Troubleshooting

Build Errors

standard_128_insecure not found (E0599)

error[E0599]: no function or associated item named `standard_128_insecure` found

Cause: Insecure config methods are gated behind #[cfg(any(test, debug_assertions, feature = "allow_insecure"))]. They don’t exist in release builds.

Fix: Use SecureConfig::secure_128().into_config() instead:

use nine65::params::secure_configs::SecureConfig;
let config = SecureConfig::secure_128().into_config();

light_rns not found

Cause: Method was renamed to light_rns_insecure().

Fix: Use FHEConfig::light_rns_insecure() (only available in test/debug builds).

Profile warnings for exact_transcendentals

warning: profiles for the non root package will be ignored

Cause: exact_transcendentals/Cargo.toml has its own [profile] section. Cargo ignores per-crate profiles in a workspace.

Impact: None. Harmless warning.

Test Failures

fhe_demo expect_eq("add", sum, a + b) fails

Cause: The demo doesn’t wrap a + b modulo t. When a + b >= t, BFV correctly wraps but the expected value doesn’t. Cosmetic demo bug only.

Impact: Demo display issue. Core FHE is correct.

NoiseExhausted after 2-3 multiplications

Cause: Normal behavior for symmetric mode without bootstrap. secure_128 exhausts at depth 2, secure_192 at depth 3.

Fix: Use AutoBootstrapEvaluator for deeper circuits:

let mut eval = AutoBootstrapEvaluator::new(&ctx, &boot, &bsk, &ksk, &evk, &config);
let result = eval.mul_auto(&ct1, &ct2)?;  // auto-bootstraps when needed

Auto-bootstrap never triggers (0 bootstraps)

Cause: This was a bug fixed in commit 50f221f. The budget.consume() error was being discarded.

Verify: Make sure you’re on a commit after 50f221f. Check ops/auto_bootstrap.rs line ~70 — it should say:

let budget_exhausted = self.budget.consume(NoiseOpType::MulCt, mul_cost).is_err();
if budget_exhausted || self.budget.should_bootstrap(self.trigger_permille) {

Not the old version:

let _ = self.budget.consume(NoiseOpType::MulCt, mul_cost);
if self.budget.should_bootstrap(self.trigger_permille) {

Runtime Issues

Tests are extremely slow

Cause: Running in debug mode (no --release flag).

Fix: Always use --release for test runs:

cargo test -p nine65 --lib --release    # correct
cargo test -p nine65 --lib              # 10-50x slower

11 compiler warnings

Cause: Unused variables in symmetric_bootstrap.rs and shadow_entropy_monitor.rs. Expected and harmless.

Impact: None. Zero errors is what matters.

cargo miri fails with “not available for stable”

Cause: Miri requires the nightly toolchain. See Rust Toolchain.

Fix:

rustup toolchain install nightly
rustup +nightly component add miri
cargo +nightly miri test -p nine65 --lib -- specific_test_name

Common Mistakes

Using float literals

NINE65 forbids all floating-point. If you add 3.14 or f64 anywhere, clippy (and eventually compilation) will reject it.

Fix: Use integer or rational arithmetic for everything.

Forgetting –exclude flags

# Wrong --- will try to build Python/WASM bindings and fail
cargo build --release --workspace

# Right
cargo build --release --workspace --exclude nine65-python --exclude nine65-wasm

Using insecure configs in production code

Test configs (test_fast_insecure, test_medium_insecure, light_rns_insecure) are blocked in release builds. This is intentional — they have inadequate security for real use.

Always use SecureConfig::secure_128() or higher for anything beyond testing.