new solver" feature gated under -Znext-solver).

Standard library complexity hotspots:

• library/core — iter, slice, num, intrinsics, str are the heaviest modules
• library/std — platform-conditional code under std/src/sys/ is the largest single area
• library/stdarch — generated SIMD intrinsics across many architectures (vendored as a subtree)

Dependencies

The workspace Cargo.toml lists ~45 published members. External crates.io dependencies are kept deliberately small in compiler/ and library/:

• library/ uses essentially zero external dependencies (only the in-tree runtime crates and compiler-builtins)
• compiler/ uses a curated set: tracing, bitflags, serde, thin-vec, smallvec, parking_lot, regex (in selected places), and FFI bindings (rustc_llvm, tikv-jemalloc-sys)
• New crates.io dependencies require team review and an entry in src/tools/tidy/src/extdeps.rs (allow-list)

Test coverage

The tests/ directory is the ground truth for compiler correctness:

Suite	Approx. test count	Purpose
ui/	14,000+	Compiler diagnostics and successful compilation
run-make/	400+	Cargo-/rustc-driven end-to-end builds
incremental/	200+	Caching correctness across edits
mir-opt/	600+	MIR pass output snapshots
codegen-llvm/	1,000+	Specific LLVM IR / asm patterns
assembly-llvm/	dozens	Final assembly output
debuginfo/	dozens	LLDB / GDB / CDB debugger interaction
rustdoc-html/	hundreds	HTML output of rustdoc
rustdoc-json/	dozens	JSON output of rustdoc
crashes/	hundreds	Known-bad inputs that previously ICE'd

The tests/crashes/ suite is unusual: it contains files that should currently trigger an ICE — once they no longer do, they get moved to tests/ui/ with a regression label.