s-b-repo-rustsploit/plans/compile-time-improvement-plan.md

Plan: Improve Cargo Build/Run Compile Times

Context

Clean build takes 14m 39s (879s) across 431 compilation units. Incremental rebuilds are already fast (0.6s). The goal is to reduce clean/cold build times — critical for CI, fresh clones, and dependency updates.

The biggest bottlenecks (from cargo --timings):

• rustsploit final crate: 427s (361 source files compiled as one unit)
• aws-lc-sys: 317s (C library build for rustls crypto — pulled by reqwest & rustls)
• dbus: 116s (solely from btleplug — used by 1 file)
• tokio: 105s
• darling_core + strum_macros: 182s (solely from ratatui — used by 1 file)
• regex-automata (×2): 175s
• clap_builder: 76s
• h2: 71s
• serde_derive + async-trait: 135s
• libssh2-sys: 62s (C library for ssh2)
• hickory-proto: 60s (used by 1 file)

---

Changes (ordered by impact / risk)

1. Configure lld linker

Savings: ~30-60s | Risk: None | Effort: 2 min

lld is installed at /usr/bin/lld but not configured. The default GNU ld is slow for a 110K-line binary.

Create .cargo/config.toml:

[target.x86_64-unknown-linux-gnu]
linker = "clang"
rustflags = ["-C", "link-arg=-fuse-ld=lld"]

---

2. Switch rustls crypto from aws-lc-rs to ring

Savings: ~250-280s | Risk: Low | Effort: 5 min

aws-lc-sys (317s) compiles a massive C library via cmake. It's pulled in because rustls 0.23 defaults to aws-lc-rs. The ring backend is functionally equivalent and compiles in ~30-50s.

cargo tree -i aws-lc-sys confirms the chain: aws-lc-sys → aws-lc-rs → rustls → {reqwest, tokio-rustls, rustsploit}.

In Cargo.toml:

rustls = { version = "0.23", default-features = false, features = ["ring", "logging", "std", "tls12"] }

No source code changes — ring and aws-lc-rs expose the same rustls::crypto::CryptoProvider API.

File: Cargo.toml line 50

---

3. Feature-gate btleplug + ratatui + crossterm

Savings: ~300s | Risk: Medium | Effort: 30 min

These three crates are used by exactly one file: src/modules/exploits/bluetooth/wpair.rs. Their transitive cost:

Dep chain	Compile time
btleplug → dbus	116s
btleplug → async-trait	68s
ratatui → strum_macros	85s
ratatui → darling_core	97s
ratatui → ratatui-core	42s
crossterm	~15s

Confirmed via cargo tree -i dbus, cargo tree -i strum_macros, cargo tree -i darling_core — all solely from btleplug/ratatui.

Changes:

Cargo.toml — add features section, make deps optional:

[features]
default = []
bluetooth = ["dep:btleplug", "dep:ratatui", "dep:crossterm"]

btleplug = { version = "0.12", optional = true }
ratatui = { version = "0.30", optional = true }
crossterm = { version = "0.29", optional = true }

src/modules/exploits/bluetooth/mod.rs — gate the module:

#[cfg(feature = "bluetooth")]
pub mod wpair;

build.rs — skip bluetooth dir when feature is absent. In generate_dispatch() (or the find_modules walk), check env::var("CARGO_FEATURE_BLUETOOTH") and skip paths containing bluetooth/ when it's not set. This prevents the generated dispatch from referencing wpair::run when the module doesn't exist.

When bluetooth is needed: cargo build --features bluetooth or cargo run --features bluetooth.

---

4. Clean up tokio feature flags

Savings: ~5-10s | Risk: None | Effort: 2 min

Current line is redundant — "full" already includes every named feature plus extras like test-util:

tokio = { version = "1.51", features = ["full", "process", "fs", "io-std", "rt-multi-thread", "macros", "rt"] }

Replace with only what's actually used:

tokio = { version = "1.51", features = ["rt-multi-thread", "macros", "net", "io-util", "io-std", "fs", "process", "sync", "time", "signal"] }

File: Cargo.toml line 20

---

5. Feature-gate hickory DNS

Savings: ~60s | Risk: Low | Effort: 15 min

hickory-proto (60s) + hickory-client are used by exactly one file: src/modules/scanners/dns_recursion.rs.

[features]
dns = ["dep:hickory-client", "dep:hickory-proto"]

hickory-client = { version = "0.25", optional = true }
hickory-proto = { version = "0.25", optional = true }

Gate in the scanner's mod.rs with #[cfg(feature = "dns")] and update build.rs to skip the module when the feature is absent.

---

Files to modify

File	Changes
.cargo/config.toml	Create — lld linker config
Cargo.toml	rustls features, optional deps, [features] section, tokio cleanup
build.rs	Skip feature-gated module dirs during code generation
src/modules/exploits/bluetooth/mod.rs	#[cfg(feature = "bluetooth")] gate
Scanner mod.rs for dns_recursion	#[cfg(feature = "dns")] gate

---

Verification

1. cargo clean && cargo build --timings 2>&1 — compare total time to baseline 879s
2. cargo build --features bluetooth,dns --timings — verify full build still works
3. cargo run -- --help — verify binary starts correctly
4. cargo run — enter shell, run a non-bluetooth module (e.g. use scanners/port_scanner, set target 127.0.0.1, run) to confirm dispatch works
5. cargo build --features bluetooth — verify bluetooth module compiles and appears in list modules

Expected result: Clean build drops from ~879s to ~250-350s (60-70% reduction), with changes 1-3 providing the bulk of the savings.