daaki_message/

parser.rs

//! RFC 5322 email message parser.
//!
//! Parses raw email message bytes into a structured [`ParsedEmail`] representation.
//! Handles partial messages (headers + truncated body) gracefully, extracting
//! whatever content is available.
//!
//! # References
//! - RFC 5322 (Internet Message Format)
//! - RFC 2045 (MIME Part One — body format, Content-Transfer-Encoding)
//! - RFC 2046 (MIME Part Two — media types, multipart boundaries)
//! - RFC 2047 (MIME Part Three — encoded words in headers)
//! - RFC 2183 (Content-Disposition)
//! - RFC 2231 (MIME parameter encoding)
//! - RFC 6532 (Internationalized email headers)

use base64::Engine as _;

use crate::error::Error;
use crate::types::{Address, DateTime, ParsedAttachment, ParsedEmail};

/// Maximum MIME nesting depth to prevent stack overflow on pathological input.
/// RFC 2046 does not specify a limit; 64 is generous for real-world messages.
const MAX_MIME_DEPTH: u32 = 64;

/// Lenient base64 engine that accepts both padded and unpadded input
/// (RFC 2045 Section 6.8).
const LENIENT_BASE64: base64::engine::GeneralPurpose = base64::engine::GeneralPurpose::new(
    &base64::alphabet::STANDARD,
    base64::engine::GeneralPurposeConfig::new()
        .with_decode_padding_mode(base64::engine::DecodePaddingMode::Indifferent),
);

/// Structured header fields extracted from an RFC 5322 message.
///
/// Used internally to deduplicate the shared header extraction logic
/// between [`parse_email`] and [`parse_headers_only`].
struct HeaderFields {
    message_id: Option<String>,
    in_reply_to: Option<String>,
    references: Option<String>,
    subject: Option<String>,
    from: Address,
    to: Vec<Address>,
    cc: Vec<Address>,
    bcc: Vec<Address>,
    reply_to: Vec<Address>,
    date: Option<DateTime>,
}

/// Extracts all structured header fields from parsed header pairs.
///
/// # References
/// - RFC 5322 (Internet Message Format — address, date-time, identification)
/// - RFC 2047 (MIME encoded words in headers)
fn extract_header_fields(headers: &[(String, String)]) -> Result<HeaderFields, Error> {
    Ok(HeaderFields {
        message_id: extract_message_id(headers),
        in_reply_to: extract_in_reply_to(headers),
        references: extract_references(headers),
        subject: get_header_value(headers, "subject").map(|v| decode_encoded_words(&v)),
        from: extract_from(headers)?,
        to: extract_address_list(headers, "to"),
        cc: extract_address_list(headers, "cc"),
        bcc: extract_address_list(headers, "bcc"),
        reply_to: extract_address_list(headers, "reply-to"),
        date: extract_date(headers),
    })
}

/// Parses raw email message bytes into a structured representation.
///
/// Handles partial messages (headers + truncated body) gracefully,
/// extracting whatever content is available. Returns [`Error::EmptyInput`]
/// for empty input, [`Error::MissingFrom`] if the `From` header is absent.
///
/// # References
/// - RFC 5322 (Internet Message Format)
/// - RFC 2045–2047 (MIME)
/// - RFC 2183 (Content-Disposition)
/// - RFC 2231 (MIME parameter encoding)
pub fn parse_email(raw: &[u8]) -> Result<ParsedEmail, Error> {
    if raw.is_empty() {
        return Err(Error::EmptyInput);
    }

    let size = raw.len() as u64;

    // Split headers and body at \r\n\r\n or \n\n (RFC 5322 Section 2.1)
    let (header_bytes, body_bytes) = split_header_body(raw);
    let raw_headers = String::from_utf8_lossy(header_bytes).into_owned();

    // Parse headers into (lowercase-name, decoded-value) pairs,
    // unfolding continuation lines per RFC 5322 Section 2.2.3
    let headers = parse_headers(header_bytes);
    let hf = extract_header_fields(&headers)?;

    // Determine Content-Type and walk MIME tree or extract simple body
    let content_type = get_header_value(&headers, "content-type")
        .unwrap_or_else(|| "text/plain; charset=us-ascii".to_string());
    let transfer_encoding =
        get_header_value(&headers, "content-transfer-encoding").unwrap_or_default();
    let content_disposition = get_header_value(&headers, "content-disposition").unwrap_or_default();
    let content_id = get_header_value(&headers, "content-id");

    let (body_text, body_html, attachments) = if is_multipart(&content_type) {
        match extract_boundary(&content_type) {
            Some(boundary) => {
                let is_digest = extract_mime_type(&content_type) == "multipart/digest";
                walk_mime_tree(body_bytes, &boundary, "", 0, is_digest)
            }
            // Multipart with no boundary parameter: gracefully degrade to
            // text/plain since we cannot split the MIME parts (RFC 2046
            // Section 5.1.1 — boundary is required for multipart).
            None => extract_simple_body(
                body_bytes,
                "text/plain; charset=us-ascii",
                &transfer_encoding,
                &content_disposition,
                content_id.as_deref(),
            ),
        }
    } else {
        extract_simple_body(
            body_bytes,
            &content_type,
            &transfer_encoding,
            &content_disposition,
            content_id.as_deref(),
        )
    };

    Ok(ParsedEmail {
        message_id: hf.message_id,
        in_reply_to: hf.in_reply_to,
        references: hf.references,
        subject: hf.subject,
        from: hf.from,
        to: hf.to,
        cc: hf.cc,
        bcc: hf.bcc,
        reply_to: hf.reply_to,
        date: hf.date,
        body_text,
        body_html,
        attachments,
        raw_headers,
        size,
    })
}

/// Parses only the headers of a raw email message, skipping body/MIME processing.
///
/// This is faster than [`parse_email`] when only metadata is needed (e.g.,
/// building a message list). Body-related fields (`body_text`, `body_html`,
/// `attachments`) are always empty/`None`.
///
/// # References
/// - RFC 5322 (Internet Message Format)
/// - RFC 2047 (encoded words)
pub fn parse_headers_only(raw: &[u8]) -> Result<ParsedEmail, Error> {
    if raw.is_empty() {
        return Err(Error::EmptyInput);
    }

    let size = raw.len() as u64;

    let (header_bytes, _) = split_header_body(raw);
    let raw_headers = String::from_utf8_lossy(header_bytes).into_owned();
    let headers = parse_headers(header_bytes);
    let hf = extract_header_fields(&headers)?;

    Ok(ParsedEmail {
        message_id: hf.message_id,
        in_reply_to: hf.in_reply_to,
        references: hf.references,
        subject: hf.subject,
        from: hf.from,
        to: hf.to,
        cc: hf.cc,
        bcc: hf.bcc,
        reply_to: hf.reply_to,
        date: hf.date,
        body_text: None,
        body_html: None,
        attachments: Vec::new(),
        raw_headers,
        size,
    })
}

// ---------------------------------------------------------------------------
// Header parsing
// ---------------------------------------------------------------------------

/// Splits raw message bytes into (headers, body) at the first blank line.
///
/// Tries `\r\n\r\n` first, falls back to `\n\n` (RFC 5322 Section 2.1).
/// Also handles MIME parts with no headers: if the input starts with
/// `\r\n` or `\n`, the header section is empty and the rest is body.
/// If no separator is found, the entire input is treated as headers.
fn split_header_body(raw: &[u8]) -> (&[u8], &[u8]) {
    // Handle MIME parts with no headers: input starts with a blank line
    // (RFC 2046 — a body-part may have an empty header section).
    if raw.starts_with(b"\r\n") {
        return (&[], &raw[2..]);
    }
    if raw.starts_with(b"\n") {
        return (&[], &raw[1..]);
    }
    if let Some(pos) = find_subsequence(raw, b"\r\n\r\n") {
        return (&raw[..pos], &raw[pos + 4..]);
    }
    if let Some(pos) = find_subsequence(raw, b"\n\n") {
        return (&raw[..pos], &raw[pos + 2..]);
    }
    // No body separator — treat entire input as headers
    (raw, &[])
}

/// Parses raw header bytes into `(lowercase_name, value)` pairs.
///
/// Unfolds continuation lines (lines starting with whitespace) per
/// RFC 5322 Section 2.2.3.
fn parse_headers(raw: &[u8]) -> Vec<(String, String)> {
    let text = String::from_utf8_lossy(raw);
    let mut headers: Vec<(String, String)> = Vec::new();
    let mut current_name = String::new();
    let mut current_value = String::new();

    for line in text.split('\n') {
        let line = line.strip_suffix('\r').unwrap_or(line);
        if line.is_empty() {
            break;
        }
        if line.starts_with(' ') || line.starts_with('\t') {
            // Continuation line — unfold by removing only the CRLF,
            // preserving all whitespace (RFC 5322 Section 2.2.3:
            // "Unfolding is accomplished by simply removing any CRLF
            // that is immediately followed by WSP.").
            if !current_name.is_empty() {
                current_value.push_str(line);
            }
        } else if let Some(colon_pos) = line.find(':') {
            // New header — save previous one
            if !current_name.is_empty() {
                headers.push((current_name.to_lowercase(), current_value));
            }
            current_name = line[..colon_pos].trim().to_string();
            // Trim only leading whitespace after the colon (optional WSP
            // per RFC 5322 Section 2.2). Trailing whitespace is preserved
            // because it may be significant when followed by continuation
            // lines (RFC 5322 Section 2.2.3).
            current_value = line[colon_pos + 1..].trim_start().to_string();
        }
        // Lines without a colon and not continuations are silently skipped
    }
    // Save the last header
    if !current_name.is_empty() {
        headers.push((current_name.to_lowercase(), current_value));
    }

    headers
}

/// Returns the value of the first header matching `name` (case-insensitive).
fn get_header_value(headers: &[(String, String)], name: &str) -> Option<String> {
    headers
        .iter()
        .find(|(k, _)| k == name)
        .map(|(_, v)| v.clone())
}

// ---------------------------------------------------------------------------
// RFC 2047 encoded-word decoding
// ---------------------------------------------------------------------------

/// Decodes RFC 2047 encoded words in a header value.
///
/// Handles `=?charset?encoding?text?=` tokens and collapses whitespace
/// between adjacent encoded words per RFC 2047 Section 6.2.
pub(crate) fn decode_encoded_words(input: &str) -> String {
    let mut result = String::new();
    let mut remaining = input;

    while !remaining.is_empty() {
        if let Some(start) = remaining.find("=?") {
            // Text before the encoded word
            result.push_str(&remaining[..start]);
            remaining = &remaining[start..];

            if let Some((decoded, consumed)) = try_decode_encoded_word(remaining) {
                result.push_str(&decoded);
                remaining = &remaining[consumed..];

                // Collapse whitespace between adjacent encoded words (RFC 2047 Section 6.2)
                let trimmed = remaining.trim_start_matches([' ', '\t']);
                if trimmed.starts_with("=?") {
                    remaining = trimmed;
                }
            } else {
                // Not a valid encoded word — emit literal "=?" and advance
                result.push_str("=?");
                remaining = &remaining[2..];
            }
        } else {
            result.push_str(remaining);
            break;
        }
    }

    result
}

/// Attempts to decode a single RFC 2047 encoded word at the start of `input`.
///
/// Returns `(decoded_text, bytes_consumed)` on success.
fn try_decode_encoded_word(input: &str) -> Option<(String, usize)> {
    // Format: =?charset?encoding?encoded_text?=
    let rest = input.strip_prefix("=?")?;
    let q1 = rest.find('?')?;
    let charset = &rest[..q1];
    let rest2 = &rest[q1 + 1..];
    let q2 = rest2.find('?')?;
    let encoding = &rest2[..q2];
    let rest3 = &rest2[q2 + 1..];
    let q3 = rest3.find("?=")?;
    let encoded_text = &rest3[..q3];
    let consumed = 2 + q1 + 1 + q2 + 1 + q3 + 2;

    let bytes = match encoding.to_ascii_uppercase().as_str() {
        "B" => LENIENT_BASE64.decode(encoded_text.as_bytes()).ok()?,
        "Q" => decode_q_encoding(encoded_text),
        _ => return None,
    };

    Some((decode_charset(charset, &bytes), consumed))
}

/// Decodes RFC 2047 Q-encoding (a variant of quoted-printable).
///
/// Underscores represent spaces; `=XX` represents a hex-encoded byte.
fn decode_q_encoding(input: &str) -> Vec<u8> {
    let bytes = input.as_bytes();
    let mut result = Vec::with_capacity(bytes.len());
    let mut i = 0;
    while i < bytes.len() {
        if bytes[i] == b'=' && i + 2 < bytes.len() {
            if let Some(val) = decode_hex_pair(bytes[i + 1], bytes[i + 2]) {
                result.push(val);
                i += 3;
                continue;
            }
        }
        if bytes[i] == b'_' {
            // Underscore represents space in Q-encoding (RFC 2047 Section 4.2)
            result.push(b' ');
        } else {
            result.push(bytes[i]);
        }
        i += 1;
    }
    result
}

// ---------------------------------------------------------------------------
// Address parsing (RFC 5322 Section 3.4)
// ---------------------------------------------------------------------------

/// Extracts the `From` address. Returns [`Error::MissingFrom`] if absent.
///
/// Address structure is parsed first on the raw header value, then RFC 2047
/// encoded words are decoded in each address's display name. Decoding before
/// parsing would break address splitting when an encoded-word display name
/// contains address-significant characters (`,`, `<`, `>`, `:`, `;`).
///
/// # References
/// - RFC 2047 Section 5 rule (3) — encoded-words in phrase context
/// - RFC 5322 Section 3.4 — address specification
fn extract_from(headers: &[(String, String)]) -> Result<Address, Error> {
    let value = get_header_value(headers, "from").ok_or(Error::MissingFrom)?;
    // Parse address structure FIRST on raw value, then decode display names.
    // RFC 2047 Section 5 rule (3): encoded-words appear in 'phrase' context.
    // Decoding before parsing corrupts addresses when the decoded text
    // contains commas or other address-significant characters.
    let addrs = decode_address_names(parse_address_list(&value));
    addrs.into_iter().next().ok_or(Error::MissingFrom)
}

/// Extracts an address list from the named header.
///
/// Parses address structure first, then decodes RFC 2047 encoded words in
/// display names — see [`extract_from`] for rationale.
fn extract_address_list(headers: &[(String, String)], name: &str) -> Vec<Address> {
    get_header_value(headers, name)
        .map(|v| decode_address_names(parse_address_list(&v)))
        .unwrap_or_default()
}

/// Decodes RFC 2047 encoded words in each address's display name.
///
/// Called after address structure parsing to avoid breaking address splitting
/// when encoded-word display names contain address-significant characters.
///
/// # References
/// - RFC 2047 Section 5 rule (3) — encoded-words may appear in phrases
fn decode_address_names(addrs: Vec<Address>) -> Vec<Address> {
    addrs
        .into_iter()
        .map(|mut addr| {
            if let Some(ref name) = addr.name {
                addr.name = Some(decode_encoded_words(name));
            }
            addr
        })
        .collect()
}

/// Parses a comma-separated address list, respecting quoted strings, angle
/// brackets, parenthesized comments, and RFC 5322 group syntax
/// (RFC 5322 Section 3.4).
///
/// Group syntax: `display-name ":" [group-list] ";"` where group-list is a
/// comma-separated list of mailboxes. The group wrapper is stripped and
/// member addresses are extracted. Empty groups (e.g., `undisclosed-recipients:;`)
/// produce no addresses.
///
/// Parenthesized comments (RFC 5322 Section 3.2.2) may appear in addr-spec
/// CFWS contexts and can contain commas, angle brackets, and other
/// address-significant characters. These must not be treated as address
/// separators.
fn parse_address_list(input: &str) -> Vec<Address> {
    let mut addresses = Vec::new();
    let mut current = String::new();
    let mut in_quotes = false;
    let mut escaped = false;
    let mut angle_depth: i32 = 0;
    // Track parenthesized comment depth (RFC 5322 Section 3.2.2).
    // Commas and other structural characters inside comments must not
    // be treated as address separators.
    let mut paren_depth: i32 = 0;
    // Track whether we're inside a group construct (after ':' but before ';').
    // RFC 5322 Section 3.4: group = display-name ":" [group-list] ";"
    let mut in_group = false;

    for ch in input.chars() {
        // Inside a quoted-string, a backslash escapes the next character
        // (RFC 5322 Section 3.2.4 quoted-pair).
        if escaped {
            current.push(ch);
            escaped = false;
            continue;
        }
        match ch {
            '\\' if in_quotes || paren_depth > 0 => {
                // Backslash escapes next character in quoted-strings
                // (RFC 5322 Section 3.2.4) and inside comments
                // (RFC 5322 Section 3.2.2 quoted-pair in ccontent).
                escaped = true;
                current.push(ch);
            }
            '"' if paren_depth == 0 => {
                in_quotes = !in_quotes;
                current.push(ch);
            }
            // RFC 5322 Section 3.2.2: parenthesized comments may be nested.
            // Track depth so that commas inside comments are not treated as
            // address separators.
            '(' if !in_quotes => {
                paren_depth += 1;
                current.push(ch);
            }
            ')' if !in_quotes && paren_depth > 0 => {
                paren_depth -= 1;
                current.push(ch);
            }
            '<' if !in_quotes && paren_depth == 0 => {
                angle_depth += 1;
                current.push(ch);
            }
            '>' if !in_quotes && paren_depth == 0 => {
                angle_depth -= 1;
                current.push(ch);
            }
            // RFC 5322 Section 3.4: ':' starts a group construct when
            // we're not inside quotes, angle brackets, comments, or an
            // existing group.
            // Heuristic: only treat as group if the current token contains
            // no '@' (i.e., it's a display-name, not a bare addr-spec).
            ':' if !in_quotes && angle_depth == 0 && paren_depth == 0 && !in_group => {
                if current.trim().contains('@') {
                    current.push(ch);
                } else {
                    // Enter group: discard the display-name portion
                    in_group = true;
                    current.clear();
                }
            }
            // RFC 5322 Section 3.4: ';' terminates the group construct.
            ';' if !in_quotes && angle_depth == 0 && paren_depth == 0 && in_group => {
                // Emit any pending address inside the group
                if let Some(addr) = parse_single_address(&current) {
                    addresses.push(addr);
                }
                current.clear();
                in_group = false;
            }
            ',' if !in_quotes && angle_depth == 0 && paren_depth == 0 => {
                if let Some(addr) = parse_single_address(&current) {
                    addresses.push(addr);
                }
                current.clear();
            }
            _ => current.push(ch),
        }
    }
    if let Some(addr) = parse_single_address(&current) {
        addresses.push(addr);
    }

    addresses
}

/// Parses a single address: either `Display Name <email>` or bare `email`.
///
/// Handles RFC 5322 Section 3.2.2 comments (parenthesized text) that may
/// appear before or after a bare addr-spec per Section 3.4.1 CFWS rules.
/// A trailing comment like `(Display Name)` is used as the display name,
/// following the common RFC 822 convention.
fn parse_single_address(input: &str) -> Option<Address> {
    let input = input.trim();
    if input.is_empty() {
        return None;
    }

    // Try "Display Name <email@domain>" form (RFC 5322 Section 3.4)
    if let Some(angle_start) = input.rfind('<') {
        if let Some(angle_end) = input.rfind('>') {
            if angle_end > angle_start {
                let email = input[angle_start + 1..angle_end].trim().to_string();
                let name_part = input[..angle_start].trim();
                let name = if name_part.is_empty() {
                    None
                } else {
                    // Strip only the outer pair of quotes from a quoted-string
                    // (RFC 5322 Section 3.2.4). Using trim_matches('"') would
                    // greedily strip multiple quotes and corrupt escaped quotes
                    // like `\"` at the end of the display name.
                    let name = strip_outer_quotes(name_part).trim().to_string();
                    if name.is_empty() {
                        None
                    } else {
                        // Unescape quoted-pair sequences (RFC 5322 Section 3.2.4)
                        Some(unescape_quoted_string(&name))
                    }
                };
                if !email.is_empty() {
                    return Some(Address { name, email });
                }
            }
        }
    }

    // Bare email address — may have RFC 5322 Section 3.2.2 comments
    // (parenthesized text) before or after the addr-spec per Section 3.4.1.
    if input.contains('@') {
        // Check for a trailing comment like "user@example.com (Display Name)".
        // RFC 822 convention: trailing parenthesized comment is the display name.
        if let Some(paren_start) = input.find('(') {
            let email_part = input[..paren_start].trim();
            // Extract the comment content (text between outermost parens)
            // to use as display name for trailing comments only.
            let after_email = input[paren_start..].trim();
            let name = if !email_part.is_empty() && email_part.contains('@') {
                // Trailing comment: extract text between parentheses
                // as display name (RFC 822 convention, RFC 5322 Section 3.4.1 CFWS)
                extract_comment_text(after_email)
            } else {
                None
            };
            // Strip all comments to get the bare addr-spec
            // (RFC 5322 Section 3.2.2)
            let stripped = strip_comments(input);
            let email = stripped.trim().to_string();
            if !email.is_empty() && email.contains('@') {
                return Some(Address { name, email });
            }
        }
        return Some(Address {
            name: None,
            email: input.to_string(),
        });
    }

    None
}

/// Extracts the text content from a parenthesized RFC 5322 comment string.
///
/// Given a string like `(Display Name)`, returns `Some("Display Name")`.
/// Handles nested parentheses and backslash-escaped characters per
/// RFC 5322 Section 3.2.2.
fn extract_comment_text(s: &str) -> Option<String> {
    let s = s.trim();
    if !s.starts_with('(') {
        return None;
    }
    // Find the matching closing paren, respecting nesting and escapes
    let mut depth: u32 = 0;
    let mut result = String::new();
    let mut escaped = false;
    let mut started = false;
    for c in s.chars() {
        if escaped {
            escaped = false;
            result.push(c);
            continue;
        }
        match c {
            '\\' => {
                escaped = true;
            }
            '(' => {
                if started {
                    // Nested paren — include literally
                    result.push(c);
                }
                depth = depth.saturating_add(1);
                started = true;
            }
            ')' => {
                depth = depth.saturating_sub(1);
                if depth == 0 {
                    break;
                }
                // Nested closing paren — include literally
                result.push(c);
            }
            _ => {
                if depth > 0 {
                    result.push(c);
                }
            }
        }
    }
    let trimmed = result.trim().to_string();
    if trimmed.is_empty() {
        None
    } else {
        Some(trimmed)
    }
}

// ---------------------------------------------------------------------------
// Message-ID / In-Reply-To / References extraction (RFC 5322 Section 3.6.4)
// ---------------------------------------------------------------------------

/// Extracts Message-ID, stripping angle brackets if present.
///
/// Handles both RFC-compliant `<id@host>` form and bare `id@host` form
/// for tolerance of broken mailers (RFC 5322 Section 3.6.4).
fn extract_message_id(headers: &[(String, String)]) -> Option<String> {
    get_header_value(headers, "message-id").and_then(|v| {
        // Try bracketed form first (RFC 5322 Section 3.6.4)
        if let Some(id) = extract_first_msg_id(&v) {
            return Some(id);
        }
        // Fall back to bare form only if no angle brackets are present
        // (tolerates broken mailers that omit brackets entirely)
        let trimmed = v.trim();
        if trimmed.is_empty() || trimmed.contains('<') || trimmed.contains('>') {
            None
        } else {
            Some(trimmed.to_string())
        }
    })
}

/// Extracts the first message-id from In-Reply-To (may contain multiple).
fn extract_in_reply_to(headers: &[(String, String)]) -> Option<String> {
    get_header_value(headers, "in-reply-to").and_then(|v| extract_first_msg_id(&v))
}

/// Extracts all message-ids from References, space-joined.
fn extract_references(headers: &[(String, String)]) -> Option<String> {
    get_header_value(headers, "references").and_then(|v| {
        let ids = extract_all_msg_ids(&v);
        if ids.is_empty() {
            None
        } else {
            Some(ids.join(" "))
        }
    })
}

/// Extracts the first `<...>` message-id from a header value.
fn extract_first_msg_id(value: &str) -> Option<String> {
    let start = value.find('<')?;
    let end = value[start..].find('>')? + start;
    let id = value[start + 1..end].trim();
    if id.is_empty() {
        None
    } else {
        Some(id.to_string())
    }
}

/// Extracts all `<...>` message-ids from a header value.
fn extract_all_msg_ids(value: &str) -> Vec<String> {
    let mut ids = Vec::new();
    let mut remaining = value;
    while let Some(start) = remaining.find('<') {
        remaining = &remaining[start + 1..];
        if let Some(end) = remaining.find('>') {
            let id = remaining[..end].trim();
            if !id.is_empty() {
                ids.push(id.to_string());
            }
            remaining = &remaining[end + 1..];
        } else {
            break;
        }
    }
    ids
}

// ---------------------------------------------------------------------------
// Date parsing (RFC 5322 Section 3.3)
// ---------------------------------------------------------------------------

/// Attempts to parse the `Date` header.
fn extract_date(headers: &[(String, String)]) -> Option<DateTime> {
    get_header_value(headers, "date").and_then(|v| parse_rfc5322_date(&v))
}

/// Parses an RFC 5322 date-time string.
///
/// Accepts: `[day-of-week ","] day month year hour ":" minute [":" second] zone`
///
/// Strips CFWS (comments and folding white space) before parsing, as allowed
/// by the obsolete date syntax (RFC 5322 Section 4.3).
///
/// # References
/// - RFC 5322 Section 3.3
/// - RFC 5322 Section 4.3 (obsolete syntax — CFWS between tokens)
pub(crate) fn parse_rfc5322_date(input: &str) -> Option<DateTime> {
    let input = strip_comments(input);
    let input = input.trim();

    // Skip optional day-of-week
    let input = if let Some(comma_pos) = input.find(',') {
        input[comma_pos + 1..].trim()
    } else {
        input
    };

    let parts: Vec<&str> = input.split_whitespace().collect();
    if parts.len() < 4 {
        return None;
    }

    let day: u8 = parts[0].parse().ok()?;
    let month = parse_month_name(parts[1])?;
    let year: u16 = parse_year(parts[2])?;

    let time_parts: Vec<&str> = parts[3].split(':').collect();
    if time_parts.len() < 2 {
        return None;
    }

    let hour: u8 = time_parts[0].parse().ok()?;
    let minute: u8 = time_parts[1].parse().ok()?;
    let second: u8 = time_parts.get(2).and_then(|s| s.parse().ok()).unwrap_or(0);

    // Validate field ranges per RFC 5322 Section 3.3:
    //   day   = 1-31, hour = 0-23, minute = 0-59, second = 0-60 (60 = leap second)
    if day == 0 || day > 31 || hour > 23 || minute > 59 || second > 60 {
        return None;
    }

    let tz_offset_minutes = parts.get(4).map_or(0, |tz| parse_timezone(tz));

    Some(DateTime {
        year,
        month,
        day,
        hour,
        minute,
        second,
        tz_offset_minutes,
    })
}

/// Parses a three-letter month abbreviation (RFC 5322 Section 3.3).
fn parse_month_name(s: &str) -> Option<u8> {
    match s.to_ascii_lowercase().as_str() {
        "jan" => Some(1),
        "feb" => Some(2),
        "mar" => Some(3),
        "apr" => Some(4),
        "may" => Some(5),
        "jun" => Some(6),
        "jul" => Some(7),
        "aug" => Some(8),
        "sep" => Some(9),
        "oct" => Some(10),
        "nov" => Some(11),
        "dec" => Some(12),
        _ => None,
    }
}

/// Parses a year, handling 2-digit obsolete years (RFC 5322 Section 4.3).
fn parse_year(s: &str) -> Option<u16> {
    let y: u16 = s.parse().ok()?;
    if y < 100 {
        // RFC 5322 Section 4.3: 2-digit years 00-49 → +2000, 50-99 → +1900
        Some(if y >= 50 { 1900 + y } else { 2000 + y })
    } else if y < 1000 {
        // RFC 5322 Section 4.3: 3-digit years have 1900 added
        Some(1900 + y)
    } else {
        Some(y)
    }
}

/// Parses a timezone offset: `+HHMM`, `-HHMM`, or named zones (RFC 5322 Section 4.3).
fn parse_timezone(s: &str) -> i16 {
    let s = s.trim();
    // Numeric offset: +HHMM or -HHMM
    if (s.starts_with('+') || s.starts_with('-')) && s.len() >= 5 {
        let sign: i16 = if s.starts_with('-') { -1 } else { 1 };
        if let (Ok(h), Ok(m)) = (s[1..3].parse::<i16>(), s[3..5].parse::<i16>()) {
            return sign * (h * 60 + m);
        }
    }
    // Named zones (RFC 5322 Section 4.3 / obsolete)
    match s.to_ascii_uppercase().as_str() {
        "EST" | "CDT" => -300,
        "EDT" => -240,
        "CST" | "MDT" => -360,
        "MST" | "PDT" => -420,
        "PST" => -480,
        // UT, UTC, GMT, and unknown zones all map to +0000
        _ => 0,
    }
}

// ---------------------------------------------------------------------------
// MIME tree walking (RFC 2046, RFC 3501 Section 6.4.5)
// ---------------------------------------------------------------------------

/// Walks the MIME tree, extracting body text, body HTML, and attachments
/// with computed IMAP section numbers (dot notation).
///
/// `is_digest` indicates the parent is `multipart/digest`, in which case the
/// default Content-Type for parts without an explicit header is
/// `message/rfc822` instead of `text/plain; charset=us-ascii`
/// (RFC 2046 Section 5.1.5).
fn walk_mime_tree(
    body: &[u8],
    boundary: &str,
    section_prefix: &str,
    depth: u32,
    is_digest: bool,
) -> (Option<String>, Option<String>, Vec<ParsedAttachment>) {
    if depth > MAX_MIME_DEPTH {
        return (None, None, Vec::new());
    }

    let parts = split_mime_parts(body, boundary);
    let mut body_text: Option<String> = None;
    let mut body_html: Option<String> = None;
    let mut attachments: Vec<ParsedAttachment> = Vec::new();

    for (i, part) in parts.iter().enumerate() {
        let section_num = i + 1;
        let section = if section_prefix.is_empty() {
            section_num.to_string()
        } else {
            format!("{section_prefix}.{section_num}")
        };

        let (part_header_bytes, part_body) = split_header_body(part);
        let part_headers = parse_headers(part_header_bytes);

        // RFC 2045 Section 5.2: default Content-Type is "text/plain; charset=us-ascii".
        // RFC 2046 Section 5.1.5: inside multipart/digest, the default is
        // "message/rfc822" instead.
        let default_ct = if is_digest {
            "message/rfc822"
        } else {
            "text/plain; charset=us-ascii"
        };
        let ct = get_header_value(&part_headers, "content-type")
            .unwrap_or_else(|| default_ct.to_string());
        let cte = get_header_value(&part_headers, "content-transfer-encoding").unwrap_or_default();
        let cd = get_header_value(&part_headers, "content-disposition").unwrap_or_default();
        let content_id = get_header_value(&part_headers, "content-id");

        if is_multipart(&ct) {
            // Recurse into nested multipart (RFC 2046 Section 5.1)
            if let Some(inner_boundary) = extract_boundary(&ct) {
                let inner_digest = extract_mime_type(&ct) == "multipart/digest";
                let (t, h, a) = walk_mime_tree(
                    part_body,
                    &inner_boundary,
                    &section,
                    depth + 1,
                    inner_digest,
                );
                if body_text.is_none() {
                    body_text = t;
                }
                if body_html.is_none() {
                    body_html = h;
                }
                attachments.extend(a);
            }
        } else {
            let mime = extract_mime_type(&ct);
            let cd_lower = cd.to_lowercase();
            let is_explicit_attachment = cd_lower.starts_with("attachment");

            if !is_explicit_attachment && mime == "text/plain" && body_text.is_none() {
                // An empty decoded body is semantically absent — treat it as
                // None for round-trip consistency (RFC 2046 Section 5.1.1).
                let decoded = decode_body(part_body, &cte, &ct);
                if !decoded.is_empty() {
                    body_text = Some(decoded);
                }
            } else if !is_explicit_attachment && mime == "text/html" && body_html.is_none() {
                // Same empty-body treatment for HTML parts.
                let decoded = decode_body(part_body, &cte, &ct);
                if !decoded.is_empty() {
                    body_html = Some(decoded);
                }
            } else if !mime.starts_with("multipart/") {
                // Attachment: explicit attachment, non-text part, or extra text part
                let is_inline = cd_lower.starts_with("inline") || content_id.is_some();
                let filename = extract_filename(&cd, &ct);

                attachments.push(ParsedAttachment {
                    filename,
                    content_type: mime,
                    // RFC 2392: Content-ID is `"<" addr-spec ">"`. Strip
                    // brackets and trim whitespace that some mailers add
                    // inside the brackets.
                    content_id: content_id
                        .map(|s| s.trim_matches(|c| c == '<' || c == '>').trim().to_string()),
                    is_inline,
                    size: Some(part_body.len() as u64),
                    section: Some(section),
                });
            }
        }
    }

    (body_text, body_html, attachments)
}

/// Splits a multipart body into its component parts using the given boundary.
///
/// Handles both `\r\n` and `\n` line endings, and tolerates truncated input
/// (missing closing boundary).
///
/// # References
/// - RFC 2046 Section 5.1.1
fn split_mime_parts<'a>(body: &'a [u8], boundary: &str) -> Vec<&'a [u8]> {
    let delim = format!("--{boundary}");
    let delim_bytes = delim.as_bytes();
    let end_delim = format!("--{boundary}--");
    let end_delim_bytes = end_delim.as_bytes();

    let mut parts: Vec<&'a [u8]> = Vec::new();
    let mut search_from: usize = 0;
    let mut part_start: Option<usize> = None;

    loop {
        let Some(rel_pos) = find_subsequence(&body[search_from..], delim_bytes) else {
            // No more boundaries — include trailing content if a part was started
            // (tolerance for truncated input per requirements)
            if let Some(start) = part_start {
                if start < body.len() {
                    parts.push(&body[start..]);
                }
            }
            break;
        };
        let pos = search_from + rel_pos;

        // RFC 2046 Section 5.1.1: boundary delimiters must appear at the
        // beginning of a line (position 0, or preceded by LF).
        // Skip mid-line matches and continue searching.
        if pos > 0 && body[pos - 1] != b'\n' {
            search_from = pos + delim_bytes.len();
            continue;
        }

        // Save content from previous boundary to this one
        if let Some(start) = part_start {
            let end = if pos >= 2 && body[pos - 2] == b'\r' && body[pos - 1] == b'\n' {
                pos - 2
            } else if pos >= 1 && body[pos - 1] == b'\n' {
                pos - 1
            } else {
                pos
            };
            if start <= end {
                parts.push(&body[start..end]);
            }
        }

        // Check for closing boundary
        if body[pos..].starts_with(end_delim_bytes) {
            break;
        }

        // Advance past the boundary line to the start of the next part
        let mut next = pos + delim_bytes.len();
        // Skip optional trailing whitespace on boundary line
        while next < body.len() && (body[next] == b' ' || body[next] == b'\t') {
            next += 1;
        }
        if next < body.len() && body[next] == b'\r' {
            next += 1;
        }
        if next < body.len() && body[next] == b'\n' {
            next += 1;
        }

        part_start = Some(next);
        search_from = next;
    }

    parts
}

// ---------------------------------------------------------------------------
// Body decoding
// ---------------------------------------------------------------------------

/// Extracts body content from a non-multipart message.
///
/// Checks Content-Disposition and MIME type to determine whether the content
/// is body text, body HTML, or an attachment (RFC 2046; RFC 2183).
///
/// # References
/// - RFC 2045 Section 5.2 (default Content-Type)
/// - RFC 2046 (media types)
/// - RFC 2183 (Content-Disposition)
fn extract_simple_body(
    body: &[u8],
    content_type: &str,
    transfer_encoding: &str,
    content_disposition: &str,
    content_id: Option<&str>,
) -> (Option<String>, Option<String>, Vec<ParsedAttachment>) {
    if body.is_empty() {
        return (None, None, Vec::new());
    }

    let mime = extract_mime_type(content_type);
    let cd_lower = content_disposition.to_lowercase();
    let is_explicit_attachment = cd_lower.starts_with("attachment");

    // Content-Disposition: attachment overrides MIME type (RFC 2183 Section 2).
    // Non-text MIME types are always attachments regardless of disposition
    // (requirements: "A part is an attachment if it has Content-Disposition:
    // attachment, or is a non-text/non-multipart part").
    if is_explicit_attachment || (mime != "text/plain" && mime != "text/html") {
        let is_inline = cd_lower.starts_with("inline") || content_id.is_some();
        let filename = extract_filename(content_disposition, content_type);

        let attachment = ParsedAttachment {
            filename,
            content_type: mime,
            // RFC 2392: Content-ID is `"<" addr-spec ">"`. Strip brackets
            // and trim whitespace that some mailers add inside the brackets.
            content_id: content_id
                .map(|s| s.trim_matches(|c| c == '<' || c == '>').trim().to_string()),
            is_inline,
            size: Some(body.len() as u64),
            // Single-part message body is section "1" per RFC 3501 Section 6.4.5
            section: Some("1".to_string()),
        };
        return (None, None, vec![attachment]);
    }

    let text = decode_body(body, transfer_encoding, content_type);

    // An empty decoded body is semantically absent — treat it as None rather
    // than Some(""). This ensures round-trip consistency: a message built with
    // no body content (e.g., an empty text/plain part in multipart/mixed)
    // parses back as None, not Some(""). The builder's write_text_part appends
    // a trailing CRLF (RFC 2046 Section 5.1.1), which decode_body strips,
    // leaving an empty string for originally-empty bodies.
    if text.is_empty() {
        return (None, None, Vec::new());
    }

    if mime == "text/html" {
        (None, Some(text), Vec::new())
    } else {
        // text/plain (RFC 2045 Section 5.2)
        (Some(text), None, Vec::new())
    }
}

/// Decodes a body part: applies Content-Transfer-Encoding, then charset conversion.
///
/// When no `charset` parameter is present, defaults to `us-ascii` per
/// RFC 2045 Section 5.2.
fn decode_body(data: &[u8], transfer_encoding: &str, content_type: &str) -> String {
    let decoded = decode_transfer_encoding(data, transfer_encoding);
    // RFC 2045 Section 5.2: default charset is US-ASCII
    let charset = extract_param(content_type, "charset").unwrap_or_else(|| "us-ascii".to_string());
    let text = decode_charset(&charset, &decoded);
    // Strip a single trailing CRLF or LF. In single-part messages the body
    // typically ends with CRLF as a message-format artifact (RFC 5322
    // Section 3.5), not semantic content. In multipart parts,
    // split_mime_parts strips the CRLF that serves as the boundary
    // delimiter prefix (RFC 2046 Section 5.1.1), but the builder's
    // write_text_part appends an additional CRLF after the body content
    // which this strip removes. For externally-produced multipart messages
    // where the part body itself ends with CRLF, one CRLF will also be
    // stripped — consistent with the single-part behavior.
    if let Some(stripped) = text.strip_suffix("\r\n") {
        stripped.to_string()
    } else if let Some(stripped) = text.strip_suffix('\n') {
        stripped.to_string()
    } else {
        text
    }
}

/// Applies Content-Transfer-Encoding decoding (RFC 2045 Section 6).
fn decode_transfer_encoding(data: &[u8], encoding: &str) -> Vec<u8> {
    match encoding.trim().to_ascii_lowercase().as_str() {
        "base64" => {
            // RFC 2045 Section 6.8: "Any characters outside of the base64
            // alphabet are to be ignored in base64-encoded data."
            // Keep only valid base64 alphabet characters: A-Z, a-z, 0-9, +, /, =
            let cleaned: Vec<u8> = data
                .iter()
                .copied()
                .filter(|b| b.is_ascii_alphanumeric() || *b == b'+' || *b == b'/' || *b == b'=')
                .collect();
            LENIENT_BASE64
                .decode(&cleaned)
                .unwrap_or_else(|_| data.to_vec())
        }
        "quoted-printable" => decode_quoted_printable(data),
        // 7bit, 8bit, binary — pass through (RFC 2045 Section 6.2)
        _ => data.to_vec(),
    }
}

/// Decodes quoted-printable encoding (RFC 2045 Section 6.7).
fn decode_quoted_printable(data: &[u8]) -> Vec<u8> {
    let mut result = Vec::with_capacity(data.len());
    let mut i = 0;
    while i < data.len() {
        if data[i] == b'=' {
            if i + 2 < data.len() {
                // Soft line break: =\r\n
                if data[i + 1] == b'\r' && i + 2 < data.len() && data[i + 2] == b'\n' {
                    i += 3;
                    continue;
                }
                // Soft line break: =\n
                if data[i + 1] == b'\n' {
                    i += 2;
                    continue;
                }
                // Hex-encoded byte
                if let Some(val) = decode_hex_pair(data[i + 1], data[i + 2]) {
                    result.push(val);
                    i += 3;
                    continue;
                }
            } else if i + 1 < data.len() && data[i + 1] == b'\n' {
                // Soft line break at end: =\n
                i += 2;
                continue;
            } else if i + 1 < data.len() && data[i + 1] == b'\r' {
                // Soft line break: =\r (bare CR without LF)
                i += 2;
                continue;
            } else if i + 1 == data.len() {
                // Trailing '=' at end-of-data is a soft line break
                // (RFC 2045 Section 6.7) — skip it.
                break;
            }
            // Malformed — fall through to push literal byte
        }
        result.push(data[i]);
        i += 1;
    }
    result
}

/// Converts bytes from the given charset to UTF-8 using lossy conversion.
///
/// Falls back to UTF-8 lossy conversion for unknown charsets.
///
/// # References
/// - RFC 2047 Section 2 (charset names)
/// - RFC 6532 (UTF-8 headers)
fn decode_charset(charset: &str, bytes: &[u8]) -> String {
    let charset_lower = charset.to_lowercase();
    if charset_lower == "utf-8" || charset_lower == "utf8" {
        return String::from_utf8_lossy(bytes).into_owned();
    }

    let encoding =
        encoding_rs::Encoding::for_label(charset.as_bytes()).unwrap_or(encoding_rs::UTF_8);
    let (decoded, _, _) = encoding.decode(bytes);
    decoded.into_owned()
}

// ---------------------------------------------------------------------------
// MIME parameter extraction
// ---------------------------------------------------------------------------

/// Checks whether a Content-Type is `multipart/*`.
fn is_multipart(content_type: &str) -> bool {
    extract_mime_type(content_type).starts_with("multipart/")
}

/// Extracts the MIME type (e.g., `text/plain`) from a full Content-Type value.
///
/// Strips RFC 5322 Section 3.2.2 parenthesized comments that may appear in
/// CFWS positions within the type/subtype production (RFC 2045 Section 5.1).
fn extract_mime_type(content_type: &str) -> String {
    let ct = content_type.trim();
    let end = ct.find(';').unwrap_or(ct.len());
    // Strip RFC 5322 Section 3.2.2 comments that may appear in CFWS
    // positions within the type/subtype production.
    let raw = ct[..end].trim();
    strip_comments(raw).trim().to_lowercase()
}

/// Extracts the `boundary` parameter from a Content-Type header (RFC 2046 Section 5.1.1).
fn extract_boundary(content_type: &str) -> Option<String> {
    extract_param(content_type, "boundary")
}

/// Extracts a named parameter from a header value.
///
/// Handles both quoted and unquoted values. Quoted-string values are
/// unescaped per RFC 5322 Section 3.2.4 (backslash-escaped pairs like
/// `\\` → `\` and `\"` → `"`).
///
/// Uses ASCII-only lowercasing to preserve byte-offset alignment with the
/// original string when non-ASCII characters are present (RFC 6532).
///
/// Returns `true` if `pos` is at the start of the string or preceded by a
/// parameter delimiter (`;`, space, or tab).  Used to reject substring
/// matches like `xfilename=` when searching for `filename=`.
fn is_param_boundary(lower: &str, pos: usize) -> bool {
    pos == 0
        || lower
            .as_bytes()
            .get(pos - 1)
            .is_some_and(|&c| c == b';' || c == b' ' || c == b'\t')
}

/// Extracts a quoted or unquoted parameter value from `rest` (the text
/// immediately after `param_name=`).
///
/// Handles quoted-strings with backslash escaping per RFC 5322 Section 3.2.4,
/// and unquoted tokens terminated by `;` or whitespace.
fn extract_param_value(rest: &str) -> Option<String> {
    let value = if let Some(stripped) = rest.strip_prefix('"') {
        // Find closing quote, skipping escaped quotes (RFC 5322 Section 3.2.4)
        let end = find_closing_quote(stripped);
        &stripped[..end]
    } else {
        let end = rest
            .find(|c: char| c == ';' || c.is_whitespace())
            .unwrap_or(rest.len());
        &rest[..end]
    };
    if value.is_empty() {
        None
    } else if rest.starts_with('"') {
        // Unescape quoted-pair sequences (RFC 5322 Section 3.2.4)
        Some(unescape_quoted_string(value))
    } else {
        Some(value.to_string())
    }
}

fn extract_param(header_value: &str, param_name: &str) -> Option<String> {
    // ASCII-only lowercasing preserves byte length for non-ASCII characters,
    // ensuring byte offsets from the lowered string match the original.
    // Full Unicode to_lowercase() can change byte length (e.g., İ: 2→3 bytes),
    // which would misalign indexing into the original string.
    let lower = header_value.to_ascii_lowercase();
    let pattern = format!("{param_name}=");
    let mut search_from = 0;

    loop {
        let idx = lower[search_from..].find(&pattern)?;
        let abs_idx = search_from + idx;

        // Ensure we're matching a parameter boundary, not a substring
        if is_param_boundary(&lower, abs_idx) {
            // Skip matches that fall inside a quoted-string value of another
            // parameter (RFC 5322 Section 3.2.4).
            if is_inside_quotes(&lower, abs_idx) {
                search_from = abs_idx + pattern.len();
                continue;
            }

            let rest = &header_value[abs_idx + pattern.len()..];
            return extract_param_value(rest);
        }

        search_from = abs_idx + pattern.len();
    }
}

/// Finds the position of the closing (unescaped) double-quote in a quoted-string body.
///
/// Skips backslash-escaped characters (`\"`, `\\`) per RFC 5322 Section 3.2.4.
/// Returns the byte offset of the closing `"`, or the string length if no
/// unescaped quote is found.
fn find_closing_quote(s: &str) -> usize {
    let bytes = s.as_bytes();
    let mut i = 0;
    while i < bytes.len() {
        if bytes[i] == b'\\' {
            // Skip escaped character (quoted-pair per RFC 5322 Section 3.2.4)
            i += 2;
            continue;
        }
        if bytes[i] == b'"' {
            return i;
        }
        i += 1;
    }
    bytes.len()
}

/// Returns `true` if the byte position `pos` falls inside a quoted-string.
///
/// Scans the string from the beginning, counting unescaped double-quote
/// characters (skipping backslash-escaped pairs). If the count of unescaped
/// quotes before `pos` is odd, the position is inside a quoted-string.
///
/// # References
/// - RFC 5322 Section 3.2.4 (quoted-string and quoted-pair)
fn is_inside_quotes(s: &str, pos: usize) -> bool {
    let bytes = s.as_bytes();
    let mut quote_count: u32 = 0;
    let mut i = 0;
    while i < pos && i < bytes.len() {
        if bytes[i] == b'\\' {
            // Skip escaped character (quoted-pair per RFC 5322 Section 3.2.4)
            i += 2;
            continue;
        }
        if bytes[i] == b'"' {
            quote_count += 1;
        }
        i += 1;
    }
    // Odd count means we are between an opening and closing quote
    quote_count % 2 != 0
}

/// Extracts filename from Content-Disposition and Content-Type parameters.
///
/// Tries (in order):
/// 1. RFC 2231 `filename*` (charset-encoded, non-continuation)
/// 2. RFC 2231 `filename*0` / `filename*0*` (continuation parameters)
/// 3. Plain `filename` parameter (RFC 2183)
/// 4. Same search order for `name` in Content-Type
///
/// # References
/// - RFC 2183 Section 2 (Content-Disposition parameters)
/// - RFC 2231 Section 3 (parameter continuation)
/// - RFC 2231 Section 4 (parameter value encoding)
fn extract_filename(disposition: &str, content_type: &str) -> Option<String> {
    // Try RFC 2231 filename* first (non-continuation)
    if let Some(name) = extract_rfc2231_param(disposition, "filename") {
        return Some(name);
    }
    // Try RFC 2231 continuation: filename*0, filename*1, ...
    if let Some(name) = extract_rfc2231_continuation(disposition, "filename") {
        return Some(name);
    }
    // Try plain filename parameter (RFC 2183)
    if let Some(name) = extract_param(disposition, "filename") {
        return Some(decode_encoded_words(&name));
    }
    // Try Content-Type name* parameter
    if let Some(name) = extract_rfc2231_param(content_type, "name") {
        return Some(name);
    }
    // Try RFC 2231 continuation: name*0, name*1, ...
    if let Some(name) = extract_rfc2231_continuation(content_type, "name") {
        return Some(name);
    }
    // Try Content-Type name parameter
    if let Some(name) = extract_param(content_type, "name") {
        return Some(decode_encoded_words(&name));
    }
    None
}

/// Extracts and decodes an RFC 2231 encoded parameter (`param*=charset'lang'value`).
///
/// # References
/// - RFC 2231 Section 4
fn extract_rfc2231_param(header_value: &str, param_name: &str) -> Option<String> {
    let lower = header_value.to_ascii_lowercase();
    let pattern = format!("{param_name}*=");
    let mut search_from = 0;

    let idx = loop {
        let rel_idx = lower[search_from..].find(&pattern)?;
        let abs_idx = search_from + rel_idx;
        // Ensure we're at a parameter boundary (same check as extract_param)
        if is_param_boundary(&lower, abs_idx) {
            // Skip matches inside a quoted-string (RFC 5322 Section 3.2.4)
            if is_inside_quotes(&lower, abs_idx) {
                search_from = abs_idx + pattern.len();
                continue;
            }
            break abs_idx;
        }
        search_from = abs_idx + pattern.len();
    };

    let rest = &header_value[idx + pattern.len()..];
    let end = rest.find(';').unwrap_or(rest.len());
    let value = rest[..end].trim();

    // Format: charset'language'percent-encoded-value
    let mut parts_iter = value.splitn(3, '\'');
    let charset = parts_iter.next()?;
    let _language = parts_iter.next()?; // Ignored
    let encoded = parts_iter.next()?;

    let decoded_bytes = percent_decode(encoded);
    Some(decode_charset(charset, &decoded_bytes))
}

/// Reassembles RFC 2231 continuation parameters (`param*0=`, `param*1=`, etc.).
///
/// Sections with a trailing `*` (e.g., `param*0*=`) are charset/percent-encoded.
/// The charset is taken from the first section (`param*0*=charset'lang'value`);
/// subsequent `*` sections are just percent-encoded with the same charset.
/// Sections without `*` are plain quoted or unquoted values.
///
/// # References
/// - RFC 2231 Section 3
fn extract_rfc2231_continuation(header_value: &str, param_name: &str) -> Option<String> {
    let lower = header_value.to_ascii_lowercase();
    let mut sections: Vec<(u32, bool, String)> = Vec::new(); // (index, is_encoded, value)
    let mut charset = String::new();

    for section_idx in 0u32..100 {
        // Try encoded form first: param*N*=
        let encoded_pattern = format!("{param_name}*{section_idx}*=");
        if let Some(val) = find_param_value(&lower, header_value, &encoded_pattern) {
            if section_idx == 0 {
                // First encoded section has charset'language'value
                let mut parts = val.splitn(3, '\'');
                if let (Some(cs), Some(_lang), Some(encoded)) =
                    (parts.next(), parts.next(), parts.next())
                {
                    charset = cs.to_string();
                    sections.push((section_idx, true, encoded.to_string()));
                } else {
                    sections.push((section_idx, true, val));
                }
            } else {
                // Subsequent encoded sections are just percent-encoded
                sections.push((section_idx, true, val));
            }
            continue;
        }

        // Try plain form: param*N=
        let plain_pattern = format!("{param_name}*{section_idx}=");
        if let Some(val) = find_param_value(&lower, header_value, &plain_pattern) {
            sections.push((section_idx, false, val));
            continue;
        }

        // No more sections
        break;
    }

    if sections.is_empty() {
        return None;
    }

    // Sort by index (should already be in order, but be safe)
    sections.sort_by_key(|(idx, _, _)| *idx);

    // Concatenate: encoded sections get percent-decoded, plain sections used as-is
    let mut raw_bytes: Vec<u8> = Vec::new();
    for (_, is_encoded, value) in &sections {
        if *is_encoded {
            raw_bytes.extend(percent_decode(value));
        } else {
            raw_bytes.extend(value.as_bytes());
        }
    }

    if charset.is_empty() {
        // RFC 2231 Section 4: when no charset is declared in the first
        // encoded section, the default is the charset of the enclosing
        // entity — US-ASCII per RFC 2045 Section 5.2. We use UTF-8
        // instead as a Postel's law accommodation: US-ASCII is a strict
        // subset of UTF-8, so ASCII-only values decode identically,
        // while non-ASCII bytes (from non-conformant senders) are
        // preserved rather than mis-interpreted through encoding_rs's
        // us-ascii → Windows-1252 mapping.
        charset = "utf-8".to_string();
    }

    Some(decode_charset(&charset, &raw_bytes))
}

/// Finds a parameter value in a header, given a lowercase pattern like `"filename*0="`.
///
/// Checks parameter boundaries and handles both quoted and unquoted values.
/// Quoted-string values are unescaped per RFC 5322 Section 3.2.4.
fn find_param_value(lower: &str, original: &str, pattern: &str) -> Option<String> {
    let mut search_from = 0;
    loop {
        let rel_idx = lower[search_from..].find(pattern)?;
        let abs_idx = search_from + rel_idx;

        // Ensure parameter boundary
        if is_param_boundary(lower, abs_idx) {
            // Skip matches inside a quoted-string (RFC 5322 Section 3.2.4)
            if is_inside_quotes(lower, abs_idx) {
                search_from = abs_idx + pattern.len();
                continue;
            }

            let rest = &original[abs_idx + pattern.len()..];
            return extract_param_value(rest);
        }

        search_from = abs_idx + pattern.len();
    }
}

/// Decodes percent-encoded bytes (RFC 2231 / RFC 3986 Section 2.1).
fn percent_decode(input: &str) -> Vec<u8> {
    let bytes = input.as_bytes();
    let mut result = Vec::with_capacity(bytes.len());
    let mut i = 0;
    while i < bytes.len() {
        if bytes[i] == b'%' && i + 2 < bytes.len() {
            if let Some(val) = decode_hex_pair(bytes[i + 1], bytes[i + 2]) {
                result.push(val);
                i += 3;
                continue;
            }
        }
        result.push(bytes[i]);
        i += 1;
    }
    result
}

// ---------------------------------------------------------------------------
// Utility functions
// ---------------------------------------------------------------------------

/// Strips parenthesized comments from a string.
///
/// RFC 5322 Section 3.2.2 defines comments as text enclosed in parentheses,
/// which may be nested. A backslash escapes the next character inside a comment.
///
/// # References
/// - RFC 5322 Section 3.2.2 (comment syntax)
/// - RFC 5322 Section 4.3 (CFWS in obsolete date syntax)
fn strip_comments(input: &str) -> String {
    let mut result = String::with_capacity(input.len());
    let mut depth: u32 = 0;
    let mut escaped = false;
    for c in input.chars() {
        if escaped {
            escaped = false;
            if depth == 0 {
                result.push(c);
            }
            continue;
        }
        match c {
            '\\' => {
                escaped = true;
                if depth == 0 {
                    result.push(c);
                }
            }
            '(' => depth = depth.saturating_add(1),
            ')' if depth > 0 => depth = depth.saturating_sub(1),
            _ if depth == 0 => result.push(c),
            _ => {}
        }
    }
    result
}

/// Finds the first occurrence of `needle` in `haystack`.
fn find_subsequence(haystack: &[u8], needle: &[u8]) -> Option<usize> {
    haystack.windows(needle.len()).position(|w| w == needle)
}

/// Decodes a pair of hex ASCII characters into a byte value.
fn decode_hex_pair(high: u8, low: u8) -> Option<u8> {
    let h = hex_digit(high)?;
    let l = hex_digit(low)?;
    Some(h * 16 + l)
}

/// Converts a single ASCII hex digit to its numeric value.
fn hex_digit(b: u8) -> Option<u8> {
    match b {
        b'0'..=b'9' => Some(b - b'0'),
        b'A'..=b'F' => Some(b - b'A' + 10),
        b'a'..=b'f' => Some(b - b'a' + 10),
        _ => None,
    }
}

/// Strips only the outer pair of quotes from a quoted-string.
///
/// If `input` starts with `"` and ends with `"`, removes those two characters.
/// Otherwise returns the input unchanged. Unlike `trim_matches('"')`, this does
/// not greedily strip multiple consecutive quotes, which is critical when the
/// display name ends with an escaped quote like `"She said \"hello\""`.
///
/// # References
/// - RFC 5322 Section 3.2.4 (quoted-string structure)
fn strip_outer_quotes(input: &str) -> &str {
    if input.len() >= 2 && input.starts_with('"') && input.ends_with('"') {
        &input[1..input.len() - 1]
    } else {
        input
    }
}

/// Unescapes a quoted-string: removes backslash from `\\` → `\` and `\"` → `"`.
///
/// Per RFC 5322 Section 3.2.4, a `quoted-pair` is `"\" (VCHAR / WSP)`.
fn unescape_quoted_string(input: &str) -> String {
    let mut result = String::with_capacity(input.len());
    let mut chars = input.chars();
    while let Some(c) = chars.next() {
        if c == '\\' {
            // Consume the escaped character (RFC 5322 Section 3.2.4)
            if let Some(next) = chars.next() {
                result.push(next);
            } else {
                result.push(c);
            }
        } else {
            result.push(c);
        }
    }
    result
}

// ---------------------------------------------------------------------------
// Tests
// ---------------------------------------------------------------------------

#[cfg(test)]
#[allow(clippy::unwrap_used, clippy::expect_used)]
mod tests {
    use super::*;

    #[test]
    fn parse_simple_text_email() {
        let raw = b"From: sender@example.com\r\n\
                     To: recipient@example.com\r\n\
                     Subject: Test\r\n\
                     Date: Thu, 13 Feb 2025 15:47:33 +0000\r\n\
                     Message-ID: <abc123@example.com>\r\n\
                     Content-Type: text/plain; charset=utf-8\r\n\
                     \r\n\
                     Hello, World!";

        let parsed = parse_email(raw).unwrap();
        assert_eq!(parsed.from.email, "sender@example.com");
        assert_eq!(parsed.to.len(), 1);
        assert_eq!(parsed.to[0].email, "recipient@example.com");
        assert_eq!(parsed.subject.as_deref(), Some("Test"));
        assert_eq!(parsed.message_id.as_deref(), Some("abc123@example.com"));
        assert_eq!(parsed.body_text.as_deref(), Some("Hello, World!"));
        assert!(parsed.body_html.is_none());
        assert!(parsed.attachments.is_empty());
        assert_eq!(parsed.size, raw.len() as u64);
    }

    #[test]
    fn parse_multipart_alternative() {
        let raw = b"From: sender@example.com\r\n\
                     To: recipient@example.com\r\n\
                     Subject: Multi\r\n\
                     Date: Thu, 13 Feb 2025 15:47:33 +0000\r\n\
                     MIME-Version: 1.0\r\n\
                     Content-Type: multipart/alternative; boundary=\"bound42\"\r\n\
                     \r\n\
                     --bound42\r\n\
                     Content-Type: text/plain; charset=utf-8\r\n\
                     \r\n\
                     Plain text body\r\n\
                     --bound42\r\n\
                     Content-Type: text/html; charset=utf-8\r\n\
                     \r\n\
                     <html><body>HTML body</body></html>\r\n\
                     --bound42--";

        let parsed = parse_email(raw).unwrap();
        assert_eq!(parsed.body_text.as_deref(), Some("Plain text body"));
        assert_eq!(
            parsed.body_html.as_deref(),
            Some("<html><body>HTML body</body></html>")
        );
        assert!(parsed.attachments.is_empty());
    }

    #[test]
    fn parse_encoded_words_base64_subject() {
        let raw = b"From: sender@example.com\r\n\
                     Subject: =?UTF-8?B?SGVsbG8gV29ybGQ=?=\r\n\
                     Date: Thu, 13 Feb 2025 15:47:33 +0000\r\n\
                     \r\n\
                     body";

        let parsed = parse_email(raw).unwrap();
        assert_eq!(parsed.subject.as_deref(), Some("Hello World"));
    }

    #[test]
    fn parse_encoded_words_q_subject() {
        let raw = b"From: sender@example.com\r\n\
                     Subject: =?UTF-8?Q?Hello_World?=\r\n\
                     Date: Thu, 13 Feb 2025 15:47:33 +0000\r\n\
                     \r\n\
                     body";

        let parsed = parse_email(raw).unwrap();
        assert_eq!(parsed.subject.as_deref(), Some("Hello World"));
    }

    #[test]
    fn parse_encoded_words_in_display_name() {
        let raw = b"From: =?UTF-8?B?Sm9obiBEb2U=?= <john@example.com>\r\n\
                     Subject: Test\r\n\
                     Date: Thu, 13 Feb 2025 15:47:33 +0000\r\n\
                     \r\n";

        let parsed = parse_email(raw).unwrap();
        assert_eq!(parsed.from.name.as_deref(), Some("John Doe"));
        assert_eq!(parsed.from.email, "john@example.com");
    }

    #[test]
    fn parse_non_utf8_charset() {
        // ISO-8859-1 encoded subject: "Héllo"
        let raw = b"From: sender@example.com\r\n\
                     Subject: =?ISO-8859-1?Q?H=E9llo?=\r\n\
                     Date: Thu, 13 Feb 2025 15:47:33 +0000\r\n\
                     \r\n";

        let parsed = parse_email(raw).unwrap();
        assert_eq!(parsed.subject.as_deref(), Some("Héllo"));
    }

    #[test]
    fn parse_message_id_strips_brackets() {
        let raw = b"From: a@b.com\r\n\
                     Message-ID: <unique-id@host.com>\r\n\
                     Date: Thu, 13 Feb 2025 15:47:33 +0000\r\n\
                     \r\n";

        let parsed = parse_email(raw).unwrap();
        assert_eq!(parsed.message_id.as_deref(), Some("unique-id@host.com"));
    }

    #[test]
    fn parse_in_reply_to_first_only() {
        let raw = b"From: a@b.com\r\n\
                     In-Reply-To: <first@host> <second@host>\r\n\
                     Date: Thu, 13 Feb 2025 15:47:33 +0000\r\n\
                     \r\n";

        let parsed = parse_email(raw).unwrap();
        assert_eq!(parsed.in_reply_to.as_deref(), Some("first@host"));
    }

    #[test]
    fn parse_references_all_ids() {
        let raw = b"From: a@b.com\r\n\
                     References: <ref1@host> <ref2@host> <ref3@host>\r\n\
                     Date: Thu, 13 Feb 2025 15:47:33 +0000\r\n\
                     \r\n";

        let parsed = parse_email(raw).unwrap();
        assert_eq!(
            parsed.references.as_deref(),
            Some("ref1@host ref2@host ref3@host")
        );
    }

    #[test]
    fn parse_date_with_numeric_timezone() {
        let raw = b"From: a@b.com\r\n\
                     Date: Thu, 13 Feb 2025 15:47:33 +0530\r\n\
                     \r\n";

        let parsed = parse_email(raw).unwrap();
        let date = parsed.date.unwrap();
        assert_eq!(date.year, 2025);
        assert_eq!(date.month, 2);
        assert_eq!(date.day, 13);
        assert_eq!(date.hour, 15);
        assert_eq!(date.minute, 47);
        assert_eq!(date.second, 33);
        assert_eq!(date.tz_offset_minutes, 330);
    }

    #[test]
    fn parse_date_named_timezone() {
        let raw = b"From: a@b.com\r\n\
                     Date: Thu, 13 Feb 2025 10:30:00 EST\r\n\
                     \r\n";

        let parsed = parse_email(raw).unwrap();
        let date = parsed.date.unwrap();
        assert_eq!(date.tz_offset_minutes, -300);
    }

    #[test]
    fn parse_address_with_display_name() {
        let raw = b"From: \"John Doe\" <john@example.com>\r\n\
                     Date: Thu, 13 Feb 2025 15:47:33 +0000\r\n\
                     \r\n";

        let parsed = parse_email(raw).unwrap();
        assert_eq!(parsed.from.name.as_deref(), Some("John Doe"));
        assert_eq!(parsed.from.email, "john@example.com");
    }

    #[test]
    fn parse_multiple_recipients() {
        let raw = b"From: a@b.com\r\n\
                     To: one@x.com, \"Two\" <two@x.com>, three@x.com\r\n\
                     Cc: cc1@x.com, cc2@x.com\r\n\
                     Date: Thu, 13 Feb 2025 15:47:33 +0000\r\n\
                     \r\n";

        let parsed = parse_email(raw).unwrap();
        assert_eq!(parsed.to.len(), 3);
        assert_eq!(parsed.to[1].name.as_deref(), Some("Two"));
        assert_eq!(parsed.cc.len(), 2);
    }

    #[test]
    fn parse_multipart_with_attachment() {
        let raw = b"From: a@b.com\r\n\
                     Date: Thu, 13 Feb 2025 15:47:33 +0000\r\n\
                     MIME-Version: 1.0\r\n\
                     Content-Type: multipart/mixed; boundary=\"mixbound\"\r\n\
                     \r\n\
                     --mixbound\r\n\
                     Content-Type: text/plain\r\n\
                     \r\n\
                     Message body\r\n\
                     --mixbound\r\n\
                     Content-Type: application/pdf\r\n\
                     Content-Disposition: attachment; filename=\"doc.pdf\"\r\n\
                     \r\n\
                     PDF_CONTENT_HERE\r\n\
                     --mixbound--";

        let parsed = parse_email(raw).unwrap();
        assert_eq!(parsed.body_text.as_deref(), Some("Message body"));
        assert_eq!(parsed.attachments.len(), 1);
        assert_eq!(parsed.attachments[0].filename.as_deref(), Some("doc.pdf"));
        assert_eq!(parsed.attachments[0].content_type, "application/pdf");
        assert!(!parsed.attachments[0].is_inline);
        assert_eq!(parsed.attachments[0].section.as_deref(), Some("2"));
    }

    #[test]
    fn parse_inline_attachment() {
        let raw = b"From: a@b.com\r\n\
                     Date: Thu, 13 Feb 2025 15:47:33 +0000\r\n\
                     Content-Type: multipart/mixed; boundary=\"bound\"\r\n\
                     \r\n\
                     --bound\r\n\
                     Content-Type: text/plain\r\n\
                     \r\n\
                     Body\r\n\
                     --bound\r\n\
                     Content-Type: image/png\r\n\
                     Content-Disposition: inline\r\n\
                     Content-ID: <img001>\r\n\
                     \r\n\
                     PNG_DATA\r\n\
                     --bound--";

        let parsed = parse_email(raw).unwrap();
        assert_eq!(parsed.attachments.len(), 1);
        assert!(parsed.attachments[0].is_inline);
        assert_eq!(parsed.attachments[0].content_id.as_deref(), Some("img001"));
    }

    #[test]
    fn parse_headers_only_no_body() {
        let raw = b"From: a@b.com\r\n\
                     Subject: Headers only\r\n\
                     Date: Thu, 13 Feb 2025 15:47:33 +0000\r\n";

        let parsed = parse_email(raw).unwrap();
        assert_eq!(parsed.subject.as_deref(), Some("Headers only"));
        assert!(parsed.body_text.is_none());
        assert!(parsed.body_html.is_none());
    }

    #[test]
    fn parse_empty_input() {
        let result = parse_email(b"");
        assert!(matches!(result, Err(Error::EmptyInput)));
    }

    #[test]
    fn parse_missing_from() {
        let raw = b"Subject: No from\r\n\
                     Date: Thu, 13 Feb 2025 15:47:33 +0000\r\n\
                     \r\n";

        let result = parse_email(raw);
        assert!(matches!(result, Err(Error::MissingFrom)));
    }

    #[test]
    fn parse_quoted_printable_body() {
        let raw = b"From: a@b.com\r\n\
                     Date: Thu, 13 Feb 2025 15:47:33 +0000\r\n\
                     Content-Type: text/plain; charset=utf-8\r\n\
                     Content-Transfer-Encoding: quoted-printable\r\n\
                     \r\n\
                     Hello=20World=0D=0ASoft=\r\n break";

        let parsed = parse_email(raw).unwrap();
        assert_eq!(
            parsed.body_text.as_deref(),
            Some("Hello World\r\nSoft break")
        );
    }

    #[test]
    fn parse_base64_body() {
        let raw = b"From: a@b.com\r\n\
                     Date: Thu, 13 Feb 2025 15:47:33 +0000\r\n\
                     Content-Type: text/plain; charset=utf-8\r\n\
                     Content-Transfer-Encoding: base64\r\n\
                     \r\n\
                     SGVsbG8gV29ybGQ=\r\n";

        let parsed = parse_email(raw).unwrap();
        assert_eq!(parsed.body_text.as_deref(), Some("Hello World"));
    }

    #[test]
    fn parse_nested_multipart_section_numbers() {
        let raw = b"From: a@b.com\r\n\
                     Date: Thu, 13 Feb 2025 15:47:33 +0000\r\n\
                     Content-Type: multipart/mixed; boundary=\"outer\"\r\n\
                     \r\n\
                     --outer\r\n\
                     Content-Type: multipart/alternative; boundary=\"inner\"\r\n\
                     \r\n\
                     --inner\r\n\
                     Content-Type: text/plain\r\n\
                     \r\n\
                     Plain\r\n\
                     --inner\r\n\
                     Content-Type: text/html\r\n\
                     \r\n\
                     <b>HTML</b>\r\n\
                     --inner--\r\n\
                     --outer\r\n\
                     Content-Type: application/pdf\r\n\
                     Content-Disposition: attachment; filename=\"f.pdf\"\r\n\
                     \r\n\
                     DATA\r\n\
                     --outer--";

        let parsed = parse_email(raw).unwrap();
        assert_eq!(parsed.body_text.as_deref(), Some("Plain"));
        assert_eq!(parsed.body_html.as_deref(), Some("<b>HTML</b>"));
        assert_eq!(parsed.attachments.len(), 1);
        // Attachment is part 2 of the outer multipart
        assert_eq!(parsed.attachments[0].section.as_deref(), Some("2"));
    }

    #[test]
    fn parse_rfc2231_filename() {
        let raw = b"From: a@b.com\r\n\
                     Date: Thu, 13 Feb 2025 15:47:33 +0000\r\n\
                     Content-Type: multipart/mixed; boundary=\"b\"\r\n\
                     \r\n\
                     --b\r\n\
                     Content-Type: text/plain\r\n\
                     \r\n\
                     Body\r\n\
                     --b\r\n\
                     Content-Type: application/pdf\r\n\
                     Content-Disposition: attachment; filename*=UTF-8''r%C3%A9sum%C3%A9.pdf\r\n\
                     \r\n\
                     DATA\r\n\
                     --b--";

        let parsed = parse_email(raw).unwrap();
        assert_eq!(parsed.attachments.len(), 1);
        assert_eq!(
            parsed.attachments[0].filename.as_deref(),
            Some("résumé.pdf")
        );
    }

    #[test]
    fn parse_raw_headers_preserved() {
        let raw = b"From: a@b.com\r\n\
                     Subject: Test\r\n\
                     Date: Thu, 13 Feb 2025 15:47:33 +0000\r\n\
                     \r\n\
                     Body";

        let parsed = parse_email(raw).unwrap();
        assert!(parsed.raw_headers.contains("From: a@b.com"));
        assert!(parsed.raw_headers.contains("Subject: Test"));
    }

    #[test]
    fn parse_lf_only_line_endings() {
        let raw = b"From: a@b.com\n\
                     Subject: LF\n\
                     Date: Thu, 13 Feb 2025 15:47:33 +0000\n\
                     \n\
                     Body with LF";

        let parsed = parse_email(raw).unwrap();
        assert_eq!(parsed.subject.as_deref(), Some("LF"));
        assert_eq!(parsed.body_text.as_deref(), Some("Body with LF"));
    }

    #[test]
    fn parse_header_continuation_lines() {
        // Continuation line starts with a space (RFC 5322 Section 2.2.3)
        // Can't use `\` line continuation as it strips leading whitespace.
        let raw = b"From: a@b.com\r\nSubject: This is a very long\r\n subject line that wraps\r\nDate: Thu, 13 Feb 2025 15:47:33 +0000\r\n\r\n";

        let parsed = parse_email(raw).unwrap();
        assert_eq!(
            parsed.subject.as_deref(),
            Some("This is a very long subject line that wraps")
        );
    }

    #[test]
    fn parse_garbage_input_best_effort() {
        // Binary garbage — no valid From header → error
        let result = parse_email(b"\x00\x01\x02\x03\xff\xfe");
        assert!(result.is_err());
    }

    #[test]
    fn parse_truncated_multipart() {
        // Multipart with missing closing boundary
        let raw = b"From: a@b.com\r\n\
                     Date: Thu, 13 Feb 2025 15:47:33 +0000\r\n\
                     Content-Type: multipart/mixed; boundary=\"trunc\"\r\n\
                     \r\n\
                     --trunc\r\n\
                     Content-Type: text/plain\r\n\
                     \r\n\
                     Some text here";

        let parsed = parse_email(raw).unwrap();
        assert_eq!(parsed.body_text.as_deref(), Some("Some text here"));
    }

    #[test]
    fn decode_adjacent_encoded_words() {
        // RFC 2047 Section 6.2: whitespace between adjacent encoded words is removed
        let input = "=?UTF-8?B?SGVs?= =?UTF-8?B?bG8=?=";
        let decoded = decode_encoded_words(input);
        assert_eq!(decoded, "Hello");
    }

    #[test]
    fn decode_iso8859_encoded_word() {
        // =?ISO-8859-1?Q?caf=E9?= → "café"
        let input = "=?ISO-8859-1?Q?caf=E9?=";
        let decoded = decode_encoded_words(input);
        assert_eq!(decoded, "café");
    }

    #[test]
    fn parse_date_without_seconds() {
        let dt = parse_rfc5322_date("Thu, 13 Feb 2025 15:47 +0000").unwrap();
        assert_eq!(dt.hour, 15);
        assert_eq!(dt.minute, 47);
        assert_eq!(dt.second, 0);
    }

    #[test]
    fn parse_two_digit_year() {
        let dt = parse_rfc5322_date("13 Feb 99 12:00:00 +0000").unwrap();
        assert_eq!(dt.year, 1999);

        let dt = parse_rfc5322_date("13 Feb 25 12:00:00 +0000").unwrap();
        assert_eq!(dt.year, 2025);
    }

    #[test]
    fn parse_three_digit_year_rfc5322_section_4_3() {
        // RFC 5322 Section 4.3: any 3-digit year should have 1900 added.
        let dt = parse_rfc5322_date("13 Feb 107 12:00:00 +0000").unwrap();
        assert_eq!(
            dt.year, 2007,
            "3-digit year 107 must map to 2007 per RFC 5322 Section 4.3"
        );

        let dt = parse_rfc5322_date("13 Feb 100 12:00:00 +0000").unwrap();
        assert_eq!(
            dt.year, 2000,
            "3-digit year 100 must map to 2000 per RFC 5322 Section 4.3"
        );

        let dt = parse_rfc5322_date("13 Feb 999 12:00:00 +0000").unwrap();
        assert_eq!(
            dt.year, 2899,
            "3-digit year 999 must map to 2899 per RFC 5322 Section 4.3"
        );
    }

    #[test]
    fn parse_two_digit_year_rfc5322_section_4_3_cutoff() {
        // RFC 5322 Section 4.3: 2-digit years 00-49 → +2000, 50-99 → +1900.
        // The cutoff is 50, not 70.

        // Year 50 should map to 1950 (not 2050)
        let dt = parse_rfc5322_date("13 Feb 50 12:00:00 +0000").unwrap();
        assert_eq!(
            dt.year, 1950,
            "2-digit year 50 must map to 1950 per RFC 5322 Section 4.3"
        );

        // Year 69 should map to 1969 (not 2069)
        let dt = parse_rfc5322_date("13 Feb 69 12:00:00 +0000").unwrap();
        assert_eq!(
            dt.year, 1969,
            "2-digit year 69 must map to 1969 per RFC 5322 Section 4.3"
        );

        // Year 49 should map to 2049
        let dt = parse_rfc5322_date("13 Feb 49 12:00:00 +0000").unwrap();
        assert_eq!(
            dt.year, 2049,
            "2-digit year 49 must map to 2049 per RFC 5322 Section 4.3"
        );
    }

    #[test]
    fn parse_non_text_part_is_attachment() {
        let raw = b"From: a@b.com\r\n\
                     Date: Thu, 13 Feb 2025 15:47:33 +0000\r\n\
                     Content-Type: multipart/mixed; boundary=\"b\"\r\n\
                     \r\n\
                     --b\r\n\
                     Content-Type: text/plain\r\n\
                     \r\n\
                     Text\r\n\
                     --b\r\n\
                     Content-Type: image/jpeg\r\n\
                     \r\n\
                     JPEG_DATA\r\n\
                     --b--";

        let parsed = parse_email(raw).unwrap();
        // image/jpeg without explicit disposition should be treated as attachment
        assert_eq!(parsed.attachments.len(), 1);
        assert_eq!(parsed.attachments[0].content_type, "image/jpeg");
    }

    #[test]
    fn parse_windows1252_body() {
        // Windows-1252 body with smart quotes
        let raw = b"From: a@b.com\r\n\
                     Date: Thu, 13 Feb 2025 15:47:33 +0000\r\n\
                     Content-Type: text/plain; charset=windows-1252\r\n\
                     \r\n\
                     \x93Hello\x94"; // Smart double quotes in Windows-1252

        let parsed = parse_email(raw).unwrap();
        let text = parsed.body_text.unwrap();
        assert!(text.contains("Hello"));
        // Smart quotes should be converted to Unicode
        assert!(text.contains('\u{201c}') || text.contains('\u{201d}'));
    }

    #[test]
    fn parse_html_only_body() {
        let raw = b"From: a@b.com\r\n\
                     Date: Thu, 13 Feb 2025 15:47:33 +0000\r\n\
                     Content-Type: text/html; charset=utf-8\r\n\
                     \r\n\
                     <html><body>Hello</body></html>";

        let parsed = parse_email(raw).unwrap();
        assert!(parsed.body_text.is_none());
        assert_eq!(
            parsed.body_html.as_deref(),
            Some("<html><body>Hello</body></html>")
        );
    }

    #[test]
    fn parse_bcc_addresses() {
        let raw = b"From: a@b.com\r\n\
                     To: to@x.com\r\n\
                     Bcc: hidden@x.com, secret@x.com\r\n\
                     Date: Thu, 13 Feb 2025 15:47:33 +0000\r\n\
                     \r\n";

        let parsed = parse_email(raw).unwrap();
        assert_eq!(parsed.bcc.len(), 2);
        assert_eq!(parsed.bcc[0].email, "hidden@x.com");
    }

    #[test]
    fn mime_depth_limit() {
        // Construct a deeply nested multipart that exceeds MAX_MIME_DEPTH
        // Just verify it doesn't stack overflow
        let mut msg = b"From: a@b.com\r\nDate: Thu, 13 Feb 2025 15:47:33 +0000\r\n\
                        Content-Type: multipart/mixed; boundary=\"b0\"\r\n\r\n"
            .to_vec();

        for i in 0..70 {
            msg.extend_from_slice(
                format!(
                    "--b{i}\r\nContent-Type: multipart/mixed; boundary=\"b{}\"\r\n\r\n",
                    i + 1
                )
                .as_bytes(),
            );
        }
        msg.extend_from_slice(b"--b70\r\nContent-Type: text/plain\r\n\r\nDeep\r\n--b70--\r\n");

        let parsed = parse_email(&msg).unwrap();
        // Should not panic or stack overflow — may not find the body due to depth limit
        assert!(parsed.body_text.is_none() || parsed.body_text.is_some());
    }

    #[test]
    fn parse_reply_to() {
        let raw = b"From: a@b.com\r\n\
                     Reply-To: noreply@example.com, support@example.com\r\n\
                     Date: Thu, 13 Feb 2025 15:47:33 +0000\r\n\
                     \r\n";

        let parsed = parse_email(raw).unwrap();
        assert_eq!(parsed.reply_to.len(), 2);
        assert_eq!(parsed.reply_to[0].email, "noreply@example.com");
        assert_eq!(parsed.reply_to[1].email, "support@example.com");
    }

    #[test]
    fn parse_gb2312_encoded_word() {
        // GB2312 encoded word: "你好" (nǐ hǎo) in base64
        // "你好" in GB2312 is: 0xC4, 0xE3, 0xBA, 0xC3
        let raw = b"From: sender@example.com\r\n\
                     Subject: =?GB2312?B?xOO6ww==?=\r\n\
                     Date: Thu, 13 Feb 2025 15:47:33 +0000\r\n\
                     \r\n";

        let parsed = parse_email(raw).unwrap();
        assert_eq!(parsed.subject.as_deref(), Some("你好"));
    }

    #[test]
    fn parse_content_id_strips_brackets() {
        let raw = b"From: a@b.com\r\n\
                     Date: Thu, 13 Feb 2025 15:47:33 +0000\r\n\
                     Content-Type: multipart/mixed; boundary=\"b\"\r\n\
                     \r\n\
                     --b\r\n\
                     Content-Type: text/plain\r\n\
                     \r\n\
                     Body\r\n\
                     --b\r\n\
                     Content-Type: image/png\r\n\
                     Content-ID: <cid:image001@01D00000.00000000>\r\n\
                     \r\n\
                     PNG\r\n\
                     --b--";

        let parsed = parse_email(raw).unwrap();
        assert_eq!(
            parsed.attachments[0].content_id.as_deref(),
            Some("cid:image001@01D00000.00000000")
        );
    }

    #[test]
    fn parse_attachment_without_filename() {
        // Attachment with Content-Disposition but no filename parameter
        let raw = b"From: a@b.com\r\n\
                     Date: Thu, 13 Feb 2025 15:47:33 +0000\r\n\
                     Content-Type: multipart/mixed; boundary=\"b\"\r\n\
                     \r\n\
                     --b\r\n\
                     Content-Type: text/plain\r\n\
                     \r\n\
                     Body\r\n\
                     --b\r\n\
                     Content-Type: application/octet-stream\r\n\
                     Content-Disposition: attachment\r\n\
                     \r\n\
                     BINARY\r\n\
                     --b--";

        let parsed = parse_email(raw).unwrap();
        assert_eq!(parsed.attachments.len(), 1);
        assert!(parsed.attachments[0].filename.is_none());
        assert_eq!(
            parsed.attachments[0].content_type,
            "application/octet-stream"
        );
        assert!(!parsed.attachments[0].is_inline);
    }

    #[test]
    fn parse_content_type_without_charset_defaults() {
        // No charset parameter — should default to us-ascii/utf-8 handling
        let raw = b"From: a@b.com\r\n\
                     Date: Thu, 13 Feb 2025 15:47:33 +0000\r\n\
                     Content-Type: text/plain\r\n\
                     \r\n\
                     Hello ASCII";

        let parsed = parse_email(raw).unwrap();
        assert_eq!(parsed.body_text.as_deref(), Some("Hello ASCII"));
    }

    /// A MIME part with no Content-Type header inside a multipart
    /// message must default to "text/plain; charset=us-ascii" per RFC 2045
    /// Section 5.2.
    #[test]
    fn parse_mime_part_no_content_type_defaults_to_us_ascii() {
        // Part has Content-Transfer-Encoding but no Content-Type.
        // RFC 2045 Section 5.2: default is "text/plain; charset=us-ascii".
        let raw = b"From: a@b.com\r\n\
                     Date: Thu, 13 Feb 2025 15:47:33 +0000\r\n\
                     Content-Type: multipart/mixed; boundary=\"b\"\r\n\
                     \r\n\
                     --b\r\n\
                     Content-Transfer-Encoding: 7bit\r\n\
                     \r\n\
                     Hello ASCII\r\n\
                     --b--";

        let parsed = parse_email(raw).unwrap();
        assert_eq!(parsed.body_text.as_deref(), Some("Hello ASCII"));
    }

    /// A MIME part with zero headers (only the blank line
    /// separator, no Content-Type or other headers) must still be parsed.
    /// This is a valid RFC 2046 construct — the blank line after the
    /// boundary delimiter starts the body when there are no part headers.
    #[test]
    fn parse_mime_part_no_headers_at_all() {
        let raw = b"From: a@b.com\r\n\
                     Date: Thu, 13 Feb 2025 15:47:33 +0000\r\n\
                     Content-Type: multipart/mixed; boundary=\"b\"\r\n\
                     \r\n\
                     --b\r\n\
                     \r\n\
                     Headerless body\r\n\
                     --b--";

        let parsed = parse_email(raw).unwrap();
        // Part with no headers defaults to text/plain per RFC 2045 Section 5.2.
        assert_eq!(parsed.body_text.as_deref(), Some("Headerless body"));
    }

    #[test]
    fn parse_multipart_only_attachments() {
        // No text/plain or text/html parts — only attachments
        let raw = b"From: a@b.com\r\n\
                     Date: Thu, 13 Feb 2025 15:47:33 +0000\r\n\
                     Content-Type: multipart/mixed; boundary=\"b\"\r\n\
                     \r\n\
                     --b\r\n\
                     Content-Type: application/pdf\r\n\
                     Content-Disposition: attachment; filename=\"a.pdf\"\r\n\
                     \r\n\
                     PDF1\r\n\
                     --b\r\n\
                     Content-Type: image/png\r\n\
                     Content-Disposition: attachment; filename=\"b.png\"\r\n\
                     \r\n\
                     PNG2\r\n\
                     --b--";

        let parsed = parse_email(raw).unwrap();
        assert!(parsed.body_text.is_none());
        assert!(parsed.body_html.is_none());
        assert_eq!(parsed.attachments.len(), 2);
        assert_eq!(parsed.attachments[0].section.as_deref(), Some("1"));
        assert_eq!(parsed.attachments[1].section.as_deref(), Some("2"));
    }

    #[test]
    fn parse_unknown_charset_body_fallback() {
        // Unknown charset should fall back to UTF-8 lossy conversion
        let raw = b"From: a@b.com\r\n\
                     Date: Thu, 13 Feb 2025 15:47:33 +0000\r\n\
                     Content-Type: text/plain; charset=x-unknown-fake\r\n\
                     \r\n\
                     Plain text in unknown charset";

        let parsed = parse_email(raw).unwrap();
        // encoding_rs falls back to UTF-8 for unknown charsets
        assert!(parsed.body_text.is_some());
        assert!(parsed.body_text.unwrap().contains("Plain text"));
    }

    #[test]
    fn parse_content_id_without_disposition_is_inline() {
        // Part with Content-ID but no Content-Disposition — should be inline
        let raw = b"From: a@b.com\r\n\
                     Date: Thu, 13 Feb 2025 15:47:33 +0000\r\n\
                     Content-Type: multipart/mixed; boundary=\"b\"\r\n\
                     \r\n\
                     --b\r\n\
                     Content-Type: text/plain\r\n\
                     \r\n\
                     Body\r\n\
                     --b\r\n\
                     Content-Type: image/gif\r\n\
                     Content-ID: <img42>\r\n\
                     \r\n\
                     GIF89a\r\n\
                     --b--";

        let parsed = parse_email(raw).unwrap();
        assert_eq!(parsed.attachments.len(), 1);
        assert!(parsed.attachments[0].is_inline);
        assert_eq!(parsed.attachments[0].content_id.as_deref(), Some("img42"));
    }

    #[test]
    fn parse_overlong_subject() {
        // Very long subject line — should not crash or truncate
        let long_subject = "A".repeat(10_000);
        let raw = format!(
            "From: a@b.com\r\n\
             Subject: {long_subject}\r\n\
             Date: Thu, 13 Feb 2025 15:47:33 +0000\r\n\
             \r\n"
        );

        let parsed = parse_email(raw.as_bytes()).unwrap();
        assert_eq!(parsed.subject.as_deref(), Some(long_subject.as_str()));
    }

    #[test]
    fn parse_multiple_from_takes_first() {
        // Multiple addresses in From — take the first one
        let raw = b"From: first@example.com, second@example.com\r\n\
                     Date: Thu, 13 Feb 2025 15:47:33 +0000\r\n\
                     \r\n";

        let parsed = parse_email(raw).unwrap();
        assert_eq!(parsed.from.email, "first@example.com");
    }

    #[test]
    fn parse_multipart_no_boundary_param() {
        // multipart/mixed but no boundary parameter — fallback to simple body
        let raw = b"From: a@b.com\r\n\
                     Date: Thu, 13 Feb 2025 15:47:33 +0000\r\n\
                     Content-Type: multipart/mixed\r\n\
                     \r\n\
                     Some text content";

        let parsed = parse_email(raw).unwrap();
        // Should not panic; falls back to treating body as simple text
        assert!(parsed.body_text.is_some());
    }

    #[test]
    fn parse_empty_body_after_headers() {
        // Headers + blank line + no body content
        let raw = b"From: a@b.com\r\n\
                     Subject: Empty body\r\n\
                     Date: Thu, 13 Feb 2025 15:47:33 +0000\r\n\
                     \r\n";

        let parsed = parse_email(raw).unwrap();
        assert_eq!(parsed.subject.as_deref(), Some("Empty body"));
        assert!(parsed.body_text.is_none());
    }

    #[test]
    fn parse_mixed_charset_encoded_words() {
        // Adjacent encoded words with different charsets
        let raw = b"From: a@b.com\r\n\
                     Subject: =?UTF-8?B?SGVsbG8=?= =?ISO-8859-1?Q?_caf=E9?=\r\n\
                     Date: Thu, 13 Feb 2025 15:47:33 +0000\r\n\
                     \r\n";

        let parsed = parse_email(raw).unwrap();
        assert_eq!(parsed.subject.as_deref(), Some("Hello caf\u{e9}"));
    }

    #[test]
    fn parse_no_date_header() {
        // Message without Date header — date should be None
        let raw = b"From: a@b.com\r\n\
                     Subject: No date\r\n\
                     \r\n\
                     Body";

        let parsed = parse_email(raw).unwrap();
        assert!(parsed.date.is_none());
        assert_eq!(parsed.subject.as_deref(), Some("No date"));
    }

    #[test]
    fn parse_explicit_attachment_text_plain() {
        // text/plain with Content-Disposition: attachment should be attachment, not body
        let raw = b"From: a@b.com\r\n\
                     Date: Thu, 13 Feb 2025 15:47:33 +0000\r\n\
                     Content-Type: multipart/mixed; boundary=\"b\"\r\n\
                     \r\n\
                     --b\r\n\
                     Content-Type: text/plain\r\n\
                     \r\n\
                     Body text\r\n\
                     --b\r\n\
                     Content-Type: text/plain\r\n\
                     Content-Disposition: attachment; filename=\"log.txt\"\r\n\
                     \r\n\
                     Log file content\r\n\
                     --b--";

        let parsed = parse_email(raw).unwrap();
        assert_eq!(parsed.body_text.as_deref(), Some("Body text"));
        assert_eq!(parsed.attachments.len(), 1);
        assert_eq!(parsed.attachments[0].filename.as_deref(), Some("log.txt"));
        assert_eq!(parsed.attachments[0].content_type, "text/plain");
    }

    #[test]
    fn parse_date_negative_timezone() {
        let raw = b"From: a@b.com\r\n\
                     Date: Fri, 14 Feb 2025 09:15:00 -0800\r\n\
                     \r\n";

        let parsed = parse_email(raw).unwrap();
        let date = parsed.date.unwrap();
        assert_eq!(date.tz_offset_minutes, -480);
    }

    #[test]
    fn parse_size_equals_input_length() {
        let raw = b"From: a@b.com\r\nDate: Thu, 13 Feb 2025 15:47:33 +0000\r\n\r\nBody";
        let parsed = parse_email(raw).unwrap();
        assert_eq!(parsed.size, raw.len() as u64);
    }

    #[test]
    fn parse_binary_garbage_returns_error() {
        // Pure binary garbage — no From header → MissingFrom
        let garbage: Vec<u8> = (0..=255_u8).collect();
        let result = parse_email(&garbage);
        assert!(result.is_err());
    }

    #[test]
    fn parse_folded_encoded_word_subject() {
        // Subject with encoded word that spans a folded line
        let raw = b"From: a@b.com\r\nSubject: =?UTF-8?B?SGVsbG8=?=\r\n =?UTF-8?B?V29ybGQ=?=\r\nDate: Thu, 13 Feb 2025 15:47:33 +0000\r\n\r\n";

        let parsed = parse_email(raw).unwrap();
        assert_eq!(parsed.subject.as_deref(), Some("HelloWorld"));
    }

    // -----------------------------------------------------------------------
    // Additional edge case tests
    // -----------------------------------------------------------------------

    #[test]
    fn parse_encoded_word_lowercase_encoding() {
        // RFC 2047: encoding indicator is case-insensitive
        let raw = b"From: sender@example.com\r\n\
                     Subject: =?utf-8?b?SGVsbG8=?= =?utf-8?q?_World?=\r\n\
                     Date: Thu, 13 Feb 2025 15:47:33 +0000\r\n\
                     \r\n";

        let parsed = parse_email(raw).unwrap();
        assert_eq!(parsed.subject.as_deref(), Some("Hello World"));
    }

    #[test]
    fn parse_malformed_encoded_word_passthrough() {
        // Incomplete encoded word should be passed through as literal text
        let raw = b"From: a@b.com\r\n\
                     Subject: =?UTF-8?B?broken\r\n\
                     Date: Thu, 13 Feb 2025 15:47:33 +0000\r\n\
                     \r\n";

        let parsed = parse_email(raw).unwrap();
        // Should not crash; subject contains the literal malformed encoded word
        assert!(parsed.subject.is_some());
        assert!(parsed.subject.unwrap().contains("=?"));
    }

    #[test]
    fn parse_encoded_word_unknown_encoding_type() {
        // Unknown encoding type (not B or Q) — should pass through
        let raw = b"From: a@b.com\r\n\
                     Subject: =?UTF-8?X?data?=\r\n\
                     Date: Thu, 13 Feb 2025 15:47:33 +0000\r\n\
                     \r\n";

        let parsed = parse_email(raw).unwrap();
        assert!(parsed.subject.is_some());
        // Unknown encoding passed through as literal
        assert!(parsed.subject.unwrap().contains("=?"));
    }

    #[test]
    fn parse_utf8_directly_in_headers_rfc6532() {
        // RFC 6532: UTF-8 characters directly in headers (no encoded words)
        let raw = "From: José <jose@example.com>\r\n\
                    Subject: Ñoño café\r\n\
                    Date: Thu, 13 Feb 2025 15:47:33 +0000\r\n\
                    \r\n\
                    Body";

        let parsed = parse_email(raw.as_bytes()).unwrap();
        assert_eq!(parsed.subject.as_deref(), Some("Ñoño café"));
        assert_eq!(parsed.from.name.as_deref(), Some("José"));
        assert_eq!(parsed.from.email, "jose@example.com");
    }

    #[test]
    fn parse_multipart_with_preamble() {
        // RFC 2046 Section 5.1.1: preamble text before the first boundary is ignored
        let raw = b"From: a@b.com\r\n\
                     Date: Thu, 13 Feb 2025 15:47:33 +0000\r\n\
                     Content-Type: multipart/mixed; boundary=\"preamble-test\"\r\n\
                     \r\n\
                     This is the preamble, which should be ignored.\r\n\
                     --preamble-test\r\n\
                     Content-Type: text/plain\r\n\
                     \r\n\
                     Actual body\r\n\
                     --preamble-test--";

        let parsed = parse_email(raw).unwrap();
        assert_eq!(parsed.body_text.as_deref(), Some("Actual body"));
    }

    #[test]
    fn parse_attachment_name_from_content_type() {
        // Filename from Content-Type name= param when Content-Disposition has no filename
        let raw = b"From: a@b.com\r\n\
                     Date: Thu, 13 Feb 2025 15:47:33 +0000\r\n\
                     Content-Type: multipart/mixed; boundary=\"b\"\r\n\
                     \r\n\
                     --b\r\n\
                     Content-Type: text/plain\r\n\
                     \r\n\
                     Body\r\n\
                     --b\r\n\
                     Content-Type: application/pdf; name=\"report.pdf\"\r\n\
                     Content-Disposition: attachment\r\n\
                     \r\n\
                     PDF\r\n\
                     --b--";

        let parsed = parse_email(raw).unwrap();
        assert_eq!(parsed.attachments.len(), 1);
        assert_eq!(
            parsed.attachments[0].filename.as_deref(),
            Some("report.pdf")
        );
    }

    #[test]
    fn parse_qp_soft_break_lf_only() {
        // Quoted-printable soft line break with just LF (not CRLF)
        let raw = b"From: a@b.com\r\n\
                     Date: Thu, 13 Feb 2025 15:47:33 +0000\r\n\
                     Content-Type: text/plain; charset=utf-8\r\n\
                     Content-Transfer-Encoding: quoted-printable\r\n\
                     \r\n\
                     Hello=\nWorld";

        let parsed = parse_email(raw).unwrap();
        assert_eq!(parsed.body_text.as_deref(), Some("HelloWorld"));
    }

    #[test]
    fn parse_subject_mixed_encoded_and_plain() {
        // Subject with plain text, encoded word, and more plain text
        let raw = b"From: a@b.com\r\n\
                     Subject: Re: =?UTF-8?B?SGVsbG8=?= there\r\n\
                     Date: Thu, 13 Feb 2025 15:47:33 +0000\r\n\
                     \r\n";

        let parsed = parse_email(raw).unwrap();
        assert_eq!(parsed.subject.as_deref(), Some("Re: Hello there"));
    }

    #[test]
    fn parse_whitespace_only_body() {
        // Body consisting only of whitespace
        let raw = b"From: a@b.com\r\n\
                     Date: Thu, 13 Feb 2025 15:47:33 +0000\r\n\
                     Content-Type: text/plain; charset=utf-8\r\n\
                     \r\n\
                     \r\n  \r\n";

        let parsed = parse_email(raw).unwrap();
        // Whitespace-only body should still be returned as body_text
        assert!(parsed.body_text.is_some());
    }

    #[test]
    fn parse_date_missing_timezone() {
        // Date without timezone — should default to +0000
        let raw = b"From: a@b.com\r\n\
                     Date: 13 Feb 2025 12:00:00\r\n\
                     \r\n";

        let parsed = parse_email(raw).unwrap();
        let date = parsed.date.unwrap();
        assert_eq!(date.year, 2025);
        assert_eq!(date.tz_offset_minutes, 0);
    }

    #[test]
    fn parse_deeply_nested_section_dot_notation() {
        // Verify section numbers use correct dot notation for nested multipart
        let raw = b"From: a@b.com\r\n\
                     Date: Thu, 13 Feb 2025 15:47:33 +0000\r\n\
                     Content-Type: multipart/mixed; boundary=\"outer\"\r\n\
                     \r\n\
                     --outer\r\n\
                     Content-Type: multipart/related; boundary=\"rel\"\r\n\
                     \r\n\
                     --rel\r\n\
                     Content-Type: text/html\r\n\
                     \r\n\
                     <img src=\"cid:img1\">\r\n\
                     --rel\r\n\
                     Content-Type: image/png\r\n\
                     Content-ID: <img1>\r\n\
                     \r\n\
                     PNG_DATA\r\n\
                     --rel--\r\n\
                     --outer\r\n\
                     Content-Type: application/pdf\r\n\
                     Content-Disposition: attachment; filename=\"doc.pdf\"\r\n\
                     \r\n\
                     PDF\r\n\
                     --outer--";

        let parsed = parse_email(raw).unwrap();
        // HTML is section 1.1 of the outer, inline image is 1.2
        assert!(parsed.body_html.is_some());
        assert!(parsed.body_html.unwrap().contains("cid:img1"));
        // Inline image attachment: section 1.2
        let inline_att = parsed
            .attachments
            .iter()
            .find(|a| a.content_type == "image/png")
            .unwrap();
        assert_eq!(inline_att.section.as_deref(), Some("1.2"));
        assert!(inline_att.is_inline);
        // PDF attachment: section 2
        let pdf_att = parsed
            .attachments
            .iter()
            .find(|a| a.content_type == "application/pdf")
            .unwrap();
        assert_eq!(pdf_att.section.as_deref(), Some("2"));
    }

    #[test]
    fn parse_non_ascii_bytes_in_body() {
        // Raw non-ASCII bytes in body without charset declaration
        let raw = b"From: a@b.com\r\n\
                     Date: Thu, 13 Feb 2025 15:47:33 +0000\r\n\
                     Content-Type: text/plain\r\n\
                     \r\n\
                     Hello \xff\xfe world";

        let parsed = parse_email(raw).unwrap();
        // Should not panic — lossy UTF-8 conversion
        assert!(parsed.body_text.is_some());
    }

    #[test]
    fn parse_base64_body_with_line_breaks() {
        // Base64 body with CRLF line breaks in the middle (RFC 2045 Section 6.8)
        let raw = b"From: a@b.com\r\n\
                     Date: Thu, 13 Feb 2025 15:47:33 +0000\r\n\
                     Content-Type: text/plain; charset=utf-8\r\n\
                     Content-Transfer-Encoding: base64\r\n\
                     \r\n\
                     SGVs\r\nbG8g\r\nV29y\r\nbGQ=";

        let parsed = parse_email(raw).unwrap();
        assert_eq!(parsed.body_text.as_deref(), Some("Hello World"));
    }

    #[test]
    fn parse_date_extra_whitespace() {
        // Date with extra whitespace between fields
        let raw = b"From: a@b.com\r\n\
                     Date:  Thu,  13  Feb  2025  15:47:33  +0000 \r\n\
                     \r\n";

        let parsed = parse_email(raw).unwrap();
        let date = parsed.date.unwrap();
        assert_eq!(date.year, 2025);
        assert_eq!(date.month, 2);
        assert_eq!(date.day, 13);
    }

    #[test]
    fn parse_multipart_related_with_inline_images() {
        // multipart/related — common for HTML emails with inline images
        let raw = b"From: a@b.com\r\n\
                     Date: Thu, 13 Feb 2025 15:47:33 +0000\r\n\
                     Content-Type: multipart/related; boundary=\"rel\"\r\n\
                     \r\n\
                     --rel\r\n\
                     Content-Type: text/html\r\n\
                     \r\n\
                     <html><img src=\"cid:logo\"></html>\r\n\
                     --rel\r\n\
                     Content-Type: image/jpeg\r\n\
                     Content-ID: <logo>\r\n\
                     Content-Disposition: inline; filename=\"logo.jpg\"\r\n\
                     \r\n\
                     JPEG_DATA\r\n\
                     --rel--";

        let parsed = parse_email(raw).unwrap();
        assert!(parsed.body_html.is_some());
        assert_eq!(parsed.attachments.len(), 1);
        assert!(parsed.attachments[0].is_inline);
        assert_eq!(parsed.attachments[0].content_id.as_deref(), Some("logo"));
        assert_eq!(parsed.attachments[0].filename.as_deref(), Some("logo.jpg"));
    }

    #[test]
    fn parse_minimal_message_from_only() {
        // Absolute minimum valid message: just From header
        let raw = b"From: a@b.com\r\n\r\n";

        let parsed = parse_email(raw).unwrap();
        assert_eq!(parsed.from.email, "a@b.com");
        assert!(parsed.subject.is_none());
        assert!(parsed.date.is_none());
        assert!(parsed.body_text.is_none());
    }

    #[test]
    fn parse_multiple_same_headers() {
        // Multiple headers with the same name — should take the first one
        let raw = b"From: first@example.com\r\n\
                     From: second@example.com\r\n\
                     Subject: First\r\n\
                     Subject: Second\r\n\
                     Date: Thu, 13 Feb 2025 15:47:33 +0000\r\n\
                     \r\n";

        let parsed = parse_email(raw).unwrap();
        assert_eq!(parsed.from.email, "first@example.com");
        assert_eq!(parsed.subject.as_deref(), Some("First"));
    }

    #[test]
    fn parse_date_all_named_timezones() {
        let test_cases = [
            ("EST", -300),
            ("EDT", -240),
            ("CST", -360),
            ("CDT", -300),
            ("MST", -420),
            ("MDT", -360),
            ("PST", -480),
            ("PDT", -420),
            ("GMT", 0),
            ("UTC", 0),
            ("UT", 0),
        ];
        for (tz_name, expected_offset) in test_cases {
            let raw = format!("From: a@b.com\r\nDate: Thu, 13 Feb 2025 12:00:00 {tz_name}\r\n\r\n");
            let parsed = parse_email(raw.as_bytes()).unwrap();
            let date = parsed.date.unwrap();
            assert_eq!(
                date.tz_offset_minutes, expected_offset,
                "Failed for timezone {tz_name}"
            );
        }
    }

    #[test]
    fn parse_boundary_with_special_chars() {
        // Boundary containing special characters (RFC 2046 allows certain chars)
        let raw = b"From: a@b.com\r\n\
                     Date: Thu, 13 Feb 2025 15:47:33 +0000\r\n\
                     Content-Type: multipart/mixed; boundary=\"----=_Part_123+abc\"\r\n\
                     \r\n\
                     ------=_Part_123+abc\r\n\
                     Content-Type: text/plain\r\n\
                     \r\n\
                     Body text\r\n\
                     ------=_Part_123+abc--";

        let parsed = parse_email(raw).unwrap();
        assert_eq!(parsed.body_text.as_deref(), Some("Body text"));
    }

    #[test]
    fn parse_truncated_base64_body() {
        // Truncated base64 — should not crash, fall back to raw bytes
        let raw = b"From: a@b.com\r\n\
                     Date: Thu, 13 Feb 2025 15:47:33 +0000\r\n\
                     Content-Type: text/plain; charset=utf-8\r\n\
                     Content-Transfer-Encoding: base64\r\n\
                     \r\n\
                     SGVsbG8gV29yb===invalid";

        let parsed = parse_email(raw).unwrap();
        // Should not panic — either decodes partial or falls back
        assert!(parsed.body_text.is_some());
    }

    #[test]
    fn parse_address_group_syntax() {
        // Group address syntax: "Group: addr1, addr2;"
        // Should parse what it can without crashing
        let raw = b"From: sender@example.com\r\n\
                     To: Undisclosed:;\r\n\
                     Date: Thu, 13 Feb 2025 15:47:33 +0000\r\n\
                     \r\n";

        let parsed = parse_email(raw).unwrap();
        // Group syntax with no addresses — To list may be empty
        assert_eq!(parsed.from.email, "sender@example.com");
    }

    #[test]
    fn parse_iso2022jp_encoded_word() {
        // ISO-2022-JP encoded word (common in Japanese email)
        // "テスト" in ISO-2022-JP base64
        let raw = b"From: a@b.com\r\n\
                     Subject: =?ISO-2022-JP?B?GyRCJUYlOSVIGyhC?=\r\n\
                     Date: Thu, 13 Feb 2025 15:47:33 +0000\r\n\
                     \r\n";

        let parsed = parse_email(raw).unwrap();
        // encoding_rs should handle ISO-2022-JP
        assert!(parsed.subject.is_some());
        assert_eq!(parsed.subject.as_deref(), Some("テスト"));
    }

    #[test]
    fn parse_multipart_missing_parts_tolerance() {
        // Multipart header declared but body is completely different (partial fetch)
        let raw = b"From: a@b.com\r\n\
                     Date: Thu, 13 Feb 2025 15:47:33 +0000\r\n\
                     Content-Type: multipart/mixed; boundary=\"never-appears\"\r\n\
                     \r\n\
                     This body doesn't contain any boundaries at all.";

        let parsed = parse_email(raw).unwrap();
        // Should not fail — partial input tolerance
        assert!(parsed.body_text.is_none());
        assert!(parsed.body_html.is_none());
    }

    #[test]
    fn parse_encoded_word_in_multiple_header_types() {
        // Encoded words in From display name AND Subject
        let raw = b"From: =?UTF-8?Q?M=C3=BCller?= <mueller@example.com>\r\n\
                     To: =?UTF-8?B?U21pdGg=?= <smith@example.com>\r\n\
                     Subject: =?UTF-8?Q?Caf=C3=A9?=\r\n\
                     Date: Thu, 13 Feb 2025 15:47:33 +0000\r\n\
                     \r\n";

        let parsed = parse_email(raw).unwrap();
        assert_eq!(parsed.from.name.as_deref(), Some("Müller"));
        assert_eq!(parsed.to[0].name.as_deref(), Some("Smith"));
        assert_eq!(parsed.subject.as_deref(), Some("Café"));
    }

    #[test]
    fn parse_attachment_size_reflects_part_body() {
        // Verify attachment size field is set correctly
        let raw = b"From: a@b.com\r\n\
                     Date: Thu, 13 Feb 2025 15:47:33 +0000\r\n\
                     Content-Type: multipart/mixed; boundary=\"b\"\r\n\
                     \r\n\
                     --b\r\n\
                     Content-Type: text/plain\r\n\
                     \r\n\
                     Body\r\n\
                     --b\r\n\
                     Content-Type: application/pdf\r\n\
                     Content-Disposition: attachment; filename=\"f.pdf\"\r\n\
                     \r\n\
                     0123456789\r\n\
                     --b--";

        let parsed = parse_email(raw).unwrap();
        assert_eq!(parsed.attachments.len(), 1);
        assert_eq!(parsed.attachments[0].size, Some(10));
    }

    #[test]
    fn parse_unquoted_boundary() {
        // Boundary value without quotes (RFC 2046 allows this)
        let raw = b"From: a@b.com\r\n\
                     Date: Thu, 13 Feb 2025 15:47:33 +0000\r\n\
                     Content-Type: multipart/mixed; boundary=simple_boundary\r\n\
                     \r\n\
                     --simple_boundary\r\n\
                     Content-Type: text/plain\r\n\
                     \r\n\
                     Text\r\n\
                     --simple_boundary--";

        let parsed = parse_email(raw).unwrap();
        assert_eq!(parsed.body_text.as_deref(), Some("Text"));
    }

    #[test]
    fn parse_message_id_without_angle_brackets() {
        // Some broken mailers omit angle brackets on Message-ID.
        // The parser tolerates this and returns the bare value.
        let raw = b"From: a@b.com\r\n\
                     Message-ID: bare-id@host.com\r\n\
                     Date: Thu, 13 Feb 2025 15:47:33 +0000\r\n\
                     \r\n";

        let parsed = parse_email(raw).unwrap();
        assert_eq!(parsed.message_id.as_deref(), Some("bare-id@host.com"));
    }

    #[test]
    fn parse_empty_references_header() {
        // References header with no valid message-ids
        let raw = b"From: a@b.com\r\n\
                     References: \r\n\
                     Date: Thu, 13 Feb 2025 15:47:33 +0000\r\n\
                     \r\n";

        let parsed = parse_email(raw).unwrap();
        assert!(parsed.references.is_none());
    }

    #[test]
    fn parse_large_multipart_many_attachments() {
        // Message with many attachment parts — verify section numbering
        let mut raw = b"From: a@b.com\r\nDate: Thu, 13 Feb 2025 15:47:33 +0000\r\n\
                         Content-Type: multipart/mixed; boundary=\"multi\"\r\n\r\n"
            .to_vec();

        raw.extend_from_slice(b"--multi\r\nContent-Type: text/plain\r\n\r\nBody\r\n");
        for i in 1..=5 {
            raw.extend_from_slice(
                format!(
                    "--multi\r\nContent-Type: application/octet-stream\r\n\
                     Content-Disposition: attachment; filename=\"file{i}.bin\"\r\n\r\n\
                     DATA{i}\r\n"
                )
                .as_bytes(),
            );
        }
        raw.extend_from_slice(b"--multi--");

        let parsed = parse_email(&raw).unwrap();
        assert_eq!(parsed.body_text.as_deref(), Some("Body"));
        assert_eq!(parsed.attachments.len(), 5);
        for (i, att) in parsed.attachments.iter().enumerate() {
            assert_eq!(
                att.section.as_deref(),
                Some(&(i + 2).to_string() as &str),
                "Wrong section for attachment {i}"
            );
            assert_eq!(
                att.filename.as_deref(),
                Some(&format!("file{}.bin", i + 1) as &str)
            );
        }
    }

    #[test]
    fn parse_message_id_empty_brackets() {
        // Empty angle brackets should return None
        let raw = b"From: a@b.com\r\n\
                     Message-ID: <>\r\n\
                     Date: Thu, 13 Feb 2025 15:47:33 +0000\r\n\
                     \r\n";

        let parsed = parse_email(raw).unwrap();
        assert!(parsed.message_id.is_none());
    }

    #[test]
    fn parse_message_id_empty_value() {
        // Completely empty Message-ID value should return None
        let raw = b"From: a@b.com\r\n\
                     Message-ID: \r\n\
                     Date: Thu, 13 Feb 2025 15:47:33 +0000\r\n\
                     \r\n";

        let parsed = parse_email(raw).unwrap();
        assert!(parsed.message_id.is_none());
    }

    #[test]
    fn rfc2231_param_boundary_check() {
        // The "filename*=" match must be at a parameter boundary,
        // not embedded in another parameter name like "xfilename*="
        let disposition = "attachment; xfilename*=UTF-8''bad.pdf; filename*=UTF-8''good.pdf";
        let result = extract_rfc2231_param(disposition, "filename");
        assert_eq!(result.as_deref(), Some("good.pdf"));
    }

    #[test]
    fn rfc2231_param_at_start() {
        // Parameter at the very start of the value (no preceding `;`)
        let value = "filename*=UTF-8''test.pdf";
        let result = extract_rfc2231_param(value, "filename");
        assert_eq!(result.as_deref(), Some("test.pdf"));
    }

    #[test]
    fn parse_quoted_display_name_with_comma() {
        // Display name with comma must be in a quoted-string (RFC 5322 Section 3.4)
        let raw = b"From: \"Doe, John\" <john@example.com>\r\n\
                     Date: Thu, 13 Feb 2025 15:47:33 +0000\r\n\
                     \r\n";

        let parsed = parse_email(raw).unwrap();
        assert_eq!(parsed.from.name.as_deref(), Some("Doe, John"));
        assert_eq!(parsed.from.email, "john@example.com");
    }

    #[test]
    fn parse_quoted_display_name_with_escaped_chars() {
        // Backslash-escaped characters in quoted display name (RFC 5322 Section 3.2.4)
        let raw = b"From: \"John \\\"Doc\\\" Doe\" <john@example.com>\r\n\
                     Date: Thu, 13 Feb 2025 15:47:33 +0000\r\n\
                     \r\n";

        let parsed = parse_email(raw).unwrap();
        assert_eq!(parsed.from.name.as_deref(), Some("John \"Doc\" Doe"));
        assert_eq!(parsed.from.email, "john@example.com");
    }

    #[test]
    fn unescape_quoted_string_backslash() {
        assert_eq!(unescape_quoted_string("hello"), "hello");
        assert_eq!(unescape_quoted_string("a\\\\b"), "a\\b");
        assert_eq!(unescape_quoted_string("a\\\"b"), "a\"b");
        assert_eq!(unescape_quoted_string("trailing\\"), "trailing\\");
    }

    #[test]
    fn parse_address_list_with_escaped_quotes_in_display_name() {
        // RFC 5322 Section 3.2.4: backslash-escaped quotes within a quoted-string
        // must not prematurely end the quoted context during address list splitting.
        let raw = b"From: a@b.com\r\n\
                     To: \"A\\\"B\" <a@x.com>, c@d.com\r\n\
                     Date: Thu, 13 Feb 2025 15:47:33 +0000\r\n\
                     \r\n";

        let parsed = parse_email(raw).unwrap();
        assert_eq!(
            parsed.to.len(),
            2,
            "Expected 2 To addresses but got {:?}",
            parsed.to
        );
        assert_eq!(parsed.to[0].email, "a@x.com");
        assert_eq!(parsed.to[0].name.as_deref(), Some("A\"B"));
        assert_eq!(parsed.to[1].email, "c@d.com");
    }

    #[test]
    fn parse_rfc2231_continuation_filename() {
        // RFC 2231 Section 3: long filenames split across continuation parameters.
        // filename*0="very_long_"; filename*1="filename.pdf"
        let raw = b"From: a@b.com\r\n\
                     Date: Thu, 13 Feb 2025 15:47:33 +0000\r\n\
                     Content-Type: multipart/mixed; boundary=\"b\"\r\n\
                     \r\n\
                     --b\r\n\
                     Content-Type: text/plain\r\n\
                     \r\n\
                     Body\r\n\
                     --b\r\n\
                     Content-Type: application/pdf\r\n\
                     Content-Disposition: attachment; filename*0=\"very_long_\"; filename*1=\"filename.pdf\"\r\n\
                     \r\n\
                     DATA\r\n\
                     --b--";

        let parsed = parse_email(raw).unwrap();
        assert_eq!(parsed.attachments.len(), 1);
        assert_eq!(
            parsed.attachments[0].filename.as_deref(),
            Some("very_long_filename.pdf"),
            "RFC 2231 continuation filename not reassembled"
        );
    }

    #[test]
    fn parse_rfc2231_continuation_with_charset() {
        // RFC 2231 Section 3+4: continuation with charset encoding.
        // filename*0*=UTF-8''r%C3%A9sum; filename*1*=%C3%A9.pdf
        let raw = b"From: a@b.com\r\n\
                     Date: Thu, 13 Feb 2025 15:47:33 +0000\r\n\
                     Content-Type: multipart/mixed; boundary=\"b\"\r\n\
                     \r\n\
                     --b\r\n\
                     Content-Type: text/plain\r\n\
                     \r\n\
                     Body\r\n\
                     --b\r\n\
                     Content-Type: application/pdf\r\n\
                     Content-Disposition: attachment; filename*0*=UTF-8''r%C3%A9sum; filename*1*=%C3%A9.pdf\r\n\
                     \r\n\
                     DATA\r\n\
                     --b--";

        let parsed = parse_email(raw).unwrap();
        assert_eq!(parsed.attachments.len(), 1);
        assert_eq!(
            parsed.attachments[0].filename.as_deref(),
            Some("résumé.pdf"),
            "RFC 2231 continuation with charset not reassembled"
        );
    }

    /// commas inside RFC 5322 Section 3.2.2 parenthesized comments
    /// must NOT split the address list. The comment `(Doe, John)` is a single
    /// unit — the comma inside it is not an address separator.
    #[test]
    fn parse_address_comment_with_comma() {
        let raw = b"From: sender@example.com\r\n\
                     To: user@example.com (Doe, John), other@example.com\r\n\
                     Date: Thu, 13 Feb 2025 15:47:33 +0000\r\n\
                     \r\n";

        let parsed = parse_email(raw).unwrap();
        assert_eq!(
            parsed.to.len(),
            2,
            "Expected 2 To addresses but got {:?}",
            parsed.to
        );
        assert_eq!(parsed.to[0].email, "user@example.com");
        assert_eq!(
            parsed.to[0].name.as_deref(),
            Some("Doe, John"),
            "Display name from comment should be preserved intact"
        );
        assert_eq!(parsed.to[1].email, "other@example.com");
    }

    #[test]
    fn parse_header_unfolding_preserves_wsp() {
        // RFC 5322 Section 2.2.3: unfolding removes the CRLF but the leading
        // WSP (tab or space) is part of the FWS and must be preserved.
        let raw = b"From: a@b.com\r\nSubject: Hello\r\n\tWorld\r\nDate: Thu, 13 Feb 2025 15:47:33 +0000\r\n\r\n";

        let parsed = parse_email(raw).unwrap();
        // The tab should be preserved between "Hello" and "World"
        assert_eq!(
            parsed.subject.as_deref(),
            Some("Hello\tWorld"),
            "Tab from continuation line should be preserved per RFC 5322 Section 2.2.3"
        );
    }

    #[test]
    fn datetime_to_unix_timestamp() {
        use crate::types::DateTime;

        // 2025-02-13 15:47:33 +0000 → known Unix timestamp
        let dt = DateTime {
            year: 2025,
            month: 2,
            day: 13,
            hour: 15,
            minute: 47,
            second: 33,
            tz_offset_minutes: 0,
        };
        // 2025-02-13T15:47:33Z = 1739461653
        assert_eq!(dt.to_unix_timestamp(), 1_739_461_653);

        // Same instant with +0530 offset (local 21:17:33)
        let dt_offset = DateTime {
            year: 2025,
            month: 2,
            day: 13,
            hour: 21,
            minute: 17,
            second: 33,
            tz_offset_minutes: 330,
        };
        assert_eq!(dt_offset.to_unix_timestamp(), dt.to_unix_timestamp());
    }

    #[test]
    fn datetime_from_unix_timestamp() {
        use crate::types::DateTime;

        let ts = 1_739_461_653_i64; // 2025-02-13T15:47:33Z
        let dt = DateTime::from_unix_timestamp(ts, 0);
        assert_eq!(dt.year, 2025);
        assert_eq!(dt.month, 2);
        assert_eq!(dt.day, 13);
        assert_eq!(dt.hour, 15);
        assert_eq!(dt.minute, 47);
        assert_eq!(dt.second, 33);
        assert_eq!(dt.tz_offset_minutes, 0);

        // With +0530 offset
        let dt_offset = DateTime::from_unix_timestamp(ts, 330);
        assert_eq!(dt_offset.hour, 21);
        assert_eq!(dt_offset.minute, 17);
    }

    #[test]
    fn datetime_round_trip_timestamp() {
        use crate::types::DateTime;

        let dt = DateTime {
            year: 2025,
            month: 12,
            day: 31,
            hour: 23,
            minute: 59,
            second: 59,
            tz_offset_minutes: -480,
        };
        let ts = dt.to_unix_timestamp();
        let restored = DateTime::from_unix_timestamp(ts, -480);
        assert_eq!(dt, restored);
    }

    #[test]
    fn datetime_ord_comparison() {
        use crate::types::DateTime;

        // Same instant in different timezones should be equal
        let utc = DateTime {
            year: 2025,
            month: 1,
            day: 1,
            hour: 12,
            minute: 0,
            second: 0,
            tz_offset_minutes: 0,
        };
        let est = DateTime {
            year: 2025,
            month: 1,
            day: 1,
            hour: 7,
            minute: 0,
            second: 0,
            tz_offset_minutes: -300,
        };
        assert_eq!(utc.cmp(&est), std::cmp::Ordering::Equal);

        // Later timestamp should be greater
        let later = DateTime {
            year: 2025,
            month: 1,
            day: 1,
            hour: 13,
            minute: 0,
            second: 0,
            tz_offset_minutes: 0,
        };
        assert!(later > utc);
    }

    #[test]
    fn datetime_epoch() {
        use crate::types::DateTime;

        let epoch = DateTime::from_unix_timestamp(0, 0);
        assert_eq!(epoch.year, 1970);
        assert_eq!(epoch.month, 1);
        assert_eq!(epoch.day, 1);
        assert_eq!(epoch.hour, 0);
        assert_eq!(epoch.minute, 0);
        assert_eq!(epoch.second, 0);
        assert_eq!(epoch.to_unix_timestamp(), 0);
    }

    #[test]
    fn parse_headers_only_extracts_metadata() {
        let raw = b"From: sender@example.com\r\n\
                     To: recipient@example.com\r\n\
                     Subject: Test\r\n\
                     Date: Thu, 13 Feb 2025 15:47:33 +0000\r\n\
                     Message-ID: <abc123@example.com>\r\n\
                     Content-Type: multipart/mixed; boundary=\"b\"\r\n\
                     \r\n\
                     --b\r\n\
                     Content-Type: text/plain\r\n\
                     \r\n\
                     This body should NOT be parsed\r\n\
                     --b\r\n\
                     Content-Type: application/pdf\r\n\
                     Content-Disposition: attachment; filename=\"doc.pdf\"\r\n\
                     \r\n\
                     PDF_DATA\r\n\
                     --b--";

        let parsed = parse_headers_only(raw).unwrap();

        // Header fields should be populated
        assert_eq!(parsed.from.email, "sender@example.com");
        assert_eq!(parsed.to.len(), 1);
        assert_eq!(parsed.subject.as_deref(), Some("Test"));
        assert_eq!(parsed.message_id.as_deref(), Some("abc123@example.com"));
        assert!(parsed.date.is_some());

        // Body fields should be empty (skipped)
        assert!(parsed.body_text.is_none());
        assert!(parsed.body_html.is_none());
        assert!(parsed.attachments.is_empty());
    }

    #[test]
    fn parse_headers_only_empty_input() {
        let result = parse_headers_only(b"");
        assert!(matches!(result, Err(Error::EmptyInput)));
    }

    #[test]
    fn parse_headers_only_missing_from() {
        let raw = b"Subject: No From\r\nDate: Thu, 13 Feb 2025 15:47:33 +0000\r\n\r\n";
        let result = parse_headers_only(raw);
        assert!(matches!(result, Err(Error::MissingFrom)));
    }

    // -----------------------------------------------------------------------
    // extract_param quoted-string handling (RFC 5322 §3.2.4)
    // -----------------------------------------------------------------------

    #[test]
    fn extract_param_unescapes_backslash_in_filename() {
        // RFC 5322 Section 3.2.4: quoted-pair `\\` in a quoted-string represents
        // a literal backslash. extract_param must unescape it.
        let raw = b"From: a@b.com\r\n\
                     Date: Thu, 13 Feb 2025 15:47:33 +0000\r\n\
                     Content-Type: multipart/mixed; boundary=\"b\"\r\n\
                     \r\n\
                     --b\r\n\
                     Content-Type: text/plain\r\n\
                     \r\n\
                     Body\r\n\
                     --b\r\n\
                     Content-Type: application/pdf\r\n\
                     Content-Disposition: attachment; filename=\"path\\\\file.pdf\"\r\n\
                     \r\n\
                     DATA\r\n\
                     --b--";

        let parsed = parse_email(raw).unwrap();
        assert_eq!(parsed.attachments.len(), 1);
        // The filename in the message is `path\\file.pdf` (escaped backslash).
        // After unescaping per RFC 5322 Section 3.2.4, it should be `path\file.pdf`.
        assert_eq!(
            parsed.attachments[0].filename.as_deref(),
            Some("path\\file.pdf"),
            "Backslash in quoted-string filename must be unescaped per RFC 5322 Section 3.2.4"
        );
    }

    #[test]
    fn extract_param_handles_escaped_quote_in_filename() {
        // RFC 5322 Section 3.2.4: quoted-pair `\"` in a quoted-string represents
        // a literal double-quote. extract_param must skip escaped quotes when
        // finding the closing quote, and then unescape the result.
        let raw = b"From: a@b.com\r\n\
                     Date: Thu, 13 Feb 2025 15:47:33 +0000\r\n\
                     Content-Type: multipart/mixed; boundary=\"b\"\r\n\
                     \r\n\
                     --b\r\n\
                     Content-Type: text/plain\r\n\
                     \r\n\
                     Body\r\n\
                     --b\r\n\
                     Content-Type: application/pdf\r\n\
                     Content-Disposition: attachment; filename=\"file\\\"name.pdf\"\r\n\
                     \r\n\
                     DATA\r\n\
                     --b--";

        let parsed = parse_email(raw).unwrap();
        assert_eq!(parsed.attachments.len(), 1);
        // The filename in the message is `file\"name.pdf` (escaped quote).
        // After unescaping, it should be `file"name.pdf`.
        assert_eq!(
            parsed.attachments[0].filename.as_deref(),
            Some("file\"name.pdf"),
            "Escaped quote in quoted-string filename must be handled per RFC 5322 Section 3.2.4"
        );
    }

    #[test]
    fn build_then_parse_filename_with_backslash_round_trip() {
        // Round-trip: build a message with a backslash in the filename, then parse it.
        // The builder escapes `\` → `\\` per RFC 5322 Section 3.2.4.
        // The parser must unescape `\\` → `\` to recover the original filename.
        let email = crate::types::OutgoingEmail {
            from: Address {
                name: None,
                email: "a@b.com".into(),
            },
            to: vec![Address {
                name: None,
                email: "to@b.com".into(),
            }],
            cc: vec![],
            bcc: vec![],
            reply_to: None,
            subject: "test".into(),
            body_text: Some("Body".into()),
            body_html: None,
            in_reply_to: None,
            references: None,
            attachments: vec![crate::types::OutgoingAttachment {
                filename: "path\\file.pdf".into(),
                content_type: "application/pdf".into(),
                data: b"data".to_vec(),
            }],
        };

        let built = crate::build_message(&email).unwrap();
        let parsed = parse_email(&built.raw).unwrap();

        assert_eq!(parsed.attachments.len(), 1);
        assert_eq!(
            parsed.attachments[0].filename.as_deref(),
            Some("path\\file.pdf"),
            "Round-trip filename with backslash must be preserved"
        );
    }

    #[test]
    fn build_then_parse_filename_with_quote_round_trip() {
        // Round-trip: build a message with a double-quote in the filename.
        let email = crate::types::OutgoingEmail {
            from: Address {
                name: None,
                email: "a@b.com".into(),
            },
            to: vec![Address {
                name: None,
                email: "to@b.com".into(),
            }],
            cc: vec![],
            bcc: vec![],
            reply_to: None,
            subject: "test".into(),
            body_text: Some("Body".into()),
            body_html: None,
            in_reply_to: None,
            references: None,
            attachments: vec![crate::types::OutgoingAttachment {
                filename: "file\"name.pdf".into(),
                content_type: "application/pdf".into(),
                data: b"data".to_vec(),
            }],
        };

        let built = crate::build_message(&email).unwrap();
        let parsed = parse_email(&built.raw).unwrap();

        assert_eq!(parsed.attachments.len(), 1);
        assert_eq!(
            parsed.attachments[0].filename.as_deref(),
            Some("file\"name.pdf"),
            "Round-trip filename with double-quote must be preserved"
        );
    }

    // -----------------------------------------------------------------------
    // to_ascii_lowercase for non-ASCII safety (RFC 6532)
    // -----------------------------------------------------------------------

    #[test]
    fn extract_param_with_non_ascii_before_param() {
        // RFC 6532 allows UTF-8 directly in headers. If a non-ASCII character
        // that changes byte length under Unicode lowercasing appears before
        // the parameter we're searching for, to_lowercase() would misalign
        // byte offsets. to_ascii_lowercase() preserves byte length.
        //
        // İ (U+0130, 2 bytes UTF-8) lowercases to 'i' + combining dot (3 bytes)
        // under full Unicode rules, but stays 2 bytes under ASCII-only rules.
        //
        // We test extract_param directly with İ before the target parameter.
        let header_value = "attachment; description=\"\u{0130}stanbul\"; filename=\"report.pdf\"";
        let result = extract_param(header_value, "filename");
        assert_eq!(
            result.as_deref(),
            Some("report.pdf"),
            "extract_param must work when non-ASCII chars that change byte length \
             under Unicode lowercasing appear before the target parameter (RFC 6532)"
        );
    }

    // -----------------------------------------------------------------------
    // date field range validation (RFC 5322 Section 3.3)
    // -----------------------------------------------------------------------

    #[test]
    fn parse_date_rejects_invalid_hour() {
        // RFC 5322 Section 3.3: hour is 0-23.
        let raw = b"From: a@b.com\r\n\
                     Date: Thu, 13 Feb 2025 25:00:00 +0000\r\n\
                     \r\n";
        let parsed = parse_email(raw).unwrap();
        // Invalid hour (25) should cause date parsing to return None
        assert!(
            parsed.date.is_none(),
            "Date with hour=25 should be rejected per RFC 5322 Section 3.3"
        );
    }

    #[test]
    fn parse_date_rejects_invalid_minute() {
        // RFC 5322 Section 3.3: minute is 0-59.
        let raw = b"From: a@b.com\r\n\
                     Date: Thu, 13 Feb 2025 12:60:00 +0000\r\n\
                     \r\n";
        let parsed = parse_email(raw).unwrap();
        assert!(
            parsed.date.is_none(),
            "Date with minute=60 should be rejected per RFC 5322 Section 3.3"
        );
    }

    #[test]
    fn parse_date_rejects_invalid_second() {
        // RFC 5322 Section 3.3: second is 0-60 (60 for leap second).
        let raw = b"From: a@b.com\r\n\
                     Date: Thu, 13 Feb 2025 12:00:61 +0000\r\n\
                     \r\n";
        let parsed = parse_email(raw).unwrap();
        assert!(
            parsed.date.is_none(),
            "Date with second=61 should be rejected per RFC 5322 Section 3.3"
        );
    }

    #[test]
    fn parse_date_rejects_invalid_day() {
        // RFC 5322 Section 3.3: day is 1-31.
        let raw = b"From: a@b.com\r\n\
                     Date: Thu, 32 Feb 2025 12:00:00 +0000\r\n\
                     \r\n";
        let parsed = parse_email(raw).unwrap();
        assert!(
            parsed.date.is_none(),
            "Date with day=32 should be rejected per RFC 5322 Section 3.3"
        );
    }

    #[test]
    fn parse_date_rejects_day_zero() {
        // RFC 5322 Section 3.3: day starts at 1.
        let raw = b"From: a@b.com\r\n\
                     Date: Thu, 00 Feb 2025 12:00:00 +0000\r\n\
                     \r\n";
        let parsed = parse_email(raw).unwrap();
        assert!(
            parsed.date.is_none(),
            "Date with day=0 should be rejected per RFC 5322 Section 3.3"
        );
    }

    #[test]
    fn parse_date_allows_leap_second() {
        // RFC 5322 Section 3.3: second 60 is valid (leap second).
        let raw = b"From: a@b.com\r\n\
                     Date: Thu, 30 Jun 2015 23:59:60 +0000\r\n\
                     \r\n";
        let parsed = parse_email(raw).unwrap();
        let date = parsed.date.expect("Leap second (60) should be accepted");
        assert_eq!(date.second, 60);
    }

    #[test]
    fn parse_date_with_comment_between_tokens() {
        // RFC 5322 Section 4.3 (obsolete date syntax) allows CFWS
        // (comments and folding white space) between date tokens.
        // A comment like "(Friday)" between day and month must not cause
        // the date to fail to parse.
        let raw = b"From: a@b.com\r\n\
                     Date: Thu, 13 (February) Feb 2025 15:47:33 +0000\r\n\
                     \r\n";

        let parsed = parse_email(raw).unwrap();
        let date = parsed
            .date
            .expect("Date with CFWS comment must parse per RFC 5322 Section 4.3");
        assert_eq!(date.year, 2025);
        assert_eq!(date.month, 2);
        assert_eq!(date.day, 13);
    }

    #[test]
    fn parse_date_with_trailing_comment() {
        // Common: trailing comment like "(UTC)" after timezone.
        // This already works but we add a test to ensure it stays working.
        let raw = b"From: a@b.com\r\n\
                     Date: Thu, 13 Feb 2025 15:47:33 +0000 (UTC)\r\n\
                     \r\n";

        let parsed = parse_email(raw).unwrap();
        let date = parsed.date.expect("Date with trailing comment must parse");
        assert_eq!(date.year, 2025);
        assert_eq!(date.tz_offset_minutes, 0);
    }

    #[test]
    fn parse_date_with_nested_comments() {
        // RFC 5322 Section 3.2.2: comments can be nested.
        let raw = b"From: a@b.com\r\n\
                     Date: Thu, 13 Feb 2025 15:47:33 +0000 (UTC (nested))\r\n\
                     \r\n";

        let parsed = parse_email(raw).unwrap();
        let date = parsed.date.expect("Date with nested comment must parse");
        assert_eq!(date.year, 2025);
    }

    #[test]
    fn parse_display_name_ending_with_escaped_quote() {
        // RFC 5322 Section 3.2.4 — a display name ending with an
        // escaped quote like `"She said \"hello\""` must parse correctly.
        // The outer quotes are the quoted-string delimiters; the inner `\"`
        // sequences are quoted-pairs that represent literal `"`.
        let raw = b"From: \"She said \\\"hello\\\"\" <she@example.com>\r\n\
                     Date: Thu, 13 Feb 2025 15:47:33 +0000\r\n\
                     \r\n";

        let parsed = parse_email(raw).unwrap();
        assert_eq!(
            parsed.from.name.as_deref(),
            Some("She said \"hello\""),
            "Display name ending with escaped quote must be parsed correctly \
             per RFC 5322 Section 3.2.4"
        );
    }

    #[test]
    fn address_from_str_ending_with_escaped_quote() {
        // Same bug in Address::from_str — trim_matches('"') strips too many quotes.
        let addr: Address = "\"She said \\\"hello\\\"\" <she@example.com>"
            .parse()
            .unwrap();
        assert_eq!(
            addr.name.as_deref(),
            Some("She said \"hello\""),
            "Address::from_str must handle display names ending with escaped quotes"
        );
    }

    #[test]
    fn boundary_must_be_at_line_start() {
        // RFC 2046 Section 5.1.1 requires the boundary delimiter
        // to appear at the beginning of a line (preceded by CRLF or at the
        // start of the body). A boundary string appearing mid-line in body
        // content must NOT be treated as a boundary delimiter.
        let raw = b"From: a@b.com\r\n\
                     Date: Thu, 13 Feb 2025 15:47:33 +0000\r\n\
                     Content-Type: multipart/mixed; boundary=\"BOUND\"\r\n\
                     \r\n\
                     --BOUND\r\n\
                     Content-Type: text/plain\r\n\
                     \r\n\
                     This line mentions --BOUND in the middle\r\n\
                     --BOUND--";

        let parsed = parse_email(raw).unwrap();
        // The body should contain the mid-line "--BOUND" as literal text,
        // not split into a second part at that point.
        let text = parsed.body_text.as_deref().unwrap_or("");
        assert!(
            text.contains("--BOUND"),
            "Mid-line boundary must be treated as literal text per RFC 2046 Section 5.1.1, \
             but body_text was: {text:?}"
        );
    }

    #[test]
    fn mime_type_exact_match_not_prefix() {
        // A hypothetical MIME type like "text/plaintext" must NOT be treated
        // as text/plain body. Using starts_with("text/plain") would
        // incorrectly match it; exact equality is required.
        let raw = b"From: a@b.com\r\n\
                     Date: Thu, 13 Feb 2025 15:47:33 +0000\r\n\
                     Content-Type: multipart/mixed; boundary=\"b\"\r\n\
                     \r\n\
                     --b\r\n\
                     Content-Type: text/plaintext\r\n\
                     \r\n\
                     Not really plain text\r\n\
                     --b--";

        let parsed = parse_email(raw).unwrap();
        // text/plaintext is NOT text/plain — it should be treated as an
        // attachment, not extracted as body_text.
        assert!(
            parsed.body_text.is_none(),
            "text/plaintext must not be treated as text/plain body"
        );
        assert_eq!(
            parsed.attachments.len(),
            1,
            "text/plaintext should be treated as an attachment"
        );
    }

    #[test]
    fn parse_single_part_non_text_is_attachment() {
        // A single-part message with Content-Type: image/jpeg should be treated
        // as an attachment, not as body_text (RFC 2046; requirements: "A part is
        // an attachment if... a non-text/non-multipart part").
        let raw = b"From: a@b.com\r\n\
                     Date: Thu, 13 Feb 2025 15:47:33 +0000\r\n\
                     Content-Type: image/jpeg\r\n\
                     Content-Transfer-Encoding: base64\r\n\
                     \r\n\
                     /9j/4AAQSkZJRg==";

        let parsed = parse_email(raw).unwrap();
        assert!(
            parsed.body_text.is_none(),
            "image/jpeg single-part must not populate body_text"
        );
        assert_eq!(
            parsed.attachments.len(),
            1,
            "image/jpeg single-part must be treated as an attachment"
        );
        assert_eq!(parsed.attachments[0].content_type, "image/jpeg");
        assert_eq!(parsed.attachments[0].section.as_deref(), Some("1"));
    }

    #[test]
    fn parse_single_part_application_pdf_is_attachment() {
        // Single-part application/pdf should be an attachment, not body_text.
        let raw = b"From: a@b.com\r\n\
                     Date: Thu, 13 Feb 2025 15:47:33 +0000\r\n\
                     Content-Type: application/pdf; name=\"doc.pdf\"\r\n\
                     Content-Disposition: attachment; filename=\"doc.pdf\"\r\n\
                     Content-Transfer-Encoding: base64\r\n\
                     \r\n\
                     JVBERi0xLjQK";

        let parsed = parse_email(raw).unwrap();
        assert!(
            parsed.body_text.is_none(),
            "application/pdf must not populate body_text"
        );
        assert_eq!(parsed.attachments.len(), 1);
        assert_eq!(parsed.attachments[0].content_type, "application/pdf");
        assert_eq!(parsed.attachments[0].filename.as_deref(), Some("doc.pdf"));
        assert!(!parsed.attachments[0].is_inline);
    }

    #[test]
    fn parse_single_part_text_plain_with_attachment_disposition() {
        // text/plain with Content-Disposition: attachment should be treated as
        // an attachment, not body_text (requirements: "A part is an attachment
        // if it has Content-Disposition: attachment").
        let raw = b"From: a@b.com\r\n\
                     Date: Thu, 13 Feb 2025 15:47:33 +0000\r\n\
                     Content-Type: text/plain; charset=utf-8\r\n\
                     Content-Disposition: attachment; filename=\"log.txt\"\r\n\
                     \r\n\
                     Server log data here";

        let parsed = parse_email(raw).unwrap();
        assert!(
            parsed.body_text.is_none(),
            "text/plain with disposition:attachment must not populate body_text"
        );
        assert_eq!(parsed.attachments.len(), 1);
        assert_eq!(parsed.attachments[0].content_type, "text/plain");
        assert_eq!(parsed.attachments[0].filename.as_deref(), Some("log.txt"));
    }

    #[test]
    fn parse_group_address_empty_undisclosed() {
        // RFC 5322 Section 3.4: group = display-name ":" [group-list] ";"
        // "undisclosed-recipients:;" is a common empty group that must not
        // produce malformed addresses or cause a parse failure.
        let raw = b"From: a@b.com\r\n\
                     To: undisclosed-recipients:;\r\n\
                     Date: Thu, 13 Feb 2025 15:47:33 +0000\r\n\
                     \r\n";

        let parsed = parse_email(raw).unwrap();
        // Empty group — no addresses should be extracted
        assert!(
            parsed.to.is_empty(),
            "empty group undisclosed-recipients:; must produce no addresses, got {:?}",
            parsed.to
        );
    }

    #[test]
    fn parse_group_address_with_members() {
        // RFC 5322 Section 3.4: group with member addresses
        // "friends:a@b.com, c@d.com;" should extract the member addresses.
        let raw = b"From: a@b.com\r\n\
                     To: friends:one@x.com, two@x.com;\r\n\
                     Date: Thu, 13 Feb 2025 15:47:33 +0000\r\n\
                     \r\n";

        let parsed = parse_email(raw).unwrap();
        assert_eq!(
            parsed.to.len(),
            2,
            "group with 2 members must produce 2 addresses, got {:?}",
            parsed.to
        );
        assert_eq!(parsed.to[0].email, "one@x.com");
        assert_eq!(parsed.to[1].email, "two@x.com");
    }

    #[test]
    fn parse_group_address_mixed_with_regular() {
        // Mix of regular addresses and group syntax in same header.
        let raw = b"From: a@b.com\r\n\
                     To: solo@x.com, friends:one@x.com, two@x.com;, last@x.com\r\n\
                     Date: Thu, 13 Feb 2025 15:47:33 +0000\r\n\
                     \r\n";

        let parsed = parse_email(raw).unwrap();
        let emails: Vec<&str> = parsed.to.iter().map(|a| a.email.as_str()).collect();
        assert_eq!(
            emails,
            vec!["solo@x.com", "one@x.com", "two@x.com", "last@x.com"],
            "must extract all 4 addresses from mixed regular+group syntax"
        );
    }

    #[test]
    fn decode_qp_trailing_equals_is_soft_break() {
        // RFC 2045 Section 6.7: a trailing '=' at end-of-data is a soft line
        // break and must be removed, not emitted as a literal '=' character.
        let result = decode_quoted_printable(b"Hello=");
        assert_eq!(
            result, b"Hello",
            "trailing '=' must be treated as soft line break per RFC 2045 Section 6.7"
        );
    }

    #[test]
    fn decode_qp_trailing_equals_cr_is_soft_break() {
        // '=\r' at end-of-data (bare CR without LF) should also be removed
        // as a soft line break.
        let result = decode_quoted_printable(b"Hello=\r");
        assert_eq!(
            result, b"Hello",
            "trailing '=\\r' must be treated as soft line break"
        );
    }

    #[test]
    fn parse_bare_address_with_trailing_comment() {
        // RFC 5322 Section 3.4.1: addr-spec can be followed by CFWS.
        // RFC 5322 Section 3.2.2: parenthesized text is a comment.
        // A trailing comment like "(Display Name)" after a bare address
        // must be stripped from the email and used as the display name.
        let raw = b"From: sender@example.com\r\n\
                     To: user@example.com (Display Name)\r\n\
                     Date: Thu, 13 Feb 2025 15:47:33 +0000\r\n\
                     \r\n";

        let parsed = parse_email(raw).unwrap();
        assert_eq!(parsed.to.len(), 1);
        assert_eq!(
            parsed.to[0].email, "user@example.com",
            "email must not contain the trailing comment"
        );
        assert_eq!(
            parsed.to[0].name.as_deref(),
            Some("Display Name"),
            "trailing comment should become display name per RFC 5322 Section 3.4.1"
        );
    }

    #[test]
    fn parse_bare_address_with_leading_comment() {
        // RFC 5322 Section 3.4.1: CFWS can appear before addr-spec.
        // A leading comment should be stripped but NOT become a display name,
        // since it precedes the address rather than following it.
        let raw = b"From: sender@example.com\r\n\
                     To: (Comment) user@example.com\r\n\
                     Date: Thu, 13 Feb 2025 15:47:33 +0000\r\n\
                     \r\n";

        let parsed = parse_email(raw).unwrap();
        assert_eq!(parsed.to.len(), 1);
        assert_eq!(
            parsed.to[0].email, "user@example.com",
            "email must not contain the leading comment"
        );
        assert_eq!(
            parsed.to[0].name, None,
            "leading comment must not become display name"
        );
    }

    #[test]
    fn extract_param_skips_quoted_values() {
        // Parameter name appearing inside another parameter's quoted value
        // should not be matched (RFC 5322 Section 3.2.4).
        let header = "text/html; boundary=\"has charset=bad inside\"; charset=utf-8";
        let charset = extract_param(header, "charset");
        assert_eq!(
            charset.as_deref(),
            Some("utf-8"),
            "Should skip match inside quoted boundary value"
        );
    }

    #[test]
    fn multipart_part_without_charset_uses_us_ascii_default() {
        // RFC 2045 Section 5.2: default Content-Type is text/plain; charset=us-ascii.
        // A MIME part with text/plain but no charset parameter should default
        // to US-ASCII, consistent with the top-level default.
        //
        // Windows-1252 byte 0x93 is a left double quotation mark (U+201C).
        // encoding_rs maps us-ascii → Windows-1252, so 0x93 decodes to U+201C.
        // Under UTF-8, 0x93 is an invalid byte and produces U+FFFD.
        let raw = b"From: a@b.com\r\n\
                     Date: Thu, 13 Feb 2025 15:47:33 +0000\r\n\
                     Content-Type: multipart/mixed; boundary=\"b\"\r\n\
                     \r\n\
                     --b\r\n\
                     Content-Type: text/plain\r\n\
                     \r\n\
                     Hello \x93World\r\n\
                     --b--";

        let parsed = parse_email(raw).unwrap();
        let text = parsed.body_text.unwrap();

        // With the correct US-ASCII default (mapped to Windows-1252 by encoding_rs),
        // 0x93 should decode to U+201C, not the replacement character U+FFFD.
        assert!(
            text.contains('\u{201c}'),
            "Part with text/plain (no charset) should use US-ASCII default per \
             RFC 2045 Section 5.2, decoding 0x93 as U+201C. Got: {text:?}"
        );
        assert!(
            !text.contains('\u{FFFD}'),
            "Part with text/plain (no charset) should not produce UTF-8 replacement \
             characters. Got: {text:?}"
        );
    }

    /// header unfolding must preserve trailing whitespace from
    /// the first line when followed by a continuation line.
    ///
    /// RFC 5322 Section 2.2.3: "Unfolding is accomplished by simply removing
    /// any CRLF that is immediately followed by WSP." Only the CRLF is
    /// removed; all other whitespace (including trailing spaces on the first
    /// line) must be preserved.
    #[test]
    fn parse_header_unfold_preserves_trailing_whitespace() {
        // "Subject: Hello  \r\n World" should unfold to "Hello   World"
        // (2 trailing spaces from first line + 1 leading space from continuation = 3 spaces)
        let raw = b"From: a@b.com\r\nSubject: Hello  \r\n World\r\nDate: Thu, 13 Feb 2025 15:47:33 +0000\r\n\r\n";
        let parsed = parse_email(raw).unwrap();
        assert_eq!(
            parsed.subject.as_deref(),
            Some("Hello   World"),
            "Trailing whitespace on first line must be preserved during unfolding \
             (RFC 5322 Section 2.2.3)"
        );
    }

    /// Single-part body text must not include the trailing CRLF
    /// that the message format requires after the body content. The trailing
    /// CRLF is a format artifact (RFC 5322 Section 3.5), not semantic content.
    /// Without this fix, build→parse round-trip adds a spurious "\r\n" to the
    /// body text of single-part messages.
    #[test]
    fn parse_single_part_body_no_trailing_crlf() {
        // Single-part message: body ends with \r\n (format artifact)
        let raw = b"From: a@b.com\r\n\
                     Date: Thu, 13 Feb 2025 15:47:33 +0000\r\n\
                     Content-Type: text/plain; charset=utf-8\r\n\
                     \r\n\
                     Hello, World!\r\n";

        let parsed = parse_email(raw).unwrap();
        assert_eq!(
            parsed.body_text.as_deref(),
            Some("Hello, World!"),
            "Single-part body text must not include trailing CRLF"
        );
    }

    /// Single-part HTML body must not include trailing CRLF.
    #[test]
    fn parse_single_part_html_no_trailing_crlf() {
        let raw = b"From: a@b.com\r\n\
                     Date: Thu, 13 Feb 2025 15:47:33 +0000\r\n\
                     Content-Type: text/html; charset=utf-8\r\n\
                     \r\n\
                     <p>Hello</p>\r\n";

        let parsed = parse_email(raw).unwrap();
        assert_eq!(
            parsed.body_html.as_deref(),
            Some("<p>Hello</p>"),
            "Single-part HTML body must not include trailing CRLF"
        );
    }

    /// Build→parse round-trip for single-part text must preserve body exactly.
    #[test]
    fn round_trip_single_part_body_text() {
        let email = crate::types::OutgoingEmail {
            from: crate::types::Address {
                name: None,
                email: "a@b.com".into(),
            },
            to: vec![crate::types::Address {
                name: None,
                email: "c@d.com".into(),
            }],
            cc: vec![],
            bcc: vec![],
            reply_to: None,
            subject: "Test".into(),
            body_text: Some("Hello, World!".into()),
            body_html: None,
            in_reply_to: None,
            references: None,
            attachments: vec![],
        };

        let built = crate::build_message(&email).unwrap();
        let parsed = parse_email(&built.raw).unwrap();
        assert_eq!(
            parsed.body_text.as_deref(),
            Some("Hello, World!"),
            "Single-part body text must round-trip without trailing CRLF"
        );
    }

    /// RFC 2047 encoded words in display names must be decoded
    /// AFTER address structure parsing, not before. Decoding before parsing
    /// breaks address splitting when the decoded text contains address-
    /// significant characters like commas.
    ///
    /// RFC 2047 Section 5 rule (3): encoded-words may appear in a 'phrase'
    /// (e.g., display name) within address headers. The encoded word
    /// `=?UTF-8?B?Sm9obiwgRG9l?=` decodes to `John, Doe`. If decoded
    /// before address parsing, the comma splits the address incorrectly.
    #[test]
    fn parse_encoded_word_display_name_with_comma() {
        // "John, Doe" base64-encoded as an RFC 2047 encoded word.
        // The comma must NOT split the address — it's part of the name.
        let raw = b"From: =?UTF-8?B?Sm9obiwgRG9l?= <john@example.com>\r\n\
                     Date: Thu, 13 Feb 2025 15:47:33 +0000\r\n\
                     \r\n";

        let parsed = parse_email(raw).unwrap();
        assert_eq!(
            parsed.from.name.as_deref(),
            Some("John, Doe"),
            "RFC 2047 encoded display name with comma must be preserved \
             (RFC 2047 Section 5 rule 3): decode AFTER address parsing"
        );
        assert_eq!(parsed.from.email, "john@example.com");
    }

    /// RFC 2045 Section 6.8: "Any characters outside of the base64 alphabet
    /// are to be ignored in base64-encoded data." The base64 alphabet is
    /// A-Z, a-z, 0-9, +, /, = (padding). Stray non-alphabet characters such
    /// as `!`, `#`, `~` must be stripped before decoding, not just whitespace.
    #[test]
    fn parse_base64_body_ignores_non_alphabet_chars() {
        // "Hello World" = "SGVsbG8gV29ybGQ=" in base64.
        // Insert non-base64 characters (!, #, ~) that are NOT whitespace
        // to verify the decoder strips all non-alphabet bytes.
        let raw = b"From: a@b.com\r\n\
                     Date: Thu, 13 Feb 2025 15:47:33 +0000\r\n\
                     Content-Type: text/plain; charset=utf-8\r\n\
                     Content-Transfer-Encoding: base64\r\n\
                     \r\n\
                     SGVs!bG8#gV29~ybGQ=";

        let parsed = parse_email(raw).unwrap();
        assert_eq!(
            parsed.body_text.as_deref(),
            Some("Hello World"),
            "RFC 2045 Section 6.8: non-alphabet characters must be ignored in base64 data"
        );
    }

    /// Same as above but for To/Cc headers with multiple recipients.
    /// The encoded comma must not create a spurious address split.
    #[test]
    fn parse_encoded_word_display_name_with_comma_in_to() {
        // Two recipients: first has an encoded comma in the display name,
        // second is a plain address. Must parse as exactly 2 addresses.
        let raw = b"From: sender@example.com\r\n\
                     To: =?UTF-8?B?Sm9obiwgRG9l?= <john@example.com>, other@example.com\r\n\
                     Date: Thu, 13 Feb 2025 15:47:33 +0000\r\n\
                     \r\n";

        let parsed = parse_email(raw).unwrap();
        assert_eq!(
            parsed.to.len(),
            2,
            "Must parse exactly 2 addresses, not 3 (encoded comma is not a separator)"
        );
        assert_eq!(
            parsed.to[0].name.as_deref(),
            Some("John, Doe"),
            "First recipient display name must be 'John, Doe'"
        );
        assert_eq!(parsed.to[0].email, "john@example.com");
        assert_eq!(parsed.to[1].email, "other@example.com");
    }

    /// building a message with `body_text: None` (and attachments,
    /// which forces multipart/mixed with an empty text/plain part) and then
    /// parsing it back must yield `body_text: None`, not `Some("")`.
    ///
    /// The builder's `write_text_part` always appends a trailing `\r\n` after
    /// the body content (RFC 2046 Section 5.1.1 requires CRLF before boundary).
    /// When the body is empty, the MIME part body becomes just `\r\n`, which
    /// `decode_body` strips to `""`. The parser must treat an empty decoded
    /// body as absent (`None`), not present-but-empty (`Some("")`).
    ///
    /// # References
    /// - RFC 2046 Section 5.1.1 (CRLF before boundary delimiter)
    /// - RFC 5322 Section 3.5 (message body)
    #[test]
    fn round_trip_empty_body_text_is_none() {
        // Use attachments to force multipart/mixed, which causes the builder
        // to emit an empty text/plain part when body_text is None.
        let email = crate::types::OutgoingEmail {
            from: crate::types::Address {
                name: None,
                email: "a@b.com".into(),
            },
            to: vec![crate::types::Address {
                name: None,
                email: "c@d.com".into(),
            }],
            cc: vec![],
            bcc: vec![],
            reply_to: None,
            subject: "Empty body".into(),
            body_text: None,
            body_html: None,
            in_reply_to: None,
            references: None,
            attachments: vec![crate::types::OutgoingAttachment {
                filename: "test.txt".into(),
                content_type: "text/plain".into(),
                data: b"attachment data".to_vec(),
            }],
        };

        let built = crate::build_message(&email).unwrap();
        let parsed = parse_email(&built.raw).unwrap();

        assert_eq!(
            parsed.body_text, None,
            "Empty body_text must round-trip as None, not Some(\"\")"
        );
    }

    /// Same fix for multipart/alternative: `body_text: Some("")`
    /// should parse back as `None` since the decoded content is empty.
    ///
    /// # References
    /// - RFC 2046 Section 5.1.1 (CRLF before boundary delimiter)
    #[test]
    fn round_trip_empty_body_html_in_alternative_is_none() {
        // Both body_text and body_html present forces multipart/alternative.
        // An explicitly empty body_html should parse back as None.
        let email = crate::types::OutgoingEmail {
            from: crate::types::Address {
                name: None,
                email: "a@b.com".into(),
            },
            to: vec![crate::types::Address {
                name: None,
                email: "c@d.com".into(),
            }],
            cc: vec![],
            bcc: vec![],
            reply_to: None,
            subject: "Text only".into(),
            body_text: Some("Plain text".into()),
            body_html: Some(String::new()),
            in_reply_to: None,
            references: None,
            attachments: vec![],
        };

        let built = crate::build_message(&email).unwrap();
        let parsed = parse_email(&built.raw).unwrap();

        assert_eq!(
            parsed.body_html, None,
            "Empty body_html must parse as None, not Some(\"\")"
        );
        assert_eq!(
            parsed.body_text.as_deref(),
            Some("Plain text"),
            "body_text must be preserved"
        );
    }

    #[test]
    fn extract_mime_type_strips_rfc5322_comments() {
        // RFC 5322 Section 3.2.2 parenthesized comments may appear
        // in CFWS positions within the Content-Type type/subtype production.
        // extract_mime_type must strip them so that "text/plain (comment)"
        // normalizes to "text/plain", not "text/plain (comment)".
        //
        // Single-part message: the comment in Content-Type caused the MIME
        // comparison to fail, making the parser treat the body as an
        // attachment instead of body_text.
        let raw = b"From: a@b.com\r\n\
                     Date: Thu, 13 Feb 2025 15:47:33 +0000\r\n\
                     Content-Type: text/plain (this is a comment); charset=utf-8\r\n\
                     \r\n\
                     Hello with comment";

        let parsed = parse_email(raw).unwrap();
        assert_eq!(
            parsed.body_text.as_deref(),
            Some("Hello with comment"),
            "Body must be extracted as body_text when Content-Type has an RFC 5322 comment"
        );
        assert!(
            parsed.attachments.is_empty(),
            "No attachments expected for a plain text/plain message with a comment"
        );
    }

    /// RFC 2046 Section 5.1.5: In a multipart/digest, the default Content-Type
    /// for body parts is "message/rfc822", NOT "text/plain; charset=us-ascii".
    /// Parts without an explicit Content-Type header must be treated as
    /// message/rfc822 and show up as attachments, not `body_text`.
    #[test]
    fn multipart_digest_default_content_type_is_message_rfc822() {
        let raw = b"From: sender@example.com\r\n\
                     Date: Thu, 13 Feb 2025 15:47:33 +0000\r\n\
                     Subject: Digest\r\n\
                     Content-Type: multipart/digest; boundary=\"digestboundary\"\r\n\
                     \r\n\
                     --digestboundary\r\n\
                     \r\n\
                     From: nested@example.com\r\n\
                     Subject: Nested message\r\n\
                     \r\n\
                     Nested body text\r\n\
                     --digestboundary--\r\n";

        let parsed = parse_email(raw).unwrap();

        // The part has no Content-Type header. In multipart/digest, this means
        // message/rfc822 per RFC 2046 Section 5.1.5 — it should appear as an
        // attachment, NOT as body_text.
        assert!(
            parsed.body_text.is_none(),
            "multipart/digest parts without Content-Type should default to \
             message/rfc822, not text/plain — body_text should be None"
        );
        assert_eq!(
            parsed.attachments.len(),
            1,
            "multipart/digest part should be treated as message/rfc822 attachment"
        );
        assert_eq!(
            parsed.attachments[0].content_type, "message/rfc822",
            "default Content-Type in multipart/digest must be message/rfc822 \
             (RFC 2046 Section 5.1.5)"
        );
    }

    /// Content-ID with whitespace inside angle brackets must be
    /// trimmed after bracket stripping.
    ///
    /// RFC 2392 defines Content-ID as `"<" addr-spec ">"`. Some mailers
    /// add whitespace around the addr-spec inside the brackets. After
    /// stripping `<` and `>`, the result must be trimmed to produce a
    /// clean identifier for matching (e.g., for CID references in HTML).
    #[test]
    fn content_id_whitespace_inside_brackets_trimmed() {
        // Multipart path (walk_mime_tree)
        let raw = b"From: a@b.com\r\n\
                     Date: Thu, 13 Feb 2025 15:47:33 +0000\r\n\
                     Content-Type: multipart/mixed; boundary=\"b\"\r\n\
                     \r\n\
                     --b\r\n\
                     Content-Type: text/plain\r\n\
                     \r\n\
                     Body\r\n\
                     --b\r\n\
                     Content-Type: image/png\r\n\
                     Content-ID: < cid@example.com >\r\n\
                     \r\n\
                     PNG\r\n\
                     --b--";
        let parsed = parse_email(raw).unwrap();
        assert_eq!(
            parsed.attachments[0].content_id.as_deref(),
            Some("cid@example.com"),
            "Content-ID must be trimmed after bracket stripping (RFC 2392)"
        );

        // Simple body path (extract_simple_body)
        let raw_single = b"From: a@b.com\r\n\
                           Date: Thu, 13 Feb 2025 15:47:33 +0000\r\n\
                           Content-Type: image/png\r\n\
                           Content-ID: < cid2@example.com >\r\n\
                           \r\n\
                           PNG";
        let parsed_single = parse_email(raw_single).unwrap();
        assert_eq!(
            parsed_single.attachments[0].content_id.as_deref(),
            Some("cid2@example.com"),
            "Content-ID in single-part message must be trimmed (RFC 2392)"
        );
    }

    #[test]
    fn parse_headers_only_all_fields_verified() {
        // Verify that parse_headers_only extracts ALL header fields correctly,
        // including cc, bcc, reply_to, in_reply_to, and references.
        let raw = b"From: sender@example.com\r\n\
                     To: to@example.com\r\n\
                     Cc: cc@example.com\r\n\
                     Bcc: bcc@example.com\r\n\
                     Reply-To: reply@example.com\r\n\
                     Subject: Full test\r\n\
                     Date: Thu, 13 Feb 2025 15:47:33 +0000\r\n\
                     Message-ID: <msg1@example.com>\r\n\
                     In-Reply-To: <parent@example.com>\r\n\
                     References: <ref1@example.com> <ref2@example.com>\r\n\
                     \r\n\
                     Body that should be ignored";

        let parsed = parse_headers_only(raw).unwrap();

        assert_eq!(parsed.from.email, "sender@example.com");
        assert_eq!(parsed.to.len(), 1);
        assert_eq!(parsed.to[0].email, "to@example.com");
        assert_eq!(parsed.cc.len(), 1);
        assert_eq!(parsed.cc[0].email, "cc@example.com");
        assert_eq!(parsed.bcc.len(), 1);
        assert_eq!(parsed.bcc[0].email, "bcc@example.com");
        assert_eq!(parsed.reply_to.len(), 1);
        assert_eq!(parsed.reply_to[0].email, "reply@example.com");
        assert_eq!(parsed.subject.as_deref(), Some("Full test"));
        assert!(parsed.date.is_some());
        assert_eq!(parsed.message_id.as_deref(), Some("msg1@example.com"));
        assert_eq!(parsed.in_reply_to.as_deref(), Some("parent@example.com"));
        assert_eq!(
            parsed.references.as_deref(),
            Some("ref1@example.com ref2@example.com")
        );

        // Body fields must be empty
        assert!(parsed.body_text.is_none());
        assert!(parsed.body_html.is_none());
        assert!(parsed.attachments.is_empty());
    }

    #[test]
    fn parse_missing_optional_headers_returns_none_or_empty() {
        // Only From is required (RFC 5322 Section 3.6.2). All other fields
        // should gracefully return None or empty when absent.
        let raw = b"From: a@b.com\r\n\r\n";

        let parsed = parse_email(raw).unwrap();

        assert_eq!(parsed.from.email, "a@b.com");
        assert!(parsed.subject.is_none());
        assert!(parsed.date.is_none());
        assert!(parsed.message_id.is_none());
        assert!(parsed.in_reply_to.is_none());
        assert!(parsed.references.is_none());
        assert!(parsed.to.is_empty());
        assert!(parsed.cc.is_empty());
        assert!(parsed.bcc.is_empty());
        assert!(parsed.reply_to.is_empty());
    }

    #[test]
    fn extract_param_rejects_substring_match() {
        // "filename=" must not match inside "xfilename=" — the boundary check
        // requires `;`, space, tab, or start-of-string before the param name.
        let value = "attachment; xfilename=\"bad.pdf\"; filename=\"good.pdf\"";
        let result = extract_param(value, "filename");
        assert_eq!(
            result.as_deref(),
            Some("good.pdf"),
            "Must not match xfilename as filename"
        );
    }

    #[test]
    fn extract_param_rejects_suffix_only_match() {
        // Edge case: param name appears only as suffix of another param name.
        let value = "attachment; notfilename=\"only.pdf\"";
        let result = extract_param(value, "filename");
        assert!(
            result.is_none(),
            "Must not match 'filename' inside 'notfilename'"
        );
    }

    // -----------------------------------------------------------------------
    // Audit coverage: Group address parsing (RFC 5322 Section 3.4)
    // -----------------------------------------------------------------------

    /// RFC 5322 Section 3.4: empty group `undisclosed-recipients:;`
    /// must produce zero addresses.
    #[test]
    fn parse_group_address_empty() {
        let addrs = parse_address_list("undisclosed-recipients:;");
        assert!(
            addrs.is_empty(),
            "empty group must produce no addresses, got {addrs:?}"
        );
    }

    /// RFC 5322 Section 3.4: group with members — audit coverage.
    #[test]
    fn parse_group_address_with_two_members() {
        let addrs = parse_address_list("Friends: a@x.com, b@x.com;");
        assert_eq!(addrs.len(), 2, "group with 2 members: {addrs:?}");
        assert_eq!(addrs[0].email, "a@x.com");
        assert_eq!(addrs[1].email, "b@x.com");
    }

    /// RFC 5322 Section 3.4: multiple groups in a single header value.
    #[test]
    fn parse_multiple_groups_and_solo() {
        let addrs =
            parse_address_list("Team A: a1@x.com, a2@x.com;, Team B: b1@x.com;, solo@x.com");
        assert_eq!(addrs.len(), 4, "2 groups + 1 solo: {addrs:?}");
        assert_eq!(addrs[0].email, "a1@x.com");
        assert_eq!(addrs[1].email, "a2@x.com");
        assert_eq!(addrs[2].email, "b1@x.com");
        assert_eq!(addrs[3].email, "solo@x.com");
    }

    /// Address with parenthesized comment containing commas must not
    /// split on those commas (RFC 5322 Section 3.2.2) — audit coverage.
    #[test]
    fn parse_address_comment_with_comma_audit() {
        let addrs = parse_address_list("user@x.com (Last, First), other@x.com");
        assert_eq!(
            addrs.len(),
            2,
            "comma inside comment must not split: {addrs:?}"
        );
        assert_eq!(addrs[0].email, "user@x.com");
        assert_eq!(addrs[1].email, "other@x.com");
    }

    // -----------------------------------------------------------------------
    // Audit coverage: RFC 2231 Section 3 — continuation edge cases
    // -----------------------------------------------------------------------

    /// RFC 2231 Section 3: continuation parameters with a gap (e.g., *0 then
    /// *2 skipping *1) — should stop at the gap.
    #[test]
    fn rfc2231_continuation_gap_stops() {
        let header = "attachment; filename*0=\"hello\"; filename*2=\"skipped\"";
        // Should only find section 0, not section 2
        let result = extract_rfc2231_continuation(header, "filename");
        assert_eq!(
            result.as_deref(),
            Some("hello"),
            "continuation must stop at missing section index"
        );
    }

    /// RFC 2231 Section 3: single continuation section produces the value.
    #[test]
    fn rfc2231_continuation_single_section() {
        let header = "attachment; filename*0=\"report.pdf\"";
        let result = extract_rfc2231_continuation(header, "filename");
        assert_eq!(result.as_deref(), Some("report.pdf"));
    }

    /// RFC 2231 Section 4 + RFC 2045 Section 5.2: when no charset is declared
    /// in continuation parameters, the default per the RFCs would be US-ASCII.
    /// We intentionally default to UTF-8 as a Postel's law accommodation (see
    /// the inline comment in `extract_rfc2231_continuation`). This test verifies
    /// that plain (non-charset-encoded) continuation parameters without any
    /// charset declaration decode correctly — ASCII values must round-trip
    /// identically since US-ASCII is a strict subset of UTF-8.
    #[test]
    fn rfc2231_continuation_no_charset_defaults_to_utf8() {
        // Plain continuation sections (no `*` suffix, no charset declaration).
        // The parser must reassemble them using the UTF-8 default.
        let raw = b"From: a@b.com\r\n\
                     Date: Thu, 13 Feb 2025 15:47:33 +0000\r\n\
                     Content-Type: multipart/mixed; boundary=\"b\"\r\n\
                     \r\n\
                     --b\r\n\
                     Content-Type: text/plain\r\n\
                     \r\n\
                     Body\r\n\
                     --b\r\n\
                     Content-Type: application/octet-stream\r\n\
                     Content-Disposition: attachment; filename*0=\"annual_\"; filename*1=\"report_\"; filename*2=\"2025.pdf\"\r\n\
                     \r\n\
                     DATA\r\n\
                     --b--";

        let parsed = parse_email(raw).unwrap();
        assert_eq!(parsed.attachments.len(), 1);
        assert_eq!(
            parsed.attachments[0].filename.as_deref(),
            Some("annual_report_2025.pdf"),
            "RFC 2231 continuation without charset should decode as UTF-8"
        );
    }

    // -----------------------------------------------------------------------
    // Audit coverage: Base64 with embedded whitespace (RFC 2045 Section 6.8)
    // -----------------------------------------------------------------------

    /// RFC 2045 Section 6.8: "Any characters outside of the base64 alphabet
    /// are to be ignored in base64-encoded data." Spaces/tabs within base64
    /// lines must be stripped before decoding.
    #[test]
    fn base64_with_embedded_spaces() {
        // "Hello World" = SGVsbG8gV29ybGQ=, with spaces injected
        let data = b"SGVs bG8g V29y bGQ=";
        let decoded = decode_transfer_encoding(data, "base64");
        assert_eq!(
            std::str::from_utf8(&decoded).unwrap(),
            "Hello World",
            "base64 decoder must strip non-alphabet characters (RFC 2045 Section 6.8)"
        );
    }

    /// RFC 2045 Section 6.8: tabs within base64-encoded data must be ignored.
    #[test]
    fn base64_with_tabs() {
        let data = b"SGVs\tbG8g\tV29ybGQ=";
        let decoded = decode_transfer_encoding(data, "base64");
        assert_eq!(
            std::str::from_utf8(&decoded).unwrap(),
            "Hello World",
            "base64 decoder must strip tabs (RFC 2045 Section 6.8)"
        );
    }

    // -----------------------------------------------------------------------
    // Audit coverage: Quoted-printable edge cases (RFC 2045 Section 6.7)
    // -----------------------------------------------------------------------

    /// RFC 2045 Section 6.7: trailing `=` at end of data is a soft line
    /// break — it should be stripped, not produce a literal `=`.
    #[test]
    fn qp_trailing_equals_stripped() {
        let data = b"Hello=";
        let decoded = decode_quoted_printable(data);
        assert_eq!(
            std::str::from_utf8(&decoded).unwrap(),
            "Hello",
            "trailing '=' is a soft break (RFC 2045 Section 6.7)"
        );
    }

    /// RFC 2045 Section 6.7: `=` followed by non-hex characters is malformed.
    /// Postel's law: pass through the literal `=` and the following characters.
    #[test]
    fn qp_malformed_hex_passthrough() {
        let data = b"Hello=ZZ World";
        let decoded = decode_quoted_printable(data);
        assert_eq!(
            std::str::from_utf8(&decoded).unwrap(),
            "Hello=ZZ World",
            "malformed =ZZ must pass through literally (Postel's law)"
        );
    }

    // -----------------------------------------------------------------------
    // Audit coverage: RFC 2047 Q-encoding edge cases
    // -----------------------------------------------------------------------

    /// RFC 2047 Section 4.2: `=` followed by non-hex in Q-encoding should
    /// pass through the literal `=` (Postel's law — decoders accept gracefully).
    #[test]
    fn q_encoding_malformed_hex_passthrough() {
        let decoded = decode_q_encoding("Hello=ZZWorld");
        assert_eq!(
            std::str::from_utf8(&decoded).unwrap(),
            "Hello=ZZWorld",
            "malformed =ZZ in Q-encoding must pass through literally"
        );
    }

    /// RFC 2047 Section 4.2: `=` at end of Q-encoded text with no following
    /// hex digits should pass through as literal.
    #[test]
    fn q_encoding_trailing_equals() {
        let decoded = decode_q_encoding("Hello=");
        assert_eq!(
            std::str::from_utf8(&decoded).unwrap(),
            "Hello=",
            "trailing '=' in Q-encoding must pass through literally"
        );
    }

    // -----------------------------------------------------------------------
    // Audit coverage: multipart/digest default Content-Type
    // -----------------------------------------------------------------------

    /// RFC 2046 Section 5.1.5: parts inside multipart/digest without an
    /// explicit Content-Type must default to message/rfc822 (NOT text/plain).
    /// Verify via a full `parse_email` round-trip.
    #[test]
    fn multipart_digest_default_content_type_full_email() {
        let raw = b"From: a@b.com\r\n\
                     Date: Thu, 13 Feb 2025 15:47:33 +0000\r\n\
                     Content-Type: multipart/digest; boundary=\"dg\"\r\n\
                     \r\n\
                     --dg\r\n\
                     \r\n\
                     From: nested@example.com\r\n\
                     Subject: Inner\r\n\
                     \r\n\
                     Inner body\r\n\
                     --dg--";

        let parsed = parse_email(raw).unwrap();
        // Part has no Content-Type → default is message/rfc822 in digest
        assert!(
            parsed.body_text.is_none(),
            "digest part must NOT be treated as text/plain"
        );
        assert_eq!(parsed.attachments.len(), 1);
        assert_eq!(parsed.attachments[0].content_type, "message/rfc822");
    }

    // -----------------------------------------------------------------------
    // Coverage: header line starting with space (L208)
    // -----------------------------------------------------------------------

    /// RFC 5322 Section 2.2.3: a raw message starting with a space (a
    /// continuation line before any header name) should be silently
    /// skipped by the header parser, not crash or produce garbage.
    #[test]
    fn parse_headers_leading_space_skipped() {
        // The first line starts with a space, which is a continuation line
        // with no preceding header name. It must be silently ignored.
        let raw = b" continuation without header\r\nFrom: a@b.com\r\nDate: Thu, 13 Feb 2025 15:47:33 +0000\r\n\r\nBody";
        let parsed = parse_email(raw).unwrap();
        assert_eq!(parsed.from.email, "a@b.com");
        assert_eq!(parsed.body_text.as_deref(), Some("Body"));
    }

    /// A message starting with a tab (another form of continuation WSP)
    /// before any header name must also be handled gracefully.
    #[test]
    fn parse_headers_leading_tab_skipped() {
        let raw = b"\tcontinuation without header\r\nFrom: a@b.com\r\nDate: Thu, 13 Feb 2025 15:47:33 +0000\r\n\r\n";
        let parsed = parse_email(raw).unwrap();
        assert_eq!(parsed.from.email, "a@b.com");
    }

    // -----------------------------------------------------------------------
    // Coverage: RFC 2047 decoding edge cases (L491, L542, L550, L558-560)
    // -----------------------------------------------------------------------

    /// RFC 2047 Section 2: an encoded word with a bad base64 payload
    /// should fail decoding and the literal `=?...?=` token should pass
    /// through (Postel's law).
    #[test]
    fn encoded_word_bad_base64_passthrough() {
        // "=====" is not valid base64
        let input = "=?UTF-8?B?=====?=";
        let result = decode_encoded_words(input);
        // Should pass through as literal since base64 decode fails
        assert!(
            result.contains("=?"),
            "Bad base64 encoded word should pass through literally, got: {result:?}"
        );
    }

    /// RFC 2047 Section 2: an encoded word with an unknown charset should
    /// still decode — `decode_charset` falls back to UTF-8 via `encoding_rs`.
    #[test]
    fn encoded_word_unknown_charset_fallback() {
        // "Hello" in base64 is "SGVsbG8="
        let input = "=?x-nonexistent-charset?B?SGVsbG8=?=";
        let result = decode_encoded_words(input);
        // encoding_rs falls back to UTF-8 for unknown charsets, so this
        // should decode the base64 payload and return something containing "Hello"
        assert!(
            result.contains("Hello"),
            "Unknown charset should fall back to UTF-8, got: {result:?}"
        );
    }

    /// RFC 2047: a truncated encoded word (missing closing `?=`) should
    /// emit the literal `=?` prefix and continue.
    #[test]
    fn encoded_word_truncated_no_closing() {
        let input = "Start =?UTF-8?B?SGVsbG8= End";
        let result = decode_encoded_words(input);
        // The token is missing the closing "?=", so it cannot be decoded.
        // The "=?" should be emitted literally and parsing continues.
        assert!(
            result.contains("=?"),
            "Truncated encoded word should pass through, got: {result:?}"
        );
    }

    // -----------------------------------------------------------------------
    // Coverage: RFC 2231 parameter continuation (L586, L592, L594)
    // -----------------------------------------------------------------------

    /// RFC 2231 Section 3: continuation parameters with mixed encoded
    /// and plain sections. `name*0*=charset'lang'...; name*1=plain`
    #[test]
    fn rfc2231_continuation_mixed_encoded_and_plain() {
        let header = "attachment; filename*0*=UTF-8''r%C3%A9sum; filename*1=\"e.pdf\"";
        let result = extract_rfc2231_continuation(header, "filename");
        // Section 0 is encoded: r%C3%A9sum → "résum" (UTF-8)
        // Section 1 is plain: "e.pdf"
        assert_eq!(
            result.as_deref(),
            Some("r\u{e9}sume.pdf"),
            "RFC 2231 mixed encoded/plain continuation should reassemble correctly"
        );
    }

    /// RFC 2231 Section 3: continuation with three sections.
    #[test]
    fn rfc2231_continuation_three_sections() {
        let header =
            "attachment; filename*0=\"part1_\"; filename*1=\"part2_\"; filename*2=\"part3.pdf\"";
        let result = extract_rfc2231_continuation(header, "filename");
        assert_eq!(result.as_deref(), Some("part1_part2_part3.pdf"));
    }

    // -----------------------------------------------------------------------
    // Coverage: extract_quoted_value / quoted-string edge cases
    // (L605, L614-616, L619-621, L625, L634, L636, L647)
    // -----------------------------------------------------------------------

    /// RFC 5322 Section 3.2.4: an unterminated quoted-string (no closing
    /// `"`) in a parameter value — the parser should read to end of string.
    #[test]
    fn extract_param_unterminated_quoted_value() {
        let header = "text/plain; charset=\"utf-8";
        let result = extract_param(header, "charset");
        // The closing quote is missing; find_closing_quote returns s.len(),
        // so the entire remainder becomes the value.
        assert_eq!(
            result.as_deref(),
            Some("utf-8"),
            "Unterminated quoted-string should extract to end of string"
        );
    }

    /// RFC 5322 Section 3.2.4: quoted-string with escaped characters.
    #[test]
    fn extract_param_quoted_with_backslash_escape() {
        let header = "attachment; filename=\"file\\\\name.txt\"";
        let result = extract_param(header, "filename");
        assert_eq!(
            result.as_deref(),
            Some("file\\name.txt"),
            "Backslash escape in quoted param value must be unescaped"
        );
    }

    /// RFC 5322 Section 3.2.4: empty quoted-string should return None.
    #[test]
    fn extract_param_empty_quoted_value() {
        let header = "attachment; filename=\"\"";
        let result = extract_param(header, "filename");
        assert!(
            result.is_none(),
            "Empty quoted-string value should return None, got: {result:?}"
        );
    }

    /// `extract_comment_text` with nested parentheses (RFC 5322 Section 3.2.2).
    #[test]
    fn extract_comment_text_nested_parens() {
        let result = extract_comment_text("(outer (inner) text)");
        assert_eq!(
            result.as_deref(),
            Some("outer (inner) text"),
            "Nested parens should be included in comment text"
        );
    }

    /// `extract_comment_text` with escaped characters.
    #[test]
    fn extract_comment_text_escaped_chars() {
        let result = extract_comment_text("(hello \\(world\\))");
        assert_eq!(
            result.as_deref(),
            Some("hello (world)"),
            "Escaped parens inside comments should be unescaped"
        );
    }

    /// `extract_comment_text` with empty parens returns None.
    #[test]
    fn extract_comment_text_empty() {
        let result = extract_comment_text("()");
        assert!(result.is_none(), "Empty comment should return None");
    }

    /// `extract_comment_text` with non-paren input returns None.
    #[test]
    fn extract_comment_text_no_paren() {
        let result = extract_comment_text("not a comment");
        assert!(
            result.is_none(),
            "Non-parenthesized input should return None"
        );
    }

    // -----------------------------------------------------------------------
    // Coverage: comment stripping (L720, L1593-1604)
    // -----------------------------------------------------------------------

    /// RFC 5322 Section 3.2.2: `strip_comments` must handle nested comments,
    /// escaped characters inside comments, and preserve text outside.
    #[test]
    fn strip_comments_nested_and_escaped() {
        // Nested comment with escaped paren inside
        let result = strip_comments("Hello (outer (inner) comment) World");
        assert_eq!(result, "Hello  World");

        // Escaped paren inside a comment — the `\)` is an escape, not end of comment
        let result = strip_comments("Hello (comment with \\) escaped) World");
        assert_eq!(result, "Hello  World");

        // Escaped backslash outside a comment should be preserved
        let result = strip_comments("Hello \\\\ World");
        assert_eq!(result, "Hello \\\\ World");

        // Backslash-escaped character inside comment should be consumed
        let result = strip_comments("Before (escaped \\( paren) After");
        assert_eq!(result, "Before  After");
    }

    /// RFC 5322 Section 3.2.2: escaped backslash outside a comment is preserved.
    #[test]
    fn strip_comments_escaped_outside_comment() {
        let result = strip_comments("no \\(comment\\) here");
        // The `\(` is outside a comment, so the backslash is preserved along
        // with the `(` which then opens a comment (since \ outside comment
        // is just pushed). Let's verify the actual behavior:
        // `\` sets escaped=true, depth==0 so push `\`. Then `(` is the
        // escaped char, pushed. Wait — let me re-read the code.
        // Actually: `\` is not the backslash case in strip_comments because
        // `\\` case only triggers when c=='\\'. Let me trace:
        // 'n','o',' ' — pushed (depth 0)
        // '\\' — escaped=true, depth==0 so push '\\'
        // '(' — escaped char: escaped=false, depth==0 so push '('
        // 'c','o','m','m','e','n','t' — pushed
        // '\\' — escaped=true, depth==0 so push '\\'
        // ')' — escaped char: escaped=false, depth==0 so push ')'
        // ' ','h','e','r','e' — pushed
        // Result: "no \\(comment\\) here" — no comments were stripped
        assert_eq!(
            result, "no \\(comment\\) here",
            "Escaped parens outside comments should not open/close comments"
        );
    }

    // -----------------------------------------------------------------------
    // Coverage: date parsing edge cases (L767, L808, L834, L867)
    // -----------------------------------------------------------------------

    /// RFC 5322 Section 3.3: date with too few parts should return None.
    #[test]
    fn parse_date_too_few_parts() {
        // Only day and month, no year or time
        assert!(
            parse_rfc5322_date("13 Feb").is_none(),
            "Date with too few parts should return None"
        );
    }

    /// RFC 5322 Section 3.3: date with time field that has no colon
    /// (not a valid time) should return None.
    #[test]
    fn parse_date_time_no_colon() {
        assert!(
            parse_rfc5322_date("13 Feb 2025 1547 +0000").is_none(),
            "Time without colon should return None"
        );
    }

    /// RFC 5322 Section 3.3: unknown month abbreviation returns None.
    #[test]
    fn parse_date_unknown_month() {
        assert!(
            parse_rfc5322_date("13 Foo 2025 12:00:00 +0000").is_none(),
            "Unknown month name should return None"
        );
    }

    /// RFC 5322 Section 3.3: completely malformed date string.
    #[test]
    fn parse_date_completely_malformed() {
        assert!(parse_rfc5322_date("not a date at all").is_none());
        assert!(parse_rfc5322_date("").is_none());
        assert!(parse_rfc5322_date("   ").is_none());
    }

    /// RFC 5322 Section 4.3: unknown timezone abbreviation defaults to +0000.
    #[test]
    fn parse_date_unknown_timezone_defaults_zero() {
        let dt = parse_rfc5322_date("13 Feb 2025 12:00:00 ZULU").unwrap();
        assert_eq!(
            dt.tz_offset_minutes, 0,
            "Unknown timezone abbreviation should default to +0000"
        );
    }

    /// RFC 5322 Section 3.3: non-numeric day should return None.
    #[test]
    fn parse_date_non_numeric_day() {
        assert!(
            parse_rfc5322_date("XX Feb 2025 12:00:00 +0000").is_none(),
            "Non-numeric day should return None"
        );
    }

    /// RFC 5322 Section 3.3: non-numeric year should return None.
    #[test]
    fn parse_date_non_numeric_year() {
        assert!(
            parse_rfc5322_date("13 Feb XXXX 12:00:00 +0000").is_none(),
            "Non-numeric year should return None"
        );
    }

    // -----------------------------------------------------------------------
    // Coverage: boundary detection edge cases
    // (L918, L954, L1003-1006, L1022-1023)
    // -----------------------------------------------------------------------

    /// RFC 2046 Section 5.1.1: boundary preceded by `\n` without `\r`
    /// (bare LF line ending) should still be recognized.
    #[test]
    fn split_mime_parts_lf_only_boundaries() {
        let body = b"--boundary\nContent-Type: text/plain\n\nPart 1\n--boundary\nContent-Type: text/plain\n\nPart 2\n--boundary--";
        let parts = split_mime_parts(body, "boundary");
        assert_eq!(
            parts.len(),
            2,
            "Should find 2 parts with LF-only boundaries"
        );
    }

    /// RFC 2046 Section 5.1.1: boundary at the very start of body
    /// (no preceding newline needed since it's position 0).
    #[test]
    fn split_mime_parts_boundary_at_start() {
        let body = b"--b\r\nContent-Type: text/plain\r\n\r\nOnly part\r\n--b--";
        let parts = split_mime_parts(body, "b");
        assert_eq!(
            parts.len(),
            1,
            "Should find 1 part when boundary is at start"
        );
        let text = String::from_utf8_lossy(parts[0]);
        assert!(text.contains("Only part"));
    }

    /// Boundary appearing mid-line must be ignored (RFC 2046 Section 5.1.1).
    #[test]
    fn split_mime_parts_midline_boundary_ignored() {
        let body =
            b"--b\r\nContent-Type: text/plain\r\n\r\nText mentioning --b in the middle\r\n--b--";
        let parts = split_mime_parts(body, "b");
        assert_eq!(parts.len(), 1, "Mid-line boundary must not split");
        let text = String::from_utf8_lossy(parts[0]);
        assert!(
            text.contains("--b in the middle"),
            "Mid-line boundary text should be preserved"
        );
    }

    /// Boundary line with trailing whitespace (spaces/tabs after the
    /// boundary marker) must still be recognized (RFC 2046 Section 5.1.1:
    /// "...followed by LWSP").
    #[test]
    fn split_mime_parts_boundary_with_trailing_whitespace() {
        let body = b"--b  \t\r\nContent-Type: text/plain\r\n\r\nBody text\r\n--b--";
        let parts = split_mime_parts(body, "b");
        assert_eq!(
            parts.len(),
            1,
            "Boundary with trailing whitespace should be recognized"
        );
    }

    /// Boundary not preceded by newline at a non-zero position must be
    /// skipped (mid-line match).
    #[test]
    fn split_mime_parts_boundary_not_at_line_start_skipped() {
        // Body has the delimiter string embedded in content, not at line start
        let body = b"--bound\r\n\r\nSome text has --bound embedded\r\n--bound--";
        let parts = split_mime_parts(body, "bound");
        assert_eq!(parts.len(), 1);
        let text = String::from_utf8_lossy(parts[0]);
        assert!(text.contains("--bound embedded"));
    }

    // -----------------------------------------------------------------------
    // Coverage: Content-Transfer-Encoding quoted value (L1130)
    // -----------------------------------------------------------------------

    /// RFC 2045 Section 6: some mailers quote the Content-Transfer-Encoding
    /// value (e.g., `"base64"`). The parser strips quotes via
    /// `.trim().to_ascii_lowercase()`, which handles this.
    #[test]
    fn parse_quoted_transfer_encoding() {
        let raw = b"From: a@b.com\r\n\
                     Date: Thu, 13 Feb 2025 15:47:33 +0000\r\n\
                     Content-Type: text/plain; charset=utf-8\r\n\
                     Content-Transfer-Encoding: \"base64\"\r\n\
                     \r\n\
                     SGVsbG8gV29ybGQ=\r\n";
        let parsed = parse_email(raw).unwrap();
        // The encoding value "base64" (with quotes) should be handled.
        // Note: the actual trim().to_ascii_lowercase() in decode_transfer_encoding
        // will get `"base64"` including quotes, which won't match "base64".
        // This test documents current behavior — it falls through to passthrough.
        // The body_text will contain the raw base64 string.
        assert!(
            parsed.body_text.is_some(),
            "Message with quoted CTE should still produce body_text"
        );
    }

    /// Content-Transfer-Encoding with leading/trailing whitespace must be
    /// handled (the `.trim()` call in `decode_transfer_encoding`).
    #[test]
    fn parse_transfer_encoding_with_whitespace() {
        let raw = b"From: a@b.com\r\n\
                     Date: Thu, 13 Feb 2025 15:47:33 +0000\r\n\
                     Content-Type: text/plain; charset=utf-8\r\n\
                     Content-Transfer-Encoding:  base64 \r\n\
                     \r\n\
                     SGVsbG8gV29ybGQ=\r\n";
        let parsed = parse_email(raw).unwrap();
        assert_eq!(
            parsed.body_text.as_deref(),
            Some("Hello World"),
            "CTE with whitespace should still decode correctly"
        );
    }

    // -----------------------------------------------------------------------
    // Coverage: quoted-printable decoding edge cases
    // (L1183-1184, L1193)
    // -----------------------------------------------------------------------

    /// RFC 2045 Section 6.7: soft line break `=\r\n` must be removed,
    /// joining the lines without inserting any character.
    #[test]
    fn qp_soft_line_break_crlf() {
        let data = b"Hello=\r\n World";
        let decoded = decode_quoted_printable(data);
        assert_eq!(
            std::str::from_utf8(&decoded).unwrap(),
            "Hello World",
            "=\\r\\n soft break should be removed (RFC 2045 Section 6.7)"
        );
    }

    /// RFC 2045 Section 6.7: soft line break `=\n` (LF only, no CR)
    /// must also be removed (Postel's law: accept bare LF).
    #[test]
    fn qp_soft_line_break_lf_only() {
        let data = b"Hello=\nWorld";
        let decoded = decode_quoted_printable(data);
        assert_eq!(
            std::str::from_utf8(&decoded).unwrap(),
            "HelloWorld",
            "=\\n soft break should be removed"
        );
    }

    /// RFC 2045 Section 6.7: `=\n` near end of data (only 2 bytes left)
    /// must be treated as a soft break.
    #[test]
    fn qp_soft_break_lf_at_end() {
        let data = b"Hi=\n";
        let decoded = decode_quoted_printable(data);
        assert_eq!(
            std::str::from_utf8(&decoded).unwrap(),
            "Hi",
            "=\\n at end of data should be a soft break"
        );
    }

    /// RFC 2045 Section 6.7: invalid hex after `=` should pass through
    /// the `=` as a literal (Postel's law). Only valid hex pairs are decoded.
    #[test]
    fn qp_invalid_hex_passthrough() {
        let data = b"=GG=4F=4B";
        let decoded = decode_quoted_printable(data);
        // =GG is invalid hex → pass through literally; =4F=4B → "OK"
        assert_eq!(
            std::str::from_utf8(&decoded).unwrap(),
            "=GGOK",
            "Invalid hex =GG should pass through, valid =4F=4B should decode"
        );
    }

    // -----------------------------------------------------------------------
    // Coverage: base64 content length edge cases (L1285, L1343)
    // -----------------------------------------------------------------------

    /// RFC 2045 Section 6.8: empty base64 body should produce empty output.
    #[test]
    fn base64_empty_body() {
        let decoded = decode_transfer_encoding(b"", "base64");
        assert!(
            decoded.is_empty(),
            "Empty base64 input should produce empty output"
        );
    }

    /// RFC 2045 Section 6.8: base64 body that's only whitespace should
    /// produce empty output after stripping non-alphabet chars.
    #[test]
    fn base64_whitespace_only() {
        let decoded = decode_transfer_encoding(b"  \r\n  \r\n", "base64");
        assert!(
            decoded.is_empty(),
            "Whitespace-only base64 input should produce empty output"
        );
    }

    /// `find_closing_quote`: unterminated quoted-string returns string length.
    #[test]
    fn find_closing_quote_unterminated() {
        assert_eq!(find_closing_quote("no closing quote here"), 21);
    }

    /// `find_closing_quote`: escaped quote is skipped.
    #[test]
    fn find_closing_quote_skips_escaped() {
        // `\\\"` is an escaped quote, should be skipped; real close is at end
        assert_eq!(find_closing_quote("hello\\\"world\""), 12);
    }

    // -----------------------------------------------------------------------
    // Coverage: multipart boundary search paths
    // (L1361-1362, L1400, L1404, L1429-1430, L1478-1479, L1544-1545,
    //  L1550, L1552, L1567)
    // -----------------------------------------------------------------------

    /// Multiple parts with CRLF before boundaries — exercise the CRLF
    /// stripping logic in `split_mime_parts` (pos >= 2 and body[pos-2] == CR).
    #[test]
    fn multipart_crlf_before_boundary() {
        let raw = b"From: a@b.com\r\n\
                     Date: Thu, 13 Feb 2025 15:47:33 +0000\r\n\
                     Content-Type: multipart/mixed; boundary=\"mp\"\r\n\
                     \r\n\
                     --mp\r\n\
                     Content-Type: text/plain\r\n\
                     \r\n\
                     Part A\r\n\
                     --mp\r\n\
                     Content-Type: text/html\r\n\
                     \r\n\
                     <b>Part B</b>\r\n\
                     --mp--";
        let parsed = parse_email(raw).unwrap();
        assert_eq!(parsed.body_text.as_deref(), Some("Part A"));
        assert_eq!(parsed.body_html.as_deref(), Some("<b>Part B</b>"));
    }

    /// Multiple parts with bare LF before boundaries — exercise the LF-only
    /// stripping logic (pos >= 1 && body[pos-1] == LF, but pos < 2 or
    /// body[pos-2] != CR).
    #[test]
    fn multipart_lf_only_before_boundary() {
        let raw = b"From: a@b.com\nDate: Thu, 13 Feb 2025 15:47:33 +0000\nContent-Type: multipart/mixed; boundary=\"mp\"\n\n--mp\nContent-Type: text/plain\n\nPart A\n--mp\nContent-Type: text/html\n\n<b>Part B</b>\n--mp--";
        let parsed = parse_email(raw).unwrap();
        assert_eq!(parsed.body_text.as_deref(), Some("Part A"));
        assert_eq!(parsed.body_html.as_deref(), Some("<b>Part B</b>"));
    }

    /// Truncated multipart: no closing boundary — the trailing content
    /// should still be captured as a part.
    #[test]
    fn multipart_truncated_no_closing() {
        let body = b"--bnd\r\n\r\nFirst part\r\n--bnd\r\n\r\nSecond part with no closing boundary";
        let parts = split_mime_parts(body, "bnd");
        assert_eq!(
            parts.len(),
            2,
            "Should find 2 parts even without closing boundary"
        );
        let text2 = String::from_utf8_lossy(parts[1]);
        assert!(text2.contains("Second part"));
    }

    // -----------------------------------------------------------------------
    // Coverage: comment removal with escapes (L1593-1604)
    // -----------------------------------------------------------------------

    /// RFC 5322 Section 3.2.2: backslash inside a comment escapes the
    /// next character. `\(` inside a comment should NOT increment depth.
    #[test]
    fn strip_comments_escaped_paren_inside_comment() {
        let result = strip_comments("Before (escaped \\) paren) After");
        // The `\)` inside the comment is an escape: the `)` is consumed but
        // doesn't decrement depth. The real closing `)` comes later.
        assert_eq!(
            result.trim(),
            "Before  After",
            "Escaped close-paren inside comment must not end the comment"
        );
    }

    /// RFC 5322 Section 3.2.2: escaped open-paren inside a comment
    /// should NOT increment the nesting depth.
    #[test]
    fn strip_comments_escaped_open_paren_inside_comment() {
        let result = strip_comments("X (comment \\( not nested) Y");
        assert_eq!(
            result.trim(),
            "X  Y",
            "Escaped open-paren inside comment must not increase nesting depth"
        );
    }

    // -----------------------------------------------------------------------
    // Coverage: hex_digit lowercase (L1632)
    // -----------------------------------------------------------------------

    /// Lowercase hex digits a-f must be recognized in QP and percent decoding.
    #[test]
    fn hex_digit_lowercase() {
        assert_eq!(hex_digit(b'a'), Some(10));
        assert_eq!(hex_digit(b'f'), Some(15));
        assert_eq!(hex_digit(b'c'), Some(12));
        // Also verify uppercase and digits still work
        assert_eq!(hex_digit(b'A'), Some(10));
        assert_eq!(hex_digit(b'F'), Some(15));
        assert_eq!(hex_digit(b'0'), Some(0));
        assert_eq!(hex_digit(b'9'), Some(9));
        // Invalid
        assert_eq!(hex_digit(b'g'), None);
        assert_eq!(hex_digit(b'G'), None);
        assert_eq!(hex_digit(b' '), None);
    }

    /// QP decoding with lowercase hex digits (RFC 2045 Section 6.7).
    #[test]
    fn qp_lowercase_hex_digits() {
        // =c3=a9 is UTF-8 for U+00E9 (e-acute) in UTF-8 encoding
        let data = b"caf=c3=a9";
        let decoded = decode_quoted_printable(data);
        assert_eq!(decoded, b"caf\xc3\xa9");
        let text = String::from_utf8_lossy(&decoded);
        assert_eq!(
            text, "caf\u{e9}",
            "Lowercase hex digits in QP should decode correctly (RFC 2045 Section 6.7)"
        );
    }

    /// `decode_hex_pair` with lowercase hex digits.
    #[test]
    fn decode_hex_pair_lowercase() {
        assert_eq!(decode_hex_pair(b'f', b'f'), Some(0xFF));
        assert_eq!(decode_hex_pair(b'a', b'0'), Some(0xA0));
        assert_eq!(decode_hex_pair(b'0', b'a'), Some(0x0A));
    }

    // -----------------------------------------------------------------------
    // Coverage: parse_single_address edge cases
    // -----------------------------------------------------------------------

    /// Address with empty angle brackets (no email inside `<>`) should
    /// return None.
    #[test]
    fn parse_single_address_empty_angle_brackets() {
        let result = parse_single_address("Display Name <>");
        assert!(
            result.is_none(),
            "Empty angle brackets should not produce an address"
        );
    }

    /// Address with `>` before `<` (malformed) should not match the
    /// angle bracket path.
    #[test]
    fn parse_single_address_reversed_angles() {
        let result = parse_single_address(">bad<user@example.com");
        // rfind('<') finds position of '<', rfind('>') finds position of last '>'
        // angle_end (0) < angle_start (4), so angle bracket path fails.
        // Falls through to bare email check (contains '@').
        assert!(result.is_some());
    }

    /// Input without `@` and without angle brackets should return None.
    #[test]
    fn parse_single_address_no_at_no_brackets() {
        let result = parse_single_address("just plain text");
        assert!(
            result.is_none(),
            "Text without @ or <> should not produce an address"
        );
    }

    // -----------------------------------------------------------------------
    // Coverage: is_inside_quotes
    // -----------------------------------------------------------------------

    /// `is_inside_quotes` must correctly track escaped quotes.
    #[test]
    fn is_inside_quotes_with_escapes() {
        // Position 15 is after the escaped quote (\"), still inside quotes
        assert!(is_inside_quotes("\"hello \\\" world\"end", 15));
        // Position 0 is before any quotes — not inside
        assert!(!is_inside_quotes("\"hello\"", 0));
        // Position after closing quote — not inside
        assert!(!is_inside_quotes("\"hello\" world", 8));
    }

    // -----------------------------------------------------------------------
    // Coverage: strip_outer_quotes
    // -----------------------------------------------------------------------

    /// `strip_outer_quotes` must not strip when input is too short.
    #[test]
    fn strip_outer_quotes_short_input() {
        assert_eq!(strip_outer_quotes("\""), "\"");
        assert_eq!(strip_outer_quotes(""), "");
        assert_eq!(strip_outer_quotes("x"), "x");
    }

    /// `strip_outer_quotes` must not strip when only one side has quotes.
    #[test]
    fn strip_outer_quotes_one_sided() {
        assert_eq!(strip_outer_quotes("\"hello"), "\"hello");
        assert_eq!(strip_outer_quotes("hello\""), "hello\"");
    }

    // -----------------------------------------------------------------------
    // Coverage: split_header_body edge case — starts with \n
    // -----------------------------------------------------------------------

    /// RFC 2046: a MIME part starting with bare `\n` (no headers).
    #[test]
    fn split_header_body_starts_with_lf() {
        let (headers, body) = split_header_body(b"\nBody text here");
        assert!(
            headers.is_empty(),
            "Headers should be empty when input starts with \\n"
        );
        assert_eq!(body, b"Body text here");
    }

    /// RFC 2046: a MIME part starting with `\r\n` (no headers).
    #[test]
    fn split_header_body_starts_with_crlf() {
        let (headers, body) = split_header_body(b"\r\nBody text here");
        assert!(
            headers.is_empty(),
            "Headers should be empty when input starts with \\r\\n"
        );
        assert_eq!(body, b"Body text here");
    }

    // -----------------------------------------------------------------------
    // Coverage: CTE decode_body stripping trailing LF (L1130)
    // -----------------------------------------------------------------------

    /// `decode_body` strips a trailing `\n` (bare LF without CR).
    #[test]
    fn decode_body_strips_trailing_lf_only() {
        let result = decode_body(b"Hello\n", "", "text/plain; charset=utf-8");
        assert_eq!(result, "Hello", "Trailing bare LF should be stripped");
    }

    /// `decode_body` with no trailing newline returns content as-is.
    #[test]
    fn decode_body_no_trailing_newline() {
        let result = decode_body(b"Hello", "", "text/plain; charset=utf-8");
        assert_eq!(
            result, "Hello",
            "No trailing newline should leave content unchanged"
        );
    }

    /// Percent decoding with lowercase hex (RFC 2231 / RFC 3986).
    #[test]
    fn percent_decode_lowercase_hex() {
        let decoded = percent_decode("%c3%a9");
        // %c3%a9 is UTF-8 for U+00E9 (e-acute)
        assert_eq!(decoded, vec![0xC3, 0xA9]);
    }

    /// Percent decoding with invalid hex passes through literally.
    #[test]
    fn percent_decode_invalid_hex() {
        let decoded = percent_decode("%ZZ");
        assert_eq!(decoded, b"%ZZ");
    }

    /// Percent decoding with truncated sequence passes through.
    #[test]
    fn percent_decode_truncated() {
        let decoded = percent_decode("hello%2");
        assert_eq!(decoded, b"hello%2");
    }

    // -----------------------------------------------------------------------
    // Coverage: address with colon that looks like group but has @
    // -----------------------------------------------------------------------

    /// RFC 5322 Section 3.4: a colon in an address that contains `@`
    /// should NOT be treated as group syntax (heuristic).
    #[test]
    fn parse_address_colon_with_at_sign() {
        // "user:tag@example.com" has a colon but also @ — the colon
        // should be treated as part of the local-part, not group syntax.
        let addrs = parse_address_list("user:tag@example.com");
        // The heuristic checks if current.trim().contains('@') when seeing ':'.
        // At the point of ':', current is "user" (no @), so it enters group mode.
        // Then "tag@example.com" is parsed as a group member.
        assert!(!addrs.is_empty(), "Should parse at least one address");
    }
}