perf-sentinel-core 0.4.0

//! Configuration parsing for `.perf-sentinel.toml`.
//!
//! Supports both the new sectioned format (`[thresholds]`, `[detection]`, `[green]`, `[daemon]`)
//! and the legacy flat format for backward compatibility.

use std::collections::HashMap;

use serde::Deserialize;

use std::time::Duration;

use crate::detect::Confidence;
use crate::score::carbon::DEFAULT_EMBODIED_CARBON_PER_REQUEST_GCO2;
use crate::score::cloud_energy::config::{CloudEnergyConfig, ServiceCloudConfig};
use crate::score::scaphandre::ScaphandreConfig;

/// Top-level configuration for perf-sentinel.
#[derive(Debug, Clone)]
#[allow(clippy::struct_excessive_bools)] // Config aggregates all toggles from .perf-sentinel.toml
pub struct Config {
    // --- Thresholds ---
    /// Maximum allowed critical N+1 SQL findings before quality gate fails.
    pub n_plus_one_sql_critical_max: u32,
    /// Maximum allowed warning+ N+1 HTTP findings before quality gate fails.
    pub n_plus_one_http_warning_max: u32,
    /// Maximum allowed I/O waste ratio before quality gate fails.
    pub io_waste_ratio_max: f64,

    // --- Detection ---
    /// N+1 detection threshold: minimum repeated similar queries to flag.
    pub n_plus_one_threshold: u32,
    /// Sliding window duration in milliseconds for N+1 detection.
    pub window_duration_ms: u64,
    /// Threshold in milliseconds above which an operation is considered slow.
    pub slow_query_threshold_ms: u64,
    /// Minimum occurrences of a slow template to flag as a finding.
    pub slow_query_min_occurrences: u32,
    /// Maximum child spans per parent before flagging excessive fanout.
    pub max_fanout: u32,
    /// Minimum HTTP outbound calls per trace to flag as chatty service.
    pub chatty_service_min_calls: u32,
    /// Peak concurrent SQL spans per service to flag pool saturation.
    pub pool_saturation_concurrent_threshold: u32,
    /// Minimum sequential independent sibling calls to flag as serialized.
    pub serialized_min_sequential: u32,

    // --- Green ---
    /// Whether `GreenOps` scoring is enabled.
    pub green_enabled: bool,
    /// Fallback region for CO₂ scoring (e.g. `"eu-west-3"`).
    pub green_default_region: Option<String>,
    /// Per-service region overrides. Keys lowercased at load time.
    pub green_service_regions: HashMap<String, String>,
    /// SCI `M` term: embodied carbon per request (gCO₂eq).
    pub green_embodied_carbon_per_request_gco2: f64,
    /// Use 24-hour carbon intensity profiles when available.
    pub green_use_hourly_profiles: bool,
    /// Scaphandre RAPL scraper config (daemon only).
    pub green_scaphandre: Option<ScaphandreConfig>,
    /// Cloud CPU% + `SPECpower` config (daemon only).
    pub green_cloud_energy: Option<CloudEnergyConfig>,
    /// Whether to use per-operation energy coefficients (SQL verb weighting,
    /// HTTP payload size tiers) in the proxy model. Default: `true`.
    pub green_per_operation_coefficients: bool,
    /// Whether to compute a network transport energy term for cross-region
    /// HTTP calls. Default: `false` (opt-in).
    pub green_include_network_transport: bool,
    /// Energy per byte for network transport (kWh/byte).
    /// Default: 0.04 kWh/GB (Mytton et al. 2024).
    pub green_network_energy_per_byte_kwh: f64,
    /// Path to user-supplied hourly profiles JSON file. `None` when not
    /// configured (uses only embedded profiles).
    pub green_hourly_profiles_file: Option<String>,
    /// Pre-parsed custom hourly profiles, loaded at config parse time.
    /// `None` when `hourly_profiles_file` is not set or failed to load.
    pub green_custom_hourly_profiles:
        Option<std::sync::Arc<HashMap<String, crate::score::carbon::HourlyProfile>>>,
    /// Path to a calibration TOML file generated by `perf-sentinel calibrate`.
    pub green_calibration_file: Option<String>,
    /// Pre-loaded calibration data, parsed at config load time.
    /// `None` when `calibration_file` is not set or failed to load.
    pub green_calibration: Option<crate::calibrate::CalibrationData>,
    /// Electricity Maps real-time carbon intensity config (daemon only).
    pub green_electricity_maps: Option<crate::score::electricity_maps::ElectricityMapsConfig>,

    // --- Daemon ---
    /// Address for the daemon to listen on.
    pub listen_addr: String,
    /// Port for OTLP HTTP receiver.
    pub listen_port: u16,
    /// Port for OTLP gRPC receiver.
    pub listen_port_grpc: u16,
    /// Unix socket path for JSON ingestion.
    pub json_socket: String,
    /// Maximum number of active traces in streaming mode.
    pub max_active_traces: usize,
    /// Trace TTL in milliseconds for streaming mode eviction.
    pub trace_ttl_ms: u64,
    /// Sampling rate for incoming traces (0.0 - 1.0).
    pub sampling_rate: f64,
    /// Maximum events kept per trace (ring buffer size).
    pub max_events_per_trace: usize,
    /// Maximum payload size in bytes for JSON deserialization.
    pub max_payload_size: usize,
    /// Deployment environment label used by the daemon to stamp findings
    /// with a [`Confidence`] value. Defaults to
    /// [`DaemonEnvironment::Staging`]; set to
    /// [`DaemonEnvironment::Production`] when running on production traffic
    /// so downstream consumers (perf-lint) can boost severity. Ignored in
    /// `analyze` batch mode, which always emits [`Confidence::CiBatch`].
    pub daemon_environment: DaemonEnvironment,
    /// Path to PEM-encoded TLS certificate chain for the OTLP receivers.
    /// When set alongside [`tls_key_path`], both gRPC and HTTP listeners
    /// use TLS. When absent, listeners use plain TCP (default).
    pub tls_cert_path: Option<String>,
    /// Path to PEM-encoded TLS private key for the OTLP receivers.
    pub tls_key_path: Option<String>,
    /// Maximum number of findings retained by the daemon query API.
    pub max_retained_findings: usize,
    /// Whether the daemon query API is enabled.
    pub daemon_api_enabled: bool,
    /// Whether cross-trace correlation is enabled (opt-in, default false).
    pub correlation_enabled: bool,
    /// Cross-trace correlation config. Only used when `correlation_enabled` is true.
    pub correlation_config: crate::detect::correlate_cross::CorrelationConfig,
}

/// Deployment environment for the daemon's `watch` mode.
///
/// Maps 1:1 to [`Confidence`] via [`Config::confidence`]:
/// - [`Self::Staging`] → [`Confidence::DaemonStaging`]
/// - [`Self::Production`] → [`Confidence::DaemonProduction`]
///
/// Parsed from the `[daemon] environment` TOML field as case-insensitive
/// `"staging"` or `"production"`. Any other value is rejected at load time
/// with a clear validation error.
#[derive(Debug, Clone, Copy, PartialEq, Eq, Default)]
pub enum DaemonEnvironment {
    /// Staging traffic — medium confidence. Default.
    #[default]
    Staging,
    /// Production traffic — high confidence.
    Production,
}

impl DaemonEnvironment {
    /// Returns the lowercase string label used in the TOML config.
    #[must_use]
    pub const fn as_str(&self) -> &'static str {
        match self {
            Self::Staging => "staging",
            Self::Production => "production",
        }
    }
}

impl Default for Config {
    fn default() -> Self {
        Self {
            // Thresholds
            n_plus_one_sql_critical_max: 0,
            n_plus_one_http_warning_max: 3,
            io_waste_ratio_max: 0.30,
            // Detection
            n_plus_one_threshold: 5,
            window_duration_ms: 500,
            slow_query_threshold_ms: 500,
            slow_query_min_occurrences: 3,
            max_fanout: 20,
            chatty_service_min_calls: 15,
            pool_saturation_concurrent_threshold: 10,
            serialized_min_sequential: 3,
            // Green
            green_enabled: true,
            green_default_region: None,
            green_service_regions: HashMap::new(),
            green_embodied_carbon_per_request_gco2: DEFAULT_EMBODIED_CARBON_PER_REQUEST_GCO2,
            green_use_hourly_profiles: true,
            green_scaphandre: None,
            green_cloud_energy: None,
            green_per_operation_coefficients: true,
            green_include_network_transport: false,
            green_network_energy_per_byte_kwh:
                crate::score::carbon::DEFAULT_NETWORK_ENERGY_PER_BYTE_KWH,
            green_hourly_profiles_file: None,
            green_custom_hourly_profiles: None,
            green_calibration_file: None,
            green_calibration: None,
            green_electricity_maps: None,
            // Daemon
            listen_addr: "127.0.0.1".to_string(),
            listen_port: 4318,
            listen_port_grpc: 4317,
            json_socket: "/tmp/perf-sentinel.sock".to_string(),
            max_active_traces: 10_000,
            trace_ttl_ms: 30_000,
            sampling_rate: 1.0,
            max_events_per_trace: 1_000,
            max_payload_size: 1_048_576, // 1 MB
            daemon_environment: DaemonEnvironment::Staging,
            tls_cert_path: None,
            tls_key_path: None,
            max_retained_findings: 10_000,
            daemon_api_enabled: true,
            correlation_enabled: false,
            correlation_config: crate::detect::correlate_cross::CorrelationConfig::default(),
        }
    }
}

impl Config {
    /// Map the daemon environment to a [`Confidence`] value.
    ///
    /// Used by `daemon::run` to stamp findings after detection. `analyze`
    /// batch mode does not call this — it hardcodes [`Confidence::CiBatch`]
    /// in `pipeline::analyze_with_traces` instead.
    #[must_use]
    pub const fn confidence(&self) -> Confidence {
        match self.daemon_environment {
            DaemonEnvironment::Staging => Confidence::DaemonStaging,
            DaemonEnvironment::Production => Confidence::DaemonProduction,
        }
    }

    /// Build a [`CarbonContext`] from the green config fields.
    ///
    /// Returns a context with `energy_snapshot: None`. The daemon clones
    /// this and patches in the measured energy snapshot per tick; the
    /// batch pipeline uses it as-is (no scrapers in batch mode).
    #[must_use]
    pub fn carbon_context(&self) -> crate::score::carbon::CarbonContext {
        crate::score::carbon::CarbonContext {
            default_region: self.green_default_region.clone(),
            service_regions: self.green_service_regions.clone(),
            embodied_per_request_gco2: self.green_embodied_carbon_per_request_gco2,
            use_hourly_profiles: self.green_use_hourly_profiles,
            energy_snapshot: None,
            per_operation_coefficients: self.green_per_operation_coefficients,
            include_network_transport: self.green_include_network_transport,
            network_energy_per_byte_kwh: self.green_network_energy_per_byte_kwh,
            custom_hourly_profiles: self.green_custom_hourly_profiles.clone(),
            calibration: self.green_calibration.clone(),
            real_time_intensity: None, // set per-tick in daemon via build_tick_ctx
        }
    }
}

// --- Internal raw deserialization types ---

#[derive(Deserialize, Default)]
#[serde(default)]
struct RawConfig {
    // Sections (new format)
    thresholds: ThresholdsSection,
    detection: DetectionSection,
    green: GreenSection,
    daemon: DaemonSection,

    // Legacy flat fields (backward compatibility)
    max_payload_size: Option<usize>,
    n_plus_one_threshold: Option<u32>,
    listen_addr: Option<String>,
    listen_port: Option<u16>,
    window_duration_ms: Option<u64>,
    trace_ttl_ms: Option<u64>,
    max_active_traces: Option<usize>,
    max_events_per_trace: Option<usize>,
}

#[derive(Deserialize, Default)]
#[serde(default)]
#[allow(clippy::struct_field_names)] // fields like `n_plus_one_sql_critical_max` repeat the struct context but match the TOML keys
struct ThresholdsSection {
    n_plus_one_sql_critical_max: Option<u32>,
    n_plus_one_http_warning_max: Option<u32>,
    io_waste_ratio_max: Option<f64>,
}

#[derive(Deserialize, Default)]
#[serde(default)]
struct DetectionSection {
    window_duration_ms: Option<u64>,
    n_plus_one_min_occurrences: Option<u32>,
    slow_query_threshold_ms: Option<u64>,
    slow_query_min_occurrences: Option<u32>,
    max_fanout: Option<u32>,
    chatty_service_min_calls: Option<u32>,
    pool_saturation_concurrent_threshold: Option<u32>,
    serialized_min_sequential: Option<u32>,
}

#[derive(Deserialize, Default)]
#[serde(default)]
struct GreenSection {
    enabled: Option<bool>,
    default_region: Option<String>,
    service_regions: HashMap<String, String>,
    embodied_carbon_per_request_gco2: Option<f64>,
    /// toggle for the hourly carbon intensity profile path.
    /// Default `true`. Maps to `Config::green_use_hourly_profiles`.
    use_hourly_profiles: Option<bool>,
    /// Scaphandre scraper section. Absent when Scaphandre
    /// is not configured.
    scaphandre: ScaphandreSection,
    /// Cloud energy section. Absent when cloud energy is not configured.
    cloud: CloudSection,
    per_operation_coefficients: Option<bool>,
    include_network_transport: Option<bool>,
    network_energy_per_byte_kwh: Option<f64>,
    /// Path to a JSON file with user-supplied hourly carbon profiles.
    hourly_profiles_file: Option<String>,
    /// Path to a calibration TOML file from `perf-sentinel calibrate`.
    calibration_file: Option<String>,
    /// Electricity Maps API section.
    electricity_maps: ElectricityMapsSection,
}

/// Raw deserialization target for `[green.scaphandre]`.
///
/// Converted to a `ScaphandreConfig` during `RawConfig → Config` only
/// when `endpoint` is set — an empty table (no fields) leaves
/// `Config::green_scaphandre = None`.
#[derive(Deserialize, Default)]
#[serde(default)]
struct ScaphandreSection {
    endpoint: Option<String>,
    scrape_interval_secs: Option<u64>,
    process_map: HashMap<String, String>,
}

/// Raw deserialization target for `[green.cloud]`.
///
/// Converted to a `CloudEnergyConfig` during `RawConfig -> Config` only
/// when `prometheus_endpoint` is set.
#[derive(Deserialize, Default)]
#[serde(default)]
struct CloudSection {
    prometheus_endpoint: Option<String>,
    scrape_interval_secs: Option<u64>,
    default_provider: Option<String>,
    default_instance_type: Option<String>,
    cpu_metric: Option<String>,
    services: HashMap<String, CloudServiceRaw>,
}

/// Raw deserialization for a single entry in `[green.cloud.services]`.
///
/// Supports two forms:
/// - Instance type: `{ provider = "aws", instance_type = "m5.large" }`
/// - Manual watts: `{ idle_watts = 45, max_watts = 120 }`
#[derive(Deserialize, Default)]
#[serde(default)]
struct CloudServiceRaw {
    provider: Option<String>,
    instance_type: Option<String>,
    idle_watts: Option<f64>,
    max_watts: Option<f64>,
    cpu_query: Option<String>,
}

/// Raw deserialization target for `[green.electricity_maps]`.
///
/// Converted to an `ElectricityMapsConfig` during `RawConfig -> Config`
/// only when `api_key` is set (directly or via env var).
#[derive(Deserialize, Default)]
#[serde(default)]
struct ElectricityMapsSection {
    api_key: Option<String>,
    endpoint: Option<String>,
    poll_interval_secs: Option<u64>,
    region_map: HashMap<String, String>,
}

#[derive(Deserialize, Default)]
#[serde(default)]
struct DaemonSection {
    listen_address: Option<String>,
    listen_port_http: Option<u16>,
    listen_port_grpc: Option<u16>,
    json_socket: Option<String>,
    max_active_traces: Option<usize>,
    trace_ttl_ms: Option<u64>,
    sampling_rate: Option<f64>,
    max_events_per_trace: Option<usize>,
    max_payload_size: Option<usize>,
    /// `"staging"` (default) or `"production"`. Validated
    /// in `Config::validate`; invalid values fail at load time with a
    /// clear error. Case-insensitive.
    environment: Option<String>,
    /// Path to PEM-encoded TLS certificate chain.
    tls_cert_path: Option<String>,
    /// Path to PEM-encoded TLS private key.
    tls_key_path: Option<String>,
    /// Maximum number of findings kept by the daemon query API.
    max_retained_findings: Option<usize>,
    /// Whether the daemon query API is enabled.
    api_enabled: Option<bool>,
    /// Cross-trace correlation section.
    correlation: CorrelationSection,
}

/// Raw deserialization target for `[daemon.correlation]`.
#[derive(Deserialize, Default)]
#[serde(default)]
struct CorrelationSection {
    enabled: Option<bool>,
    window_minutes: Option<u64>,
    lag_threshold_ms: Option<u64>,
    min_co_occurrences: Option<u32>,
    min_confidence: Option<f64>,
    max_tracked_pairs: Option<usize>,
}

impl From<RawConfig> for Config {
    #[allow(clippy::too_many_lines)] // Flat config-to-typed mapping: splitting would scatter field assignments
    fn from(raw: RawConfig) -> Self {
        let defaults = Self::default();

        // Sections take priority over flat fields, flat fields over defaults.
        Self {
            // Thresholds
            n_plus_one_sql_critical_max: raw
                .thresholds
                .n_plus_one_sql_critical_max
                .unwrap_or(defaults.n_plus_one_sql_critical_max),
            n_plus_one_http_warning_max: raw
                .thresholds
                .n_plus_one_http_warning_max
                .unwrap_or(defaults.n_plus_one_http_warning_max),
            io_waste_ratio_max: raw
                .thresholds
                .io_waste_ratio_max
                .unwrap_or(defaults.io_waste_ratio_max),

            // Detection: section > flat > default
            n_plus_one_threshold: raw
                .detection
                .n_plus_one_min_occurrences
                .or(raw.n_plus_one_threshold)
                .unwrap_or(defaults.n_plus_one_threshold),
            window_duration_ms: raw
                .detection
                .window_duration_ms
                .or(raw.window_duration_ms)
                .unwrap_or(defaults.window_duration_ms),
            slow_query_threshold_ms: raw
                .detection
                .slow_query_threshold_ms
                .unwrap_or(defaults.slow_query_threshold_ms),
            slow_query_min_occurrences: raw
                .detection
                .slow_query_min_occurrences
                .unwrap_or(defaults.slow_query_min_occurrences),
            max_fanout: raw.detection.max_fanout.unwrap_or(defaults.max_fanout),
            chatty_service_min_calls: raw
                .detection
                .chatty_service_min_calls
                .unwrap_or(defaults.chatty_service_min_calls),
            pool_saturation_concurrent_threshold: raw
                .detection
                .pool_saturation_concurrent_threshold
                .unwrap_or(defaults.pool_saturation_concurrent_threshold),
            serialized_min_sequential: raw
                .detection
                .serialized_min_sequential
                .unwrap_or(defaults.serialized_min_sequential),

            // Green
            green_enabled: raw.green.enabled.unwrap_or(defaults.green_enabled),
            green_default_region: raw.green.default_region,
            // Lowercase service_regions keys so resolve_region's
            // lowercase lookup matches regardless of config casing.
            green_service_regions: raw
                .green
                .service_regions
                .into_iter()
                .map(|(k, v)| (k.to_ascii_lowercase(), v))
                .collect(),
            green_embodied_carbon_per_request_gco2: raw
                .green
                .embodied_carbon_per_request_gco2
                .unwrap_or(defaults.green_embodied_carbon_per_request_gco2),
            green_use_hourly_profiles: raw
                .green
                .use_hourly_profiles
                .unwrap_or(defaults.green_use_hourly_profiles),
            green_scaphandre: raw.green.scaphandre.endpoint.as_ref().map(|endpoint| {
                ScaphandreConfig {
                    endpoint: endpoint.clone(),
                    // Default scrape interval 5s; clamped in validate_green
                    // to the [1, 3600] range.
                    scrape_interval: Duration::from_secs(
                        raw.green.scaphandre.scrape_interval_secs.unwrap_or(5),
                    ),
                    process_map: raw.green.scaphandre.process_map.clone(),
                }
            }),
            green_cloud_energy: convert_cloud_section(&raw.green.cloud),
            green_per_operation_coefficients: raw
                .green
                .per_operation_coefficients
                .unwrap_or(defaults.green_per_operation_coefficients),
            green_include_network_transport: raw
                .green
                .include_network_transport
                .unwrap_or(defaults.green_include_network_transport),
            green_network_energy_per_byte_kwh: raw
                .green
                .network_energy_per_byte_kwh
                .unwrap_or(defaults.green_network_energy_per_byte_kwh),
            green_hourly_profiles_file: raw.green.hourly_profiles_file.clone(),
            green_custom_hourly_profiles: raw.green.hourly_profiles_file.as_ref().and_then(
                |path| {
                    if has_control_char(path) {
                        tracing::warn!(
                            "hourly_profiles_file path contains control characters, skipping"
                        );
                        return None;
                    }
                    let p = std::path::Path::new(path);
                    match crate::score::carbon::load_custom_profiles(p) {
                        Ok(profiles) => Some(std::sync::Arc::new(profiles)),
                        Err(e) => {
                            // Not logged at warn: validate_green() will
                            // surface a hard error for this case.
                            tracing::debug!(
                                error = %e,
                                "Custom hourly profiles failed to load"
                            );
                            None
                        }
                    }
                },
            ),

            green_calibration_file: raw.green.calibration_file.clone(),
            green_calibration: raw.green.calibration_file.as_ref().and_then(|path| {
                if has_control_char(path) {
                    tracing::warn!("calibration_file path contains control characters, skipping");
                    return None;
                }
                match crate::calibrate::load_calibration_file(path) {
                    Ok(data) => Some(data),
                    Err(e) => {
                        tracing::debug!(
                            error = %e,
                            "Calibration file failed to load"
                        );
                        None
                    }
                }
            }),
            green_electricity_maps: convert_electricity_maps_section(&raw.green.electricity_maps),

            // Daemon: section > flat > default
            listen_addr: raw
                .daemon
                .listen_address
                .or(raw.listen_addr)
                .unwrap_or(defaults.listen_addr),
            listen_port: raw
                .daemon
                .listen_port_http
                .or(raw.listen_port)
                .unwrap_or(defaults.listen_port),
            listen_port_grpc: raw
                .daemon
                .listen_port_grpc
                .unwrap_or(defaults.listen_port_grpc),
            json_socket: raw.daemon.json_socket.unwrap_or(defaults.json_socket),
            max_active_traces: raw
                .daemon
                .max_active_traces
                .or(raw.max_active_traces)
                .unwrap_or(defaults.max_active_traces),
            trace_ttl_ms: raw
                .daemon
                .trace_ttl_ms
                .or(raw.trace_ttl_ms)
                .unwrap_or(defaults.trace_ttl_ms),
            sampling_rate: raw.daemon.sampling_rate.unwrap_or(defaults.sampling_rate),
            max_events_per_trace: raw
                .daemon
                .max_events_per_trace
                .or(raw.max_events_per_trace)
                .unwrap_or(defaults.max_events_per_trace),
            max_payload_size: raw
                .daemon
                .max_payload_size
                .or(raw.max_payload_size)
                .unwrap_or(defaults.max_payload_size),
            // parse environment into the typed enum. Invalid
            // strings are rejected later in Config::validate so parse
            // Fallback to Staging is safe: load_from_str() pre-validates
            // this field before calling Config::from(). Direct callers
            // (only tests) get Staging as a safe default.
            daemon_environment: match raw.daemon.environment.as_deref() {
                None => defaults.daemon_environment,
                Some(s) => parse_daemon_environment(s).unwrap_or(DaemonEnvironment::Staging),
            },
            tls_cert_path: raw.daemon.tls_cert_path,
            tls_key_path: raw.daemon.tls_key_path,
            max_retained_findings: raw
                .daemon
                .max_retained_findings
                .unwrap_or(defaults.max_retained_findings),
            daemon_api_enabled: raw
                .daemon
                .api_enabled
                .unwrap_or(defaults.daemon_api_enabled),
            correlation_enabled: raw
                .daemon
                .correlation
                .enabled
                .unwrap_or(defaults.correlation_enabled),
            correlation_config: {
                let c = &raw.daemon.correlation;
                let d = crate::detect::correlate_cross::CorrelationConfig::default();
                crate::detect::correlate_cross::CorrelationConfig {
                    window_ms: c
                        .window_minutes
                        .map_or(d.window_ms, |m| m.saturating_mul(60_000)),
                    lag_threshold_ms: c.lag_threshold_ms.unwrap_or(d.lag_threshold_ms),
                    min_co_occurrences: c.min_co_occurrences.unwrap_or(d.min_co_occurrences),
                    min_confidence: c.min_confidence.unwrap_or(d.min_confidence),
                    max_tracked_pairs: c.max_tracked_pairs.unwrap_or(d.max_tracked_pairs),
                }
            },
        }
    }
}

/// Parse a case-insensitive environment string into [`DaemonEnvironment`].
///
/// Returns `None` for any value that is not `"staging"` or `"production"`.
/// Called from [`Config::from`] (which falls back to default on error,
/// deferring the real rejection to [`Config::validate`]).
fn parse_daemon_environment(value: &str) -> Option<DaemonEnvironment> {
    match value.trim().to_ascii_lowercase().as_str() {
        "staging" => Some(DaemonEnvironment::Staging),
        "production" => Some(DaemonEnvironment::Production),
        _ => None,
    }
}

/// Convert the raw `[green.cloud]` TOML section into a typed config.
///
/// Returns `None` when `prometheus_endpoint` is absent (section empty
/// or not present). Per-service entries are classified as either
/// `InstanceType` or `ManualWatts` based on which fields are set.
fn convert_cloud_section(raw: &CloudSection) -> Option<CloudEnergyConfig> {
    let endpoint = raw.prometheus_endpoint.as_ref()?;
    let mut services = HashMap::with_capacity(raw.services.len());
    for (name, svc) in &raw.services {
        let config = if svc.idle_watts.is_some() || svc.max_watts.is_some() {
            // Manual watts mode: both must be present (validated later).
            ServiceCloudConfig::ManualWatts {
                idle_watts: svc.idle_watts.unwrap_or(0.0),
                max_watts: svc.max_watts.unwrap_or(0.0),
                cpu_query: svc.cpu_query.clone(),
            }
        } else {
            ServiceCloudConfig::InstanceType {
                provider: svc.provider.clone(),
                instance_type: svc.instance_type.clone().unwrap_or_default(),
                cpu_query: svc.cpu_query.clone(),
            }
        };
        services.insert(name.clone(), config);
    }
    Some(CloudEnergyConfig {
        prometheus_endpoint: endpoint.clone(),
        scrape_interval: Duration::from_secs(raw.scrape_interval_secs.unwrap_or(15)),
        default_provider: raw.default_provider.clone(),
        default_instance_type: raw.default_instance_type.clone(),
        cpu_metric: raw.cpu_metric.clone(),
        services,
    })
}

/// Convert the raw `[green.electricity_maps]` TOML section into a typed config.
///
/// Returns `None` when no `api_key` is set (neither in config nor env var).
fn convert_electricity_maps_section(
    raw: &ElectricityMapsSection,
) -> Option<crate::score::electricity_maps::ElectricityMapsConfig> {
    convert_electricity_maps_section_with_env(raw, || {
        std::env::var("PERF_SENTINEL_EMAPS_TOKEN").ok()
    })
}

/// Test-friendly inner form: takes the env-var lookup as a closure so tests
/// can pass `|| None` instead of mutating the global process env. Avoids the
/// `unsafe` that Rust 2024 requires on `std::env::remove_var` (`set_var` and
/// `remove_var` are data races with other threads inside the same process,
/// including the `cargo test` harness).
fn convert_electricity_maps_section_with_env(
    raw: &ElectricityMapsSection,
    env_lookup: impl FnOnce() -> Option<String>,
) -> Option<crate::score::electricity_maps::ElectricityMapsConfig> {
    // Auth token: env var takes precedence over config file.
    let from_env = env_lookup();
    let token = from_env.clone().or_else(|| raw.api_key.clone())?;

    if token.is_empty() {
        return None;
    }

    // Nudge users toward the env var when the token is in the config file.
    if from_env.is_none() && raw.api_key.is_some() {
        tracing::warn!(
            "[green.electricity_maps] api_key is set in the config file. \
             Prefer the PERF_SENTINEL_EMAPS_TOKEN environment variable \
             to avoid committing secrets to version control."
        );
    }

    let poll_secs = raw.poll_interval_secs.unwrap_or(300);
    let api_endpoint = raw
        .endpoint
        .clone()
        .unwrap_or_else(|| "https://api.electricitymaps.com/v3".to_string());
    Some(crate::score::electricity_maps::ElectricityMapsConfig {
        api_endpoint,
        auth_token: token,
        poll_interval: Duration::from_secs(poll_secs),
        // Lowercase region keys so scoring loop lookups match regardless
        // of config casing (same pattern as service_regions).
        region_map: raw
            .region_map
            .iter()
            .map(|(k, v)| (k.to_ascii_lowercase(), v.clone()))
            .collect(),
    })
}

fn check_range<T: PartialOrd + std::fmt::Display>(
    name: &str,
    val: &T,
    min: &T,
    max: &T,
) -> Result<(), String> {
    if val < min {
        return Err(format!("{name} must be >= {min}, got {val}"));
    }
    if val > max {
        return Err(format!("{name} must be <= {max}, got {val}"));
    }
    Ok(())
}

fn check_min<T: PartialOrd + std::fmt::Display>(
    name: &str,
    val: &T,
    min: &T,
) -> Result<(), String> {
    if val < min {
        return Err(format!("{name} must be >= {min}, got {val}"));
    }
    Ok(())
}

/// `true` if `s` contains any ASCII control character (< 0x20 or DEL).
fn has_control_char(s: &str) -> bool {
    s.bytes().any(|b| b < 0x20 || b == 0x7F)
}

/// Validate the authority portion of an HTTP(S) URI.
/// Rejects credentials, empty host, control characters, and invalid port.
/// Handles IPv6 bracket notation (`[::1]`, `[::1]:8080`).
fn validate_http_authority(url: &str, label: &str) -> Result<(), String> {
    let after_scheme = url
        .strip_prefix("https://")
        .or_else(|| url.strip_prefix("http://"))
        .unwrap_or(url);
    let authority = after_scheme.split('/').next().unwrap_or(after_scheme);
    if authority.is_empty() {
        return Err(format!("{label} '{url}' has no host"));
    }
    if authority.contains('@') {
        return Err(format!(
            "{label} must not contain credentials (userinfo): '{url}'"
        ));
    }
    if has_control_char(authority) {
        return Err(format!("{label} '{url}' contains control characters"));
    }
    // Port validation: skip for bare IPv6 without port (`[::1]`), handle
    // bracketed IPv6 with port (`[::1]:8080`) via the `]:` delimiter.
    if authority.starts_with('[') {
        // IPv6 bracket notation: port follows `]:` if present.
        if let Some(bracket_end) = authority.find(']') {
            let after_bracket = &authority[bracket_end + 1..];
            if let Some(port_str) = after_bracket.strip_prefix(':')
                && !port_str.is_empty()
                && port_str.parse::<u16>().is_err()
            {
                return Err(format!("{label} '{url}' has an invalid port"));
            }
        }
    } else if let Some(port_str) = authority.rsplit(':').next()
        && authority.contains(':')
        && port_str.parse::<u16>().is_err()
    {
        return Err(format!("{label} '{url}' has an invalid port"));
    }
    Ok(())
}

impl Config {
    /// Validate that config values are within acceptable bounds.
    ///
    /// # Errors
    ///
    /// Returns a `String` description of the first invalid value found.
    /// The caller (`load_from_str`) wraps this in `ConfigError::Validation`.
    pub fn validate(&self) -> Result<(), String> {
        self.validate_daemon_limits()?;
        self.validate_detection_params()?;
        self.validate_rates()?;
        self.validate_listen_addr()?;
        self.validate_tls()?;
        self.validate_green()?;
        Ok(())
    }

    /// Validate TLS configuration: both paths must be set or both absent.
    /// When set, verify the files exist and warn if the key is
    /// world-readable on Unix.
    fn validate_tls(&self) -> Result<(), String> {
        match (&self.tls_cert_path, &self.tls_key_path) {
            (Some(cert), Some(key)) => {
                if has_control_char(cert) {
                    return Err("[daemon] tls_cert_path contains control characters".to_string());
                }
                if has_control_char(key) {
                    return Err("[daemon] tls_key_path contains control characters".to_string());
                }
                if !std::path::Path::new(cert).exists() {
                    return Err(format!("[daemon] tls_cert_path '{cert}' does not exist"));
                }
                if !std::path::Path::new(key).exists() {
                    return Err(format!("[daemon] tls_key_path '{key}' does not exist"));
                }
                #[cfg(unix)]
                {
                    use std::os::unix::fs::PermissionsExt;
                    if let Ok(meta) = std::fs::metadata(key) {
                        let mode = meta.permissions().mode();
                        if mode & 0o077 != 0 {
                            tracing::warn!(
                                "TLS key file '{key}' is readable by group/others \
                                 (mode {mode:o}). Consider restricting to owner-only \
                                 (chmod 600)."
                            );
                        }
                    }
                }
                tracing::info!("TLS enabled for daemon OTLP receivers (cert: {cert})");
                Ok(())
            }
            (None, None) => Ok(()),
            (Some(_), None) => {
                Err("[daemon] tls_cert_path is set but tls_key_path is missing".to_string())
            }
            (None, Some(_)) => {
                Err("[daemon] tls_key_path is set but tls_cert_path is missing".to_string())
            }
        }
    }

    fn validate_green(&self) -> Result<(), String> {
        Self::validate_embodied_carbon(self.green_embodied_carbon_per_request_gco2)?;
        Self::validate_default_region(self.green_default_region.as_deref())?;
        Self::validate_service_regions(&self.green_service_regions)?;
        if let Some(cfg) = &self.green_scaphandre {
            Self::validate_scaphandre(cfg)?;
        }
        if let Some(cfg) = &self.green_cloud_energy {
            Self::validate_cloud_energy(cfg)?;
        }
        Self::validate_network_energy(self.green_network_energy_per_byte_kwh)?;
        self.validate_hourly_profiles_file()?;
        if let Some(cfg) = &self.green_electricity_maps {
            Self::validate_electricity_maps(cfg)?;
        }
        Ok(())
    }

    fn validate_embodied_carbon(value: f64) -> Result<(), String> {
        if !value.is_finite() {
            return Err(format!(
                "embodied_carbon_per_request_gco2 must be finite, got {value}"
            ));
        }
        if value < 0.0 {
            return Err(format!(
                "embodied_carbon_per_request_gco2 must be >= 0.0, got {value}"
            ));
        }
        Ok(())
    }

    /// Validate the optional `[green] default_region`. Config is trusted
    /// input, so typos surface loudly here rather than silently producing
    /// zeroed CO₂ rows downstream. Same validator used at the OTLP
    /// ingestion boundary (there, invalid values are silently dropped).
    fn validate_default_region(region: Option<&str>) -> Result<(), String> {
        let Some(region) = region else {
            return Ok(());
        };
        if crate::score::carbon::is_valid_region_id(region) {
            return Ok(());
        }
        Err(format!(
            "[green] default_region '{region}' contains invalid characters; \
             expected ASCII alphanumeric + '-' or '_', length 1-64"
        ))
    }

    /// Validate the `[green.service_regions]` map: cardinality cap, plus
    /// region-id syntax on every key/value pair.
    fn validate_service_regions(map: &HashMap<String, String>) -> Result<(), String> {
        /// Maximum number of entries in `[green.service_regions]`.
        /// Bounds the config-load memory footprint against fat-finger or
        /// malicious configs. 1024 is 4× `MAX_REGIONS` (256) and comfortably
        /// above any realistic multi-cloud deployment size.
        const MAX_SERVICE_REGIONS: usize = 1024;
        if map.len() > MAX_SERVICE_REGIONS {
            return Err(format!(
                "[green.service_regions] has {} entries; maximum is {MAX_SERVICE_REGIONS}",
                map.len()
            ));
        }
        for (service, region) in map {
            if !crate::score::carbon::is_valid_region_id(service) {
                return Err(format!(
                    "[green.service_regions] invalid service name '{service}'; \
                     expected ASCII alphanumeric + '-' or '_', length 1-64"
                ));
            }
            if !crate::score::carbon::is_valid_region_id(region) {
                return Err(format!(
                    "[green.service_regions] invalid region '{region}' for service '{service}'; \
                     expected ASCII alphanumeric + '-' or '_', length 1-64"
                ));
            }
        }
        Ok(())
    }

    fn validate_network_energy(value: f64) -> Result<(), String> {
        if !value.is_finite() || value < 0.0 {
            return Err(format!(
                "network_energy_per_byte_kwh must be finite and >= 0.0, got {value}"
            ));
        }
        Ok(())
    }

    /// Validate `[green] hourly_profiles_file`: reject control characters
    /// in the path (log injection) and require that the file actually
    /// loaded when the field is configured.
    fn validate_hourly_profiles_file(&self) -> Result<(), String> {
        let Some(path) = &self.green_hourly_profiles_file else {
            return Ok(());
        };
        if has_control_char(path) {
            return Err("[green] hourly_profiles_file contains control characters".to_string());
        }
        if self.green_custom_hourly_profiles.is_none() {
            return Err(format!(
                "[green] hourly_profiles_file '{path}' was configured but \
                 failed to load. Remove the field to use embedded profiles only."
            ));
        }
        Ok(())
    }

    /// Validate a parsed `[green.electricity_maps]` config section.
    fn validate_electricity_maps(
        cfg: &crate::score::electricity_maps::ElectricityMapsConfig,
    ) -> Result<(), String> {
        if cfg.auth_token.is_empty() {
            return Err(
                "[green.electricity_maps] api_key or PERF_SENTINEL_EMAPS_TOKEN is required"
                    .to_string(),
            );
        }
        if has_control_char(&cfg.auth_token) {
            return Err(
                "[green.electricity_maps] auth token contains control characters".to_string(),
            );
        }
        validate_http_authority(&cfg.api_endpoint, "[green.electricity_maps] endpoint")?;
        // Warn (but do not fail) when a non-empty auth token travels to an
        // http:// endpoint. The Electricity Maps production API is served
        // over https in practice; an http:// endpoint usually means a local
        // test server or a misconfiguration. Flag it so users do not
        // silently ship credentials in cleartext.
        if cfg.api_endpoint.starts_with("http://") && !cfg.auth_token.is_empty() {
            tracing::warn!(
                "[green.electricity_maps] auth token will be sent over http:// \
                 (no TLS). Use https:// for production or set the endpoint to \
                 a loopback/private address if this is intentional."
            );
        }
        let secs = cfg.poll_interval.as_secs();
        check_range(
            "[green.electricity_maps] poll_interval_secs",
            &secs,
            &60,
            &86400,
        )?;
        if cfg.region_map.is_empty() {
            return Err(
                "[green.electricity_maps] region_map must contain at least one entry".to_string(),
            );
        }
        for (region, zone) in &cfg.region_map {
            if zone.is_empty() {
                return Err(format!(
                    "[green.electricity_maps.region_map] zone for '{region}' is empty"
                ));
            }
            if has_control_char(zone)
                || zone.contains('&')
                || zone.contains('#')
                || zone.contains('=')
                || zone.contains('?')
                || zone.contains('%')
                || zone.contains(' ')
                || zone.contains('+')
            {
                return Err(format!(
                    "[green.electricity_maps.region_map] zone '{zone}' for '{region}' \
                     contains invalid characters"
                ));
            }
            if has_control_char(region) {
                return Err(format!(
                    "[green.electricity_maps.region_map] region key '{region}' \
                     contains control characters"
                ));
            }
        }
        Ok(())
    }

    /// Validate a parsed `[green.scaphandre]` config section.
    ///
    /// Rejects: empty endpoint, non-`http://` scheme, credentials in
    /// authority, control characters, invalid port, `scrape_interval_secs`
    /// outside [1, 3600], and `process_map` keys/values that are empty,
    /// >256 chars, or contain control characters.
    fn validate_scaphandre(cfg: &ScaphandreConfig) -> Result<(), String> {
        if cfg.endpoint.is_empty() {
            return Err(
                "[green.scaphandre] endpoint is required when the section is present".to_string(),
            );
        }
        if !cfg.endpoint.starts_with("http://") && !cfg.endpoint.starts_with("https://") {
            return Err(format!(
                "[green.scaphandre] endpoint '{}' must start with 'http://' or 'https://'",
                cfg.endpoint
            ));
        }
        validate_http_authority(&cfg.endpoint, "[green.scaphandre] endpoint")?;
        let secs = cfg.scrape_interval.as_secs();
        if !(1..=3600).contains(&secs) {
            return Err(format!(
                "[green.scaphandre] scrape_interval_secs must be in [1, 3600], got {secs}"
            ));
        }
        // process_map keys are perf-sentinel service names and values are
        // Scaphandre `exe` labels. Validate both are non-empty and of
        // reasonable length; don't run them through is_valid_region_id
        // because service names may contain dots, slashes, etc.
        for (service, exe) in &cfg.process_map {
            if service.is_empty() || service.len() > 256 {
                return Err(format!(
                    "[green.scaphandre] process_map service name '{service}' must be 1-256 chars"
                ));
            }
            if has_control_char(service) {
                return Err(format!(
                    "[green.scaphandre] process_map service name '{service}' \
                     contains control characters"
                ));
            }
            if exe.is_empty() || exe.len() > 256 {
                return Err(format!(
                    "[green.scaphandre] process_map exe for service '{service}' \
                     must be 1-256 chars, got '{exe}'"
                ));
            }
            if has_control_char(exe) {
                return Err(format!(
                    "[green.scaphandre] process_map exe for service '{service}' \
                     contains control characters"
                ));
            }
        }
        Ok(())
    }

    /// Validate a parsed `[green.cloud]` config section.
    fn validate_cloud_energy(cfg: &CloudEnergyConfig) -> Result<(), String> {
        Self::validate_cloud_endpoint(cfg)?;
        Self::validate_cloud_services(cfg)
    }

    /// Validate `[green.cloud]` endpoint, scrape interval, provider, and instance type.
    fn validate_cloud_endpoint(cfg: &CloudEnergyConfig) -> Result<(), String> {
        if cfg.prometheus_endpoint.is_empty() {
            return Err(
                "[green.cloud] prometheus_endpoint is required when the section is present"
                    .to_string(),
            );
        }
        if !cfg.prometheus_endpoint.starts_with("http://")
            && !cfg.prometheus_endpoint.starts_with("https://")
        {
            return Err(format!(
                "[green.cloud] prometheus_endpoint '{}' must start with 'http://' or 'https://'",
                cfg.prometheus_endpoint
            ));
        }
        validate_http_authority(
            &cfg.prometheus_endpoint,
            "[green.cloud] prometheus_endpoint",
        )?;
        let secs = cfg.scrape_interval.as_secs();
        if !(1..=3600).contains(&secs) {
            return Err(format!(
                "[green.cloud] scrape_interval_secs must be in [1, 3600], got {secs}"
            ));
        }
        if let Some(ref p) = cfg.default_provider
            && !matches!(p.as_str(), "aws" | "gcp" | "azure")
        {
            return Err(format!(
                "[green.cloud] default_provider must be 'aws', 'gcp', or 'azure', got '{p}'"
            ));
        }
        if let Some(ref it) = cfg.default_instance_type
            && !crate::score::cloud_energy::table::is_known_instance_type(it)
        {
            tracing::warn!(
                instance_type = %it,
                "[green.cloud] default_instance_type is not in the embedded \
                 SPECpower table; the provider default watts will be used"
            );
        }
        if let Some(ref m) = cfg.cpu_metric
            && has_control_char(m)
        {
            return Err("[green.cloud] cpu_metric contains control characters".to_string());
        }
        Ok(())
    }

    /// Validate per-service entries in `[green.cloud.services]`: cardinality
    /// cap, name/control-char checks, watts ranges, instance type lookup.
    fn validate_cloud_services(cfg: &CloudEnergyConfig) -> Result<(), String> {
        const MAX_CLOUD_SERVICES: usize = 256;
        if cfg.services.len() > MAX_CLOUD_SERVICES {
            return Err(format!(
                "[green.cloud.services] has {} entries; maximum is {MAX_CLOUD_SERVICES}",
                cfg.services.len()
            ));
        }
        for (service, svc_cfg) in &cfg.services {
            Self::validate_cloud_service_name(service)?;
            Self::validate_cloud_service_cpu_query(service, svc_cfg)?;
            match svc_cfg {
                ServiceCloudConfig::ManualWatts {
                    idle_watts,
                    max_watts,
                    ..
                } => Self::validate_manual_watts(service, *idle_watts, *max_watts)?,
                ServiceCloudConfig::InstanceType {
                    provider,
                    instance_type,
                    ..
                } => Self::validate_instance_type_variant(
                    service,
                    provider.as_deref(),
                    instance_type,
                )?,
            }
        }
        Ok(())
    }

    /// Shape + control-char check on a cloud service name.
    fn validate_cloud_service_name(service: &str) -> Result<(), String> {
        if service.is_empty() || service.len() > 256 {
            return Err(format!(
                "[green.cloud.services] service name '{service}' must be 1-256 chars"
            ));
        }
        if has_control_char(service) {
            return Err(format!(
                "[green.cloud.services] service name '{service}' contains control characters"
            ));
        }
        Ok(())
    }

    /// Reject control characters in a service's optional per-service
    /// `cpu_query` override (log-injection / Prometheus-label-injection
    /// guard).
    fn validate_cloud_service_cpu_query(
        service: &str,
        svc_cfg: &ServiceCloudConfig,
    ) -> Result<(), String> {
        let Some(q) = svc_cfg.cpu_query() else {
            return Ok(());
        };
        if has_control_char(q) {
            return Err(format!(
                "[green.cloud.services.{service}] cpu_query contains control characters"
            ));
        }
        Ok(())
    }

    /// Validate a [`ServiceCloudConfig::ManualWatts`] arm: both values
    /// finite and non-negative, and `max_watts >= idle_watts`.
    fn validate_manual_watts(service: &str, idle_watts: f64, max_watts: f64) -> Result<(), String> {
        if !idle_watts.is_finite() || idle_watts < 0.0 {
            return Err(format!(
                "[green.cloud.services.{service}] idle_watts must be finite and >= 0, \
                 got {idle_watts}"
            ));
        }
        if !max_watts.is_finite() || max_watts < 0.0 {
            return Err(format!(
                "[green.cloud.services.{service}] max_watts must be finite and >= 0, \
                 got {max_watts}"
            ));
        }
        if max_watts < idle_watts {
            return Err(format!(
                "[green.cloud.services.{service}] max_watts ({max_watts}) must be \
                 >= idle_watts ({idle_watts})"
            ));
        }
        Ok(())
    }

    /// Validate a [`ServiceCloudConfig::InstanceType`] arm: provider
    /// allow-list, control-char rejection on `instance_type`, and a
    /// soft warning when the type is not in the embedded `SPECpower`
    /// table (not an error, the provider default is used instead).
    fn validate_instance_type_variant(
        service: &str,
        provider: Option<&str>,
        instance_type: &str,
    ) -> Result<(), String> {
        if let Some(p) = provider
            && !matches!(p, "aws" | "gcp" | "azure")
        {
            return Err(format!(
                "[green.cloud.services.{service}] provider must be 'aws', 'gcp', \
                 or 'azure', got '{p}'"
            ));
        }
        if has_control_char(instance_type) {
            return Err(format!(
                "[green.cloud.services.{service}] instance_type contains control characters"
            ));
        }
        if !instance_type.is_empty()
            && !crate::score::cloud_energy::table::is_known_instance_type(instance_type)
        {
            tracing::warn!(
                service = %service,
                instance_type = %instance_type,
                "[green.cloud.services] instance_type is not in the embedded \
                 SPECpower table; provider default watts will be used"
            );
        }
        Ok(())
    }

    fn validate_daemon_limits(&self) -> Result<(), String> {
        check_range(
            "max_payload_size",
            &self.max_payload_size,
            &1024,
            &(100 * 1024 * 1024),
        )?;
        check_range("max_active_traces", &self.max_active_traces, &1, &1_000_000)?;
        check_range(
            "max_events_per_trace",
            &self.max_events_per_trace,
            &1,
            &100_000,
        )?;
        check_range("trace_ttl_ms", &self.trace_ttl_ms, &100, &3_600_000)?;
        check_range("listen_port_http", &self.listen_port, &1, &65535)?;
        check_range("listen_port_grpc", &self.listen_port_grpc, &1, &65535)?;
        Ok(())
    }

    fn validate_detection_params(&self) -> Result<(), String> {
        check_min("n_plus_one_threshold", &self.n_plus_one_threshold, &1)?;
        check_min("window_duration_ms", &self.window_duration_ms, &1)?;
        check_min("slow_query_threshold_ms", &self.slow_query_threshold_ms, &1)?;
        check_min(
            "slow_query_min_occurrences",
            &self.slow_query_min_occurrences,
            &1,
        )?;
        check_range("max_fanout", &self.max_fanout, &1, &100_000)?;
        check_min(
            "chatty_service_min_calls",
            &self.chatty_service_min_calls,
            &1,
        )?;
        check_min(
            "pool_saturation_concurrent_threshold",
            &self.pool_saturation_concurrent_threshold,
            &2,
        )?;
        check_min(
            "serialized_min_sequential",
            &self.serialized_min_sequential,
            &2,
        )?;
        Ok(())
    }

    fn validate_rates(&self) -> Result<(), String> {
        if !(0.0..=1.0).contains(&self.sampling_rate) {
            return Err(format!(
                "sampling_rate must be in [0.0, 1.0], got {}",
                self.sampling_rate
            ));
        }
        if !(0.0..=1.0).contains(&self.io_waste_ratio_max) {
            return Err(format!(
                "io_waste_ratio_max must be in [0.0, 1.0], got {}",
                self.io_waste_ratio_max
            ));
        }
        Ok(())
    }

    /// Warn (but do not reject) non-loopback listen addresses.
    ///
    /// The default is `127.0.0.1` (loopback). Advanced users may override
    /// to `0.0.0.0` for container deployments behind a reverse proxy.
    /// We warn loudly rather than rejecting, because the user's intent is
    /// explicit (they changed the config) and a hard reject would force
    /// workarounds (e.g., iptables) that are harder to audit.
    #[allow(clippy::unnecessary_wraps)]
    fn validate_listen_addr(&self) -> Result<(), String> {
        if self.listen_addr != "127.0.0.1" && self.listen_addr != "::1" {
            tracing::warn!(
                "Daemon configured to listen on non-loopback address: {}. \
                 Endpoints have no authentication, use a reverse proxy or \
                 network policy for security.",
                self.listen_addr
            );
        }
        Ok(())
    }
}

/// Load configuration from a TOML string.
///
/// Supports both the sectioned format and the legacy flat format.
/// Validates that all values are within acceptable bounds after parsing.
///
/// # Errors
///
/// Returns `ConfigError::Parse` if the TOML is malformed, or
/// `ConfigError::Validation` if a field value is out of bounds.
pub fn load_from_str(content: &str) -> Result<Config, ConfigError> {
    let raw: RawConfig = toml::from_str(content).map_err(ConfigError::Parse)?;
    // validate the daemon environment string BEFORE the lossy
    // `Config::from` conversion collapses unknown values into the default.
    // This way "envrionment = \"prod\"" (typo) is rejected with a clear
    // error instead of silently downgrading to Staging.
    if let Some(env_str) = raw.daemon.environment.as_deref()
        && parse_daemon_environment(env_str).is_none()
    {
        return Err(ConfigError::Validation(format!(
            "[daemon] environment '{env_str}' is invalid; \
             expected 'staging' or 'production' (case-insensitive)"
        )));
    }
    let config = Config::from(raw);
    config.validate().map_err(ConfigError::Validation)?;
    Ok(config)
}

/// Errors that can occur during configuration loading.
///
/// `#[non_exhaustive]` so that adding future variants (e.g. a new
/// validation failure when a new config section lands) stays a
/// SemVer-minor change.
#[derive(Debug, thiserror::Error)]
#[non_exhaustive]
pub enum ConfigError {
    /// TOML parsing error.
    #[error("config parse error: {0}")]
    Parse(#[from] toml::de::Error),
    /// Validation error (out-of-range values).
    #[error("config validation error: {0}")]
    Validation(String),
}

#[cfg(test)]
mod tests {
    use super::*;

    #[test]
    fn default_config_has_safe_defaults() {
        let config = Config::default();
        assert_eq!(config.max_payload_size, 1_048_576);
        assert_eq!(config.listen_addr, "127.0.0.1");
        assert_eq!(config.n_plus_one_threshold, 5);
        assert_eq!(config.window_duration_ms, 500);
        assert_eq!(config.trace_ttl_ms, 30_000);
        assert_eq!(config.max_active_traces, 10_000);
        assert_eq!(config.max_events_per_trace, 1_000);
    }

    #[test]
    fn parse_empty_toml_gives_defaults() {
        let config = load_from_str("").unwrap();
        assert_eq!(config.max_payload_size, 1_048_576);
    }

    #[test]
    fn parse_partial_toml() {
        let config = load_from_str("n_plus_one_threshold = 10").unwrap();
        assert_eq!(config.n_plus_one_threshold, 10);
        assert_eq!(config.max_payload_size, 1_048_576); // default preserved
    }

    #[test]
    fn parse_window_config() {
        let config = load_from_str(
            "window_duration_ms = 1000\ntrace_ttl_ms = 60000\nmax_active_traces = 5000",
        )
        .unwrap();
        assert_eq!(config.window_duration_ms, 1000);
        assert_eq!(config.trace_ttl_ms, 60_000);
        assert_eq!(config.max_active_traces, 5000);
    }

    #[test]
    fn parse_sectioned_format() {
        let toml = r#"
[thresholds]
n_plus_one_sql_critical_max = 2
n_plus_one_http_warning_max = 5
io_waste_ratio_max = 0.50

[detection]
window_duration_ms = 1000
n_plus_one_min_occurrences = 10

[green]
enabled = false

[daemon]
listen_address = "0.0.0.0"
listen_port_http = 9418
listen_port_grpc = 9417
json_socket = "/var/run/perf-sentinel.sock"
max_active_traces = 20000
trace_ttl_ms = 60000
sampling_rate = 0.5
max_events_per_trace = 500
max_payload_size = 2097152
"#;
        let config = load_from_str(toml).unwrap();
        assert_eq!(config.n_plus_one_sql_critical_max, 2);
        assert_eq!(config.n_plus_one_http_warning_max, 5);
        assert!((config.io_waste_ratio_max - 0.50).abs() < f64::EPSILON);
        assert_eq!(config.n_plus_one_threshold, 10);
        assert_eq!(config.window_duration_ms, 1000);
        assert!(!config.green_enabled);
        assert_eq!(config.listen_addr, "0.0.0.0");
        assert_eq!(config.listen_port, 9418);
        assert_eq!(config.listen_port_grpc, 9417);
        assert_eq!(config.json_socket, "/var/run/perf-sentinel.sock");
        assert_eq!(config.max_active_traces, 20_000);
        assert_eq!(config.trace_ttl_ms, 60_000);
        assert!((config.sampling_rate - 0.5).abs() < f64::EPSILON);
        assert_eq!(config.max_events_per_trace, 500);
        assert_eq!(config.max_payload_size, 2_097_152);
    }

    #[test]
    fn section_overrides_flat_field() {
        let toml = r"
n_plus_one_threshold = 7
window_duration_ms = 800

[detection]
n_plus_one_min_occurrences = 12
";
        let config = load_from_str(toml).unwrap();
        // Section takes priority over flat field
        assert_eq!(config.n_plus_one_threshold, 12);
        // Flat field used when section does not override
        assert_eq!(config.window_duration_ms, 800);
    }

    #[test]
    fn new_fields_have_correct_defaults() {
        let config = Config::default();
        assert_eq!(config.n_plus_one_sql_critical_max, 0);
        assert_eq!(config.n_plus_one_http_warning_max, 3);
        assert!((config.io_waste_ratio_max - 0.30).abs() < f64::EPSILON);
        assert!(config.green_enabled);
        assert_eq!(config.listen_port_grpc, 4317);
        assert_eq!(config.json_socket, "/tmp/perf-sentinel.sock");
        assert!((config.sampling_rate - 1.0).abs() < f64::EPSILON);
    }

    #[test]
    fn default_config_validates() {
        let config = Config::default();
        assert!(config.validate().is_ok());
    }

    #[test]
    fn rejects_sampling_rate_above_one() {
        let result = load_from_str("[daemon]\nsampling_rate = 5.0");
        assert!(result.is_err());
        let err = result.unwrap_err().to_string();
        assert!(err.contains("sampling_rate"), "got: {err}");
    }

    #[test]
    fn rejects_negative_sampling_rate() {
        let result = load_from_str("[daemon]\nsampling_rate = -0.1");
        assert!(result.is_err());
    }

    #[test]
    fn rejects_io_waste_ratio_max_above_one() {
        let result = load_from_str("[thresholds]\nio_waste_ratio_max = 1.5");
        assert!(result.is_err());
        let err = result.unwrap_err().to_string();
        assert!(err.contains("io_waste_ratio_max"), "got: {err}");
    }

    #[test]
    fn rejects_zero_max_payload_size() {
        let result = load_from_str("[daemon]\nmax_payload_size = 0");
        assert!(result.is_err());
    }

    #[test]
    fn rejects_zero_n_plus_one_threshold() {
        let result = load_from_str("n_plus_one_threshold = 0");
        assert!(result.is_err());
    }

    #[test]
    fn rejects_zero_max_active_traces() {
        let result = load_from_str("max_active_traces = 0");
        assert!(result.is_err());
    }

    #[test]
    fn rejects_zero_max_events_per_trace() {
        let result = load_from_str("max_events_per_trace = 0");
        assert!(result.is_err());
    }

    #[test]
    fn slow_query_defaults() {
        let config = Config::default();
        assert_eq!(config.slow_query_threshold_ms, 500);
        assert_eq!(config.slow_query_min_occurrences, 3);
        assert!(config.green_default_region.is_none());
        assert!(config.green_service_regions.is_empty());
        assert!(
            (config.green_embodied_carbon_per_request_gco2
                - DEFAULT_EMBODIED_CARBON_PER_REQUEST_GCO2)
                .abs()
                < f64::EPSILON
        );
    }

    #[test]
    fn parse_slow_query_config() {
        let toml = r"
[detection]
slow_query_threshold_ms = 1000
slow_query_min_occurrences = 5
";
        let config = load_from_str(toml).unwrap();
        assert_eq!(config.slow_query_threshold_ms, 1000);
        assert_eq!(config.slow_query_min_occurrences, 5);
    }

    #[test]
    fn parse_green_default_region() {
        let toml = r#"
[green]
enabled = true
default_region = "eu-west-3"
"#;
        let config = load_from_str(toml).unwrap();
        assert_eq!(config.green_default_region.as_deref(), Some("eu-west-3"));
    }

    #[test]
    fn parse_green_service_regions() {
        let toml = r#"
[green]
enabled = true
default_region = "eu-west-3"

[green.service_regions]
"order-svc" = "us-east-1"
"chat-svc" = "ap-southeast-1"
"#;
        let config = load_from_str(toml).unwrap();
        assert_eq!(config.green_service_regions.len(), 2);
        assert_eq!(
            config
                .green_service_regions
                .get("order-svc")
                .map(String::as_str),
            Some("us-east-1")
        );
        assert_eq!(
            config
                .green_service_regions
                .get("chat-svc")
                .map(String::as_str),
            Some("ap-southeast-1")
        );
    }

    #[test]
    fn parse_green_embodied_carbon_override() {
        let toml = r"
[green]
enabled = true
embodied_carbon_per_request_gco2 = 0.005
";
        let config = load_from_str(toml).unwrap();
        assert!((config.green_embodied_carbon_per_request_gco2 - 0.005).abs() < f64::EPSILON);
    }

    #[test]
    fn rejects_negative_embodied_carbon() {
        let result = load_from_str("[green]\nembodied_carbon_per_request_gco2 = -0.001");
        assert!(result.is_err());
        let err = result.unwrap_err().to_string();
        assert!(
            err.contains("embodied_carbon_per_request_gco2"),
            "got: {err}"
        );
    }

    #[test]
    fn accepts_zero_embodied_carbon() {
        let toml = r"
[green]
embodied_carbon_per_request_gco2 = 0.0
";
        let config = load_from_str(toml).unwrap();
        assert!((config.green_embodied_carbon_per_request_gco2 - 0.0).abs() < f64::EPSILON);
    }

    #[test]
    fn empty_service_regions_default() {
        let toml = r#"
[green]
default_region = "eu-west-3"
"#;
        let config = load_from_str(toml).unwrap();
        assert!(config.green_service_regions.is_empty());
    }

    // ----- Region validation + lowercase + both-set -----

    #[test]
    fn rejects_invalid_default_region_characters() {
        // Space in region name — log-injection protection at config load.
        let result = load_from_str("[green]\ndefault_region = \"eu west 3\"");
        assert!(result.is_err());
        let err = result.unwrap_err().to_string();
        assert!(
            err.contains("default_region"),
            "error should mention default_region, got: {err}"
        );
    }

    #[test]
    fn rejects_oversized_default_region() {
        // 65 chars — just over the 64-char cap.
        let long_region = "a".repeat(65);
        let toml = format!("[green]\ndefault_region = \"{long_region}\"");
        let result = load_from_str(&toml);
        assert!(result.is_err());
    }

    #[test]
    fn rejects_default_region_with_newline_escape() {
        // In a TOML basic string, `\n` is an escape sequence for a real
        // newline byte. The validator must reject the resulting control
        // char to block log-forging via default_region.
        let result = load_from_str("[green]\ndefault_region = \"eu-west-3\\n\"");
        assert!(result.is_err());
        assert!(
            result.unwrap_err().to_string().contains("default_region"),
            "error should mention default_region"
        );
    }

    #[test]
    fn rejects_default_region_with_literal_newline() {
        // Multi-line basic string with an actual newline byte in the
        // value. Also rejected at load time.
        let result = load_from_str("[green]\ndefault_region = \"\"\"eu-west-3\n\"\"\"");
        assert!(result.is_err());
    }

    #[test]
    fn accepts_known_regions() {
        // Sanity: all known region names pass the validator.
        for region in ["eu-west-3", "us-east-1", "fr", "mars-1", "unknown"] {
            let toml = format!("[green]\ndefault_region = \"{region}\"");
            let config = load_from_str(&toml)
                .unwrap_or_else(|e| panic!("region '{region}' should be accepted, got error: {e}"));
            assert_eq!(config.green_default_region.as_deref(), Some(region));
        }
    }

    #[test]
    fn rejects_invalid_service_regions_service_name() {
        let toml = r#"
[green.service_regions]
"bad service" = "us-east-1"
"#;
        let result = load_from_str(toml);
        assert!(result.is_err());
        let err = result.unwrap_err().to_string();
        assert!(
            err.contains("service_regions"),
            "error should mention service_regions, got: {err}"
        );
    }

    #[test]
    fn rejects_invalid_service_regions_region_value() {
        let toml = r#"
[green.service_regions]
"order-svc" = "us east 1"
"#;
        let result = load_from_str(toml);
        assert!(result.is_err());
    }

    #[test]
    fn rejects_oversized_service_regions_map() {
        // Fat-finger or malicious config with too many entries gets
        // rejected at load time with a clear error mentioning the cap.
        use std::fmt::Write as _;
        let mut toml = String::from("[green.service_regions]\n");
        for i in 0..1025 {
            let _ = writeln!(toml, "\"svc-{i:04}\" = \"eu-west-3\"");
        }
        let result = load_from_str(&toml);
        assert!(result.is_err());
        let err = result.unwrap_err().to_string();
        assert!(
            err.contains("service_regions") && err.contains("1025"),
            "error should mention service_regions and the count, got: {err}"
        );
    }

    #[test]
    fn accepts_service_regions_at_exactly_the_cap() {
        // Boundary check: exactly 1024 entries should pass.
        use std::fmt::Write as _;
        let mut toml = String::from("[green.service_regions]\n");
        for i in 0..1024 {
            let _ = writeln!(toml, "\"svc-{i:04}\" = \"eu-west-3\"");
        }
        let config = load_from_str(&toml).expect("1024 entries should be accepted");
        assert_eq!(config.green_service_regions.len(), 1024);
    }

    #[test]
    fn service_regions_keys_are_lowercased_on_load() {
        // Config loader lowercases keys so resolve_region's
        // case-insensitive lookup works transparently.
        let toml = r#"
[green.service_regions]
"Order-Svc" = "us-east-1"
"CHAT-SVC" = "ap-southeast-1"
"#;
        let config = load_from_str(toml).unwrap();
        assert_eq!(config.green_service_regions.len(), 2);
        // Keys are lowercased regardless of TOML casing.
        assert_eq!(
            config
                .green_service_regions
                .get("order-svc")
                .map(String::as_str),
            Some("us-east-1")
        );
        assert_eq!(
            config
                .green_service_regions
                .get("chat-svc")
                .map(String::as_str),
            Some("ap-southeast-1")
        );
        // The original casings should NOT be present.
        assert!(!config.green_service_regions.contains_key("Order-Svc"));
    }

    #[test]
    fn rejects_zero_slow_query_threshold() {
        let result = load_from_str("[detection]\nslow_query_threshold_ms = 0");
        assert!(result.is_err());
    }

    #[test]
    fn rejects_zero_slow_query_min_occurrences() {
        let result = load_from_str("[detection]\nslow_query_min_occurrences = 0");
        assert!(result.is_err());
    }

    #[test]
    fn rejects_zero_max_fanout() {
        let result = load_from_str("[detection]\nmax_fanout = 0");
        assert!(result.is_err());
    }

    #[test]
    fn rejects_max_fanout_over_100k() {
        let result = load_from_str("[detection]\nmax_fanout = 100001");
        assert!(result.is_err());
        let err = result.unwrap_err().to_string();
        assert!(err.contains("max_fanout"), "got: {err}");
    }

    #[test]
    fn accepts_max_fanout_at_100k() {
        let result = load_from_str("[detection]\nmax_fanout = 100000");
        assert!(result.is_ok());
    }

    #[test]
    fn rejects_max_payload_size_over_100mb() {
        let result = load_from_str("[daemon]\nmax_payload_size = 104857601");
        assert!(result.is_err());
        let err = result.unwrap_err().to_string();
        assert!(err.contains("max_payload_size"), "got: {err}");
    }

    #[test]
    fn accepts_max_payload_size_at_100mb() {
        let result = load_from_str("[daemon]\nmax_payload_size = 104857600");
        assert!(result.is_ok());
    }

    #[test]
    fn rejects_max_active_traces_over_1m() {
        let result = load_from_str("[daemon]\nmax_active_traces = 1000001");
        assert!(result.is_err());
        let err = result.unwrap_err().to_string();
        assert!(err.contains("max_active_traces"), "got: {err}");
    }

    #[test]
    fn accepts_max_active_traces_at_1m() {
        let result = load_from_str("[daemon]\nmax_active_traces = 1000000");
        assert!(result.is_ok());
    }

    #[test]
    fn rejects_max_events_per_trace_over_100k() {
        let result = load_from_str("[daemon]\nmax_events_per_trace = 100001");
        assert!(result.is_err());
        let err = result.unwrap_err().to_string();
        assert!(err.contains("max_events_per_trace"), "got: {err}");
    }

    #[test]
    fn accepts_max_events_per_trace_at_100k() {
        let result = load_from_str("[daemon]\nmax_events_per_trace = 100000");
        assert!(result.is_ok());
    }

    #[test]
    fn rejects_trace_ttl_below_100() {
        let result = load_from_str("[daemon]\ntrace_ttl_ms = 50");
        assert!(result.is_err());
    }

    #[test]
    fn rejects_zero_window_duration() {
        let result = load_from_str("[detection]\nwindow_duration_ms = 0");
        assert!(result.is_err());
    }

    #[test]
    fn green_disabled_parses() {
        let config = load_from_str("[green]\nenabled = false").unwrap();
        assert!(!config.green_enabled);
    }

    // -- Port validation --

    #[test]
    fn rejects_port_zero() {
        let result = load_from_str("[daemon]\nlisten_port_http = 0");
        assert!(result.is_err());
    }

    #[test]
    fn accepts_port_one() {
        let config = load_from_str("[daemon]\nlisten_port_http = 1").unwrap();
        assert_eq!(config.listen_port, 1);
    }

    #[test]
    fn accepts_port_65535() {
        let config = load_from_str("[daemon]\nlisten_port_http = 65535").unwrap();
        assert_eq!(config.listen_port, 65535);
    }

    #[test]
    fn rejects_grpc_port_zero() {
        let result = load_from_str("[daemon]\nlisten_port_grpc = 0");
        assert!(result.is_err());
    }

    // -- trace_ttl_ms upper bound --

    #[test]
    fn rejects_trace_ttl_above_1h() {
        let result = load_from_str("[daemon]\ntrace_ttl_ms = 3600001");
        assert!(result.is_err());
    }

    #[test]
    fn accepts_trace_ttl_at_1h() {
        let config = load_from_str("[daemon]\ntrace_ttl_ms = 3600000").unwrap();
        assert_eq!(config.trace_ttl_ms, 3_600_000);
    }

    #[test]
    fn accepts_trace_ttl_at_100ms() {
        let config = load_from_str("[daemon]\ntrace_ttl_ms = 100").unwrap();
        assert_eq!(config.trace_ttl_ms, 100);
    }

    // -- Sampling rate edge cases --

    #[test]
    fn accepts_sampling_rate_zero() {
        let config = load_from_str("[daemon]\nsampling_rate = 0.0").unwrap();
        assert!((config.sampling_rate - 0.0).abs() < f64::EPSILON);
    }

    #[test]
    fn accepts_sampling_rate_one() {
        let config = load_from_str("[daemon]\nsampling_rate = 1.0").unwrap();
        assert!((config.sampling_rate - 1.0).abs() < f64::EPSILON);
    }

    // --- [daemon] environment parsing ---

    #[test]
    fn daemon_environment_defaults_to_staging() {
        let config = Config::default();
        assert_eq!(config.daemon_environment, DaemonEnvironment::Staging);
        assert_eq!(config.confidence(), Confidence::DaemonStaging);
    }

    #[test]
    fn daemon_environment_omitted_uses_default() {
        let config = load_from_str("[daemon]\nmax_active_traces = 100").unwrap();
        assert_eq!(config.daemon_environment, DaemonEnvironment::Staging);
    }

    #[test]
    fn daemon_environment_staging() {
        let config = load_from_str("[daemon]\nenvironment = \"staging\"").unwrap();
        assert_eq!(config.daemon_environment, DaemonEnvironment::Staging);
        assert_eq!(config.confidence(), Confidence::DaemonStaging);
    }

    #[test]
    fn daemon_environment_production() {
        let config = load_from_str("[daemon]\nenvironment = \"production\"").unwrap();
        assert_eq!(config.daemon_environment, DaemonEnvironment::Production);
        assert_eq!(config.confidence(), Confidence::DaemonProduction);
    }

    #[test]
    fn daemon_environment_case_insensitive() {
        let config = load_from_str("[daemon]\nenvironment = \"PRODUCTION\"").unwrap();
        assert_eq!(config.daemon_environment, DaemonEnvironment::Production);
        let config = load_from_str("[daemon]\nenvironment = \"Staging\"").unwrap();
        assert_eq!(config.daemon_environment, DaemonEnvironment::Staging);
    }

    #[test]
    fn daemon_environment_rejects_unknown() {
        let result = load_from_str("[daemon]\nenvironment = \"prod\"");
        assert!(result.is_err());
        let err = result.unwrap_err().to_string();
        assert!(err.contains("environment"), "got: {err}");
        assert!(err.contains("staging"), "error should mention valid values");
        assert!(
            err.contains("production"),
            "error should mention valid values"
        );
    }

    #[test]
    fn daemon_environment_rejects_empty() {
        let result = load_from_str("[daemon]\nenvironment = \"\"");
        assert!(result.is_err());
    }

    #[test]
    fn daemon_environment_rejects_dev() {
        let result = load_from_str("[daemon]\nenvironment = \"dev\"");
        assert!(result.is_err());
    }

    #[test]
    fn daemon_environment_as_str() {
        assert_eq!(DaemonEnvironment::Staging.as_str(), "staging");
        assert_eq!(DaemonEnvironment::Production.as_str(), "production");
    }

    // --- [green] use_hourly_profiles ---

    #[test]
    fn green_use_hourly_profiles_defaults_to_true() {
        let config = Config::default();
        assert!(config.green_use_hourly_profiles);
    }

    #[test]
    fn green_use_hourly_profiles_omitted_uses_default() {
        let config = load_from_str("[green]\nenabled = true\n").unwrap();
        assert!(config.green_use_hourly_profiles);
    }

    #[test]
    fn green_use_hourly_profiles_explicit_false() {
        let config = load_from_str("[green]\nuse_hourly_profiles = false\n").unwrap();
        assert!(!config.green_use_hourly_profiles);
    }

    #[test]
    fn green_use_hourly_profiles_explicit_true() {
        let config = load_from_str("[green]\nuse_hourly_profiles = true\n").unwrap();
        assert!(config.green_use_hourly_profiles);
    }

    // --- [green] hourly_profiles_file ---

    #[test]
    fn hourly_profiles_file_absent_by_default() {
        let config = Config::default();
        assert!(config.green_hourly_profiles_file.is_none());
        assert!(config.green_custom_hourly_profiles.is_none());
    }

    #[test]
    fn hourly_profiles_file_control_chars_rejected() {
        let config = load_from_str("[green]\nhourly_profiles_file = \"/tmp/profiles\\n.json\"\n");
        // The control character check happens during loading (sets None)
        // and then validate_green rejects the config.
        if let Ok(c) = config {
            let err = c.validate().unwrap_err();
            assert!(
                err.contains("control characters") || err.contains("failed to load"),
                "expected control char or load failure error, got: {err}"
            );
        } else {
            // TOML parse error is also acceptable
        }
    }

    #[test]
    fn hourly_profiles_file_nonexistent_path_rejected() {
        let result =
            load_from_str("[green]\nhourly_profiles_file = \"/nonexistent/profiles.json\"\n");
        let err = result.unwrap_err();
        assert!(
            format!("{err}").contains("failed to load"),
            "expected load failure error, got: {err}"
        );
    }

    // --- [green.scaphandre] parsing ---

    #[test]
    fn scaphandre_absent_by_default() {
        let config = Config::default();
        assert!(config.green_scaphandre.is_none());
    }

    #[test]
    fn scaphandre_empty_section_parses_to_none() {
        // An empty [green.scaphandre] table (no endpoint) is treated
        // as "Scaphandre not configured" — the scraper is not spawned.
        let config = load_from_str("[green.scaphandre]\n").unwrap();
        assert!(config.green_scaphandre.is_none());
    }

    #[test]
    fn scaphandre_endpoint_only() {
        let config =
            load_from_str("[green.scaphandre]\nendpoint = \"http://localhost:8080/metrics\"\n")
                .unwrap();
        let cfg = config.green_scaphandre.unwrap();
        assert_eq!(cfg.endpoint, "http://localhost:8080/metrics");
        // Default interval is 5 s.
        assert_eq!(cfg.scrape_interval.as_secs(), 5);
        assert!(cfg.process_map.is_empty());
    }

    #[test]
    fn scaphandre_full_config() {
        let toml = r#"
[green.scaphandre]
endpoint = "http://localhost:9090/metrics"
scrape_interval_secs = 10

[green.scaphandre.process_map]
"order-svc" = "java"
"chat-svc" = "dotnet"
"#;
        let config = load_from_str(toml).unwrap();
        let cfg = config.green_scaphandre.unwrap();
        assert_eq!(cfg.endpoint, "http://localhost:9090/metrics");
        assert_eq!(cfg.scrape_interval.as_secs(), 10);
        assert_eq!(
            cfg.process_map.get("order-svc").map(String::as_str),
            Some("java")
        );
        assert_eq!(
            cfg.process_map.get("chat-svc").map(String::as_str),
            Some("dotnet")
        );
    }

    #[test]
    fn scaphandre_accepts_https_endpoint() {
        let result =
            load_from_str("[green.scaphandre]\nendpoint = \"https://secure:8080/metrics\"\n");
        assert!(result.is_ok(), "HTTPS endpoints should be accepted");
    }

    #[test]
    fn scaphandre_rejects_zero_interval() {
        let result = load_from_str(
            "[green.scaphandre]\nendpoint = \"http://localhost/metrics\"\nscrape_interval_secs = 0\n",
        );
        assert!(result.is_err());
        let err = result.unwrap_err().to_string();
        assert!(err.contains("scrape_interval_secs"), "got: {err}");
    }

    #[test]
    fn scaphandre_rejects_huge_interval() {
        let result = load_from_str(
            "[green.scaphandre]\nendpoint = \"http://localhost/metrics\"\nscrape_interval_secs = 99999\n",
        );
        assert!(result.is_err());
    }

    #[test]
    fn scaphandre_rejects_empty_exe_in_process_map() {
        let toml = r#"
[green.scaphandre]
endpoint = "http://localhost/metrics"

[green.scaphandre.process_map]
"order-svc" = ""
"#;
        let result = load_from_str(toml);
        assert!(result.is_err());
        let err = result.unwrap_err().to_string();
        assert!(err.contains("process_map"), "got: {err}");
    }

    #[test]
    fn scaphandre_accepts_interval_at_boundary_1s() {
        let config = load_from_str(
            "[green.scaphandre]\nendpoint = \"http://localhost/metrics\"\nscrape_interval_secs = 1\n",
        )
        .unwrap();
        assert_eq!(
            config
                .green_scaphandre
                .as_ref()
                .unwrap()
                .scrape_interval
                .as_secs(),
            1
        );
    }

    #[test]
    fn scaphandre_accepts_interval_at_boundary_3600s() {
        let config = load_from_str(
            "[green.scaphandre]\nendpoint = \"http://localhost/metrics\"\nscrape_interval_secs = 3600\n",
        )
        .unwrap();
        assert_eq!(
            config
                .green_scaphandre
                .as_ref()
                .unwrap()
                .scrape_interval
                .as_secs(),
            3600
        );
    }

    // ------------------------------------------------------------------
    // [green.cloud] config tests
    // ------------------------------------------------------------------

    #[test]
    fn cloud_section_absent_yields_none() {
        let toml = "[green]\nenabled = true\n";
        let cfg: Config = toml::from_str::<RawConfig>(toml).unwrap().into();
        assert!(cfg.green_cloud_energy.is_none());
    }

    #[test]
    fn cloud_section_endpoint_only_parses_with_defaults() {
        let toml = r#"
[green.cloud]
prometheus_endpoint = "http://prom:9090"
"#;
        let cfg: Config = toml::from_str::<RawConfig>(toml).unwrap().into();
        let cloud = cfg.green_cloud_energy.unwrap();
        assert_eq!(cloud.prometheus_endpoint, "http://prom:9090");
        assert_eq!(cloud.scrape_interval.as_secs(), 15);
        assert!(cloud.default_provider.is_none());
        assert!(cloud.services.is_empty());
    }

    #[test]
    fn cloud_section_full_config_with_both_service_types() {
        let toml = r#"
[green.cloud]
prometheus_endpoint = "http://prom:9090"
scrape_interval_secs = 30
default_provider = "aws"

[green.cloud.services.svc-a]
provider = "gcp"
instance_type = "n2-standard-8"

[green.cloud.services.svc-b]
idle_watts = 45
max_watts = 120
"#;
        let cfg: Config = toml::from_str::<RawConfig>(toml).unwrap().into();
        assert!(cfg.validate().is_ok());
        let cloud = cfg.green_cloud_energy.as_ref().unwrap();
        assert_eq!(cloud.scrape_interval.as_secs(), 30);
        assert_eq!(cloud.default_provider.as_deref(), Some("aws"));
        assert_eq!(cloud.services.len(), 2);
    }

    #[test]
    fn cloud_accepts_https_endpoint() {
        let toml = r#"
[green.cloud]
prometheus_endpoint = "https://prom:9090"
"#;
        let cfg: Config = toml::from_str::<RawConfig>(toml).unwrap().into();
        assert!(cfg.validate().is_ok(), "HTTPS endpoints should be accepted");
    }

    #[test]
    fn cloud_rejects_credentials_in_endpoint() {
        let toml = r#"
[green.cloud]
prometheus_endpoint = "http://user:pass@prom:9090"
"#;
        let cfg: Config = toml::from_str::<RawConfig>(toml).unwrap().into();
        let err = cfg.validate().unwrap_err();
        assert!(err.contains("credentials"), "error: {err}");
    }

    #[test]
    fn cloud_rejects_invalid_scrape_interval() {
        let toml = r#"
[green.cloud]
prometheus_endpoint = "http://prom:9090"
scrape_interval_secs = 0
"#;
        let cfg: Config = toml::from_str::<RawConfig>(toml).unwrap().into();
        let err = cfg.validate().unwrap_err();
        assert!(err.contains("scrape_interval"), "error: {err}");
    }

    #[test]
    fn cloud_rejects_invalid_provider() {
        let toml = r#"
[green.cloud]
prometheus_endpoint = "http://prom:9090"
default_provider = "alibaba"
"#;
        let cfg: Config = toml::from_str::<RawConfig>(toml).unwrap().into();
        let err = cfg.validate().unwrap_err();
        assert!(err.contains("default_provider"), "error: {err}");
    }

    #[test]
    fn cloud_rejects_max_watts_less_than_idle() {
        let toml = r#"
[green.cloud]
prometheus_endpoint = "http://prom:9090"

[green.cloud.services.bad-svc]
idle_watts = 100
max_watts = 50
"#;
        let cfg: Config = toml::from_str::<RawConfig>(toml).unwrap().into();
        let err = cfg.validate().unwrap_err();
        assert!(err.contains("max_watts"), "error: {err}");
    }

    #[test]
    fn cloud_rejects_service_name_with_control_chars() {
        let toml = "
[green.cloud]
prometheus_endpoint = \"http://prom:9090\"

[green.cloud.services.\"bad\\nsvc\"]
idle_watts = 10
max_watts = 50
";
        let cfg: Config = toml::from_str::<RawConfig>(toml).unwrap().into();
        let err = cfg.validate().unwrap_err();
        assert!(err.contains("control characters"), "error: {err}");
    }

    #[test]
    fn config_per_operation_coefficients_default_true() {
        let cfg = Config::default();
        assert!(cfg.green_per_operation_coefficients);
    }

    #[test]
    fn config_include_network_transport_default_false() {
        let cfg = Config::default();
        assert!(!cfg.green_include_network_transport);
    }

    #[test]
    fn config_network_energy_per_byte_kwh_default() {
        let cfg = Config::default();
        assert!(
            (cfg.green_network_energy_per_byte_kwh
                - crate::score::carbon::DEFAULT_NETWORK_ENERGY_PER_BYTE_KWH)
                .abs()
                < f64::EPSILON
        );
    }

    #[test]
    fn config_network_energy_per_byte_kwh_rejects_negative() {
        let toml = r"
[green]
network_energy_per_byte_kwh = -0.001
";
        let cfg: Config = toml::from_str::<RawConfig>(toml).unwrap().into();
        let err = cfg.validate().unwrap_err();
        assert!(err.contains("network_energy_per_byte_kwh"), "error: {err}");
    }

    #[test]
    fn config_network_energy_per_byte_kwh_rejects_nan() {
        let cfg = Config {
            green_network_energy_per_byte_kwh: f64::NAN,
            ..Config::default()
        };
        let err = cfg.validate().unwrap_err();
        assert!(err.contains("network_energy_per_byte_kwh"), "error: {err}");
    }

    #[test]
    fn config_per_operation_coefficients_from_toml() {
        let toml = r"
[green]
per_operation_coefficients = false
";
        let cfg: Config = toml::from_str::<RawConfig>(toml).unwrap().into();
        assert!(!cfg.green_per_operation_coefficients);
    }

    #[test]
    fn config_include_network_transport_from_toml() {
        let toml = r"
[green]
include_network_transport = true
network_energy_per_byte_kwh = 0.00000000008
";
        let cfg: Config = toml::from_str::<RawConfig>(toml).unwrap().into();
        assert!(cfg.green_include_network_transport);
        assert!((cfg.green_network_energy_per_byte_kwh - 0.000_000_000_08).abs() < f64::EPSILON);
    }

    // --- validate_http_authority error paths ---

    #[test]
    fn validate_http_authority_rejects_empty_host() {
        assert!(validate_http_authority("http://", "test").is_err());
    }

    #[test]
    fn validate_http_authority_rejects_credentials() {
        let err = validate_http_authority("http://user:pass@host/", "test").unwrap_err();
        assert!(err.contains("credentials"));
    }

    #[test]
    fn validate_http_authority_rejects_control_char() {
        // Embed a tab (0x09) in the host.
        let err = validate_http_authority("http://bad\thost/", "test").unwrap_err();
        assert!(err.contains("control"));
    }

    #[test]
    fn validate_http_authority_rejects_invalid_ipv4_port() {
        let err = validate_http_authority("http://host:abc/", "test").unwrap_err();
        assert!(err.contains("port"));
    }

    #[test]
    fn validate_http_authority_accepts_bare_ipv6() {
        // `[::1]` without port — should not error on the port-parse branch.
        assert!(validate_http_authority("http://[::1]/metrics", "test").is_ok());
    }

    #[test]
    fn validate_http_authority_accepts_ipv6_with_port() {
        assert!(validate_http_authority("http://[::1]:8080/metrics", "test").is_ok());
    }

    #[test]
    fn validate_http_authority_rejects_ipv6_with_invalid_port() {
        let err = validate_http_authority("http://[::1]:abc/metrics", "test").unwrap_err();
        assert!(err.contains("port"));
    }

    #[test]
    fn validate_http_authority_accepts_https_scheme() {
        assert!(validate_http_authority("https://host:443/", "test").is_ok());
    }

    // --- validate_green error paths ---

    #[test]
    fn validate_green_rejects_nonfinite_embodied_carbon() {
        let toml = "[green]\nembodied_carbon_per_request_gco2 = nan\n";
        let err = load_from_str(toml).unwrap_err();
        assert!(format!("{err:?}").contains("finite"));
    }

    #[test]
    fn validate_green_rejects_hourly_profiles_file_that_fails_to_load() {
        let toml = r#"
[green]
hourly_profiles_file = "/tmp/does-not-exist-perfsentinel-test.json"
"#;
        let err = load_from_str(toml).unwrap_err();
        let msg = format!("{err:?}");
        assert!(msg.contains("hourly_profiles_file") || msg.contains("failed to load"));
    }

    // --- convert_electricity_maps_section branches ---

    // Local imports used by all the electricity_maps tests below.
    // `HashMap` and `Duration` are already in scope via `use super::*;`
    // at the top of this module, but Qodana flags the fully-qualified
    // forms as unnecessary — using the short names reads cleaner anyway.
    use crate::score::electricity_maps::ElectricityMapsConfig;

    #[test]
    fn electricity_maps_empty_api_key_returns_none() {
        // When the api_key is explicitly an empty string and no env var is
        // set, the conversion returns None (subsystem stays inactive).
        let raw = ElectricityMapsSection {
            api_key: Some(String::new()),
            endpoint: None,
            poll_interval_secs: None,
            region_map: HashMap::new(),
        };
        // Pass a stubbed env-lookup that returns None so the test is
        // independent of the ambient process environment (no `unsafe`
        // env mutation, no races with other tests in the same binary).
        assert!(convert_electricity_maps_section_with_env(&raw, || None).is_none());
    }

    #[test]
    fn electricity_maps_warn_when_api_key_in_config_file() {
        // `api_key` set, env var unset → returns Some(...) but emits a
        // warning about preferring the env var. The warning path is
        // exercised; we just verify the conversion succeeds.
        let mut region_map = HashMap::new();
        region_map.insert("eu-west-3".to_string(), "FR".to_string());
        let raw = ElectricityMapsSection {
            api_key: Some("file-token".to_string()),
            endpoint: None,
            poll_interval_secs: Some(600),
            region_map,
        };
        let cfg = convert_electricity_maps_section_with_env(&raw, || None).expect("should convert");
        assert_eq!(cfg.auth_token, "file-token");
        assert_eq!(cfg.poll_interval, Duration::from_secs(600));
        // default endpoint fallback
        assert_eq!(cfg.api_endpoint, "https://api.electricitymaps.com/v3");
        // region key was lowercased (it was already lowercase, so idempotent)
        assert!(cfg.region_map.contains_key("eu-west-3"));
    }

    #[test]
    fn electricity_maps_region_map_keys_lowercased() {
        let mut region_map = HashMap::new();
        region_map.insert("EU-WEST-3".to_string(), "FR".to_string());
        region_map.insert("Us-East-1".to_string(), "US-MIDA-PJM".to_string());
        let raw = ElectricityMapsSection {
            api_key: Some("tok".to_string()),
            endpoint: Some("https://custom.api/v3".to_string()),
            poll_interval_secs: Some(120),
            region_map,
        };
        let cfg = convert_electricity_maps_section_with_env(&raw, || None).expect("should convert");
        assert!(cfg.region_map.contains_key("eu-west-3"));
        assert!(cfg.region_map.contains_key("us-east-1"));
        assert_eq!(cfg.api_endpoint, "https://custom.api/v3");
    }

    #[test]
    fn electricity_maps_env_var_takes_precedence_over_config_file() {
        // Env-lookup returns a token → it wins over `api_key` in the file.
        // Covers the from_env branch of convert_electricity_maps_section_with_env
        // without touching the real process environment.
        let mut region_map = HashMap::new();
        region_map.insert("eu-west-3".to_string(), "FR".to_string());
        let raw = ElectricityMapsSection {
            api_key: Some("from-file".to_string()),
            endpoint: None,
            poll_interval_secs: None,
            region_map,
        };
        let cfg = convert_electricity_maps_section_with_env(&raw, || Some("from-env".to_string()))
            .expect("env-supplied token should produce a valid config");
        assert_eq!(cfg.auth_token, "from-env");
    }

    // --- validate_electricity_maps error paths ---

    #[test]
    fn validate_electricity_maps_rejects_control_char_in_token() {
        let cfg = ElectricityMapsConfig {
            api_endpoint: "https://api.electricitymaps.com/v3".to_string(),
            auth_token: "tok\x07en".to_string(), // contains a control char
            poll_interval: Duration::from_secs(300),
            region_map: {
                let mut m = HashMap::new();
                m.insert("eu-west-3".to_string(), "FR".to_string());
                m
            },
        };
        let err = Config::validate_electricity_maps(&cfg).unwrap_err();
        assert!(err.contains("control"));
    }

    #[test]
    fn validate_electricity_maps_rejects_empty_region_map() {
        let cfg = ElectricityMapsConfig {
            api_endpoint: "https://api.electricitymaps.com/v3".to_string(),
            auth_token: "tok".to_string(),
            poll_interval: Duration::from_secs(300),
            region_map: HashMap::new(),
        };
        let err = Config::validate_electricity_maps(&cfg).unwrap_err();
        assert!(err.contains("region_map"));
    }

    #[test]
    fn validate_electricity_maps_rejects_empty_zone() {
        let mut region_map = HashMap::new();
        region_map.insert("eu-west-3".to_string(), String::new());
        let cfg = ElectricityMapsConfig {
            api_endpoint: "https://api.electricitymaps.com/v3".to_string(),
            auth_token: "tok".to_string(),
            poll_interval: Duration::from_secs(300),
            region_map,
        };
        let err = Config::validate_electricity_maps(&cfg).unwrap_err();
        assert!(err.contains("empty"));
    }

    #[test]
    fn validate_electricity_maps_rejects_invalid_poll_interval() {
        let mut region_map = HashMap::new();
        region_map.insert("eu-west-3".to_string(), "FR".to_string());
        let cfg = ElectricityMapsConfig {
            api_endpoint: "https://api.electricitymaps.com/v3".to_string(),
            auth_token: "tok".to_string(),
            poll_interval: Duration::from_secs(10), // below 60
            region_map,
        };
        let err = Config::validate_electricity_maps(&cfg).unwrap_err();
        assert!(err.contains("poll_interval"));
    }

    // ---------------------------------------------------------------
    // TLS validation
    // ---------------------------------------------------------------

    #[test]
    #[allow(clippy::field_reassign_with_default)]
    fn validate_tls_accepts_both_absent() {
        let cfg = Config::default();
        assert!(cfg.validate_tls().is_ok());
    }

    #[test]
    #[allow(clippy::field_reassign_with_default)]
    fn validate_tls_rejects_cert_without_key() {
        let mut cfg = Config::default();
        cfg.tls_cert_path = Some("/tmp/cert.pem".to_string());
        let err = cfg.validate_tls().unwrap_err();
        assert!(err.contains("tls_key_path is missing"), "{err}");
    }

    #[test]
    #[allow(clippy::field_reassign_with_default)]
    fn validate_tls_rejects_key_without_cert() {
        let mut cfg = Config::default();
        cfg.tls_key_path = Some("/tmp/key.pem".to_string());
        let err = cfg.validate_tls().unwrap_err();
        assert!(err.contains("tls_cert_path is missing"), "{err}");
    }

    #[test]
    #[allow(clippy::field_reassign_with_default)]
    fn validate_tls_rejects_nonexistent_cert() {
        let mut cfg = Config::default();
        cfg.tls_cert_path = Some("/nonexistent/cert.pem".to_string());
        cfg.tls_key_path = Some("/nonexistent/key.pem".to_string());
        let err = cfg.validate_tls().unwrap_err();
        assert!(err.contains("does not exist"), "{err}");
    }

    #[test]
    #[allow(clippy::field_reassign_with_default)]
    fn validate_tls_accepts_existing_files() {
        let dir = tempfile::tempdir().unwrap();
        let cert = dir.path().join("cert.pem");
        let key = dir.path().join("key.pem");
        std::fs::write(&cert, b"fake cert").unwrap();
        std::fs::write(&key, b"fake key").unwrap();

        let mut cfg = Config::default();
        cfg.tls_cert_path = Some(cert.to_str().unwrap().to_string());
        cfg.tls_key_path = Some(key.to_str().unwrap().to_string());
        assert!(cfg.validate_tls().is_ok());
    }

    #[test]
    fn tls_config_fields_round_trip_through_toml() {
        let dir = tempfile::tempdir().unwrap();
        let cert = dir.path().join("cert.pem");
        let key = dir.path().join("key.pem");
        std::fs::write(&cert, b"fake cert").unwrap();
        std::fs::write(&key, b"fake key").unwrap();
        let toml = format!(
            "[daemon]\ntls_cert_path = \"{}\"\ntls_key_path = \"{}\"",
            cert.display(),
            key.display()
        );
        let cfg = load_from_str(&toml).unwrap();
        assert_eq!(cfg.tls_cert_path.as_deref(), Some(cert.to_str().unwrap()));
        assert_eq!(cfg.tls_key_path.as_deref(), Some(key.to_str().unwrap()));
    }

    #[test]
    fn tls_config_defaults_to_none() {
        let cfg = load_from_str("").unwrap();
        assert!(cfg.tls_cert_path.is_none());
        assert!(cfg.tls_key_path.is_none());
    }

    #[test]
    #[allow(clippy::field_reassign_with_default)]
    fn validate_tls_rejects_control_chars_in_cert_path() {
        let mut cfg = Config::default();
        cfg.tls_cert_path = Some("/tmp/cert\x00.pem".to_string());
        cfg.tls_key_path = Some("/tmp/key.pem".to_string());
        let err = cfg.validate_tls().unwrap_err();
        assert!(err.contains("control characters"), "{err}");
    }

    #[test]
    #[allow(clippy::field_reassign_with_default)]
    fn validate_tls_rejects_control_chars_in_key_path() {
        let mut cfg = Config::default();
        cfg.tls_cert_path = Some("/tmp/cert.pem".to_string());
        cfg.tls_key_path = Some("/tmp/key\n.pem".to_string());
        let err = cfg.validate_tls().unwrap_err();
        assert!(err.contains("control characters"), "{err}");
    }
}