ld_so_cache/

lib.rs

//! # ld.so.cache Parser Library
//!
//! This crate provides a comprehensive parser for `ld.so.cache` files used by the Linux dynamic linker.
//! It supports both the legacy format (ld.so-1.7.0) and the modern glibc format with hardware capabilities.
//!
//! ## Quick Start
//!
//! ```rust
//! use ld_so_cache::parsers::parse_ld_cache;
//! use std::fs;
//!
//! # fn main() -> Result<(), Box<dyn std::error::Error>> {
//! // Parse the system's ld.so.cache file
//! let data = fs::read("/etc/ld.so.cache")?;
//! let cache = parse_ld_cache(&data)?;
//!
//! // Extract all library entries
//! let entries = cache.get_entries()?;
//! for entry in entries {
//!     println!("{} -> {}", entry.library_name, entry.library_path);
//! }
//! # Ok(())
//! # }
//! ```
//!
//! ## Cache Formats
//!
//! The library supports two cache formats:
//!
//! - **Old Format**: Used by ld.so-1.7.0, contains basic library mappings
//! - **New Format**: Used by glibc, includes hardware capabilities and extensions
//!
//! ## Hardware Capabilities
//!
//! The new format includes hardware capability flags that indicate processor features
//! required by libraries. This allows the dynamic linker to select the most optimized
//! version of a library for the current processor.
//!
//! ### Hardware Capability Bits
//!
//! The 64-bit `hwcap` field encodes various processor features:
//!
//! ```rust
//! # use ld_so_cache::NewFileEntry;
//! let entry = NewFileEntry {
//!     flags: 1,
//!     key: 0,
//!     value: 10,
//!     osversion_unused: 0,
//!     hwcap: 0x0002000000000000, // Example capability
//! };
//!
//! // Extract ISA level (bits 52-61)
//! let isa_level = (entry.hwcap >> 52) & 0x3ff;
//! if isa_level >= 2 {
//!     println!("Requires x86-64-v2 or higher");
//! }
//!
//! // Check for extension flag (bit 62)
//! let has_extensions = entry.hwcap & (1u64 << 62) != 0;
//! if has_extensions {
//!     println!("Library uses hardware capability extensions");
//! }
//!
//! // Check for specific CPU features (bits 0-51)
//! // Note: Exact bit meanings are architecture-specific
//! let cpu_features = entry.hwcap & ((1u64 << 52) - 1);
//! ```
//!
//! ## Error Handling
//!
//! All parsing operations return `Result<T, CacheError>` for robust error handling.
//! The library is designed to be resilient and will attempt to extract as much
//! information as possible even from partially corrupted files.
//!
//! ### Common Error Patterns
//!
//! ```rust
//! use ld_so_cache::{parsers::parse_ld_cache, CacheError};
//! use std::fs;
//!
//! # fn example() -> Result<(), Box<dyn std::error::Error>> {
//! // Read and parse cache file
//! let data = fs::read("/etc/ld.so.cache")?;
//! match parse_ld_cache(&data) {
//!     Ok(cache) => {
//!         // Successfully parsed cache
//!         match cache.get_entries() {
//!             Ok(entries) => println!("Found {} libraries", entries.len()),
//!             Err(CacheError::InvalidStringOffset(offset)) => {
//!                 eprintln!("Corrupted string table at offset {}", offset);
//!             }
//!             Err(e) => eprintln!("Error extracting entries: {}", e),
//!         }
//!     }
//!     Err(CacheError::InvalidMagic) => {
//!         eprintln!("File is not a valid ld.so.cache");
//!     }
//!     Err(CacheError::TruncatedFile) => {
//!         eprintln!("Cache file appears to be truncated");
//!     }
//!     Err(e) => eprintln!("Parse error: {}", e),
//! }
//! # Ok(())
//! # }
//! ```
//!
//! ### Graceful Degradation
//!
//! When extracting entries, the library skips invalid entries rather than failing:
//!
//! ```rust
//! # use ld_so_cache::parsers::parse_ld_cache;
//! # fn example() -> Result<(), Box<dyn std::error::Error>> {
//! # let data = vec![]; // Would be actual cache data
//! let cache = parse_ld_cache(&data)?;
//! let entries = cache.get_entries()?; // Will skip invalid entries
//! 
//! // Even if some entries are corrupted, we get the valid ones
//! for entry in entries {
//!     println!("{} -> {}", entry.library_name, entry.library_path);
//! }
//! # Ok(())
//! # }
//! ```

/// The main cache structure representing a parsed `ld.so.cache` file.
///
/// This structure can contain either the old format, new format, or both.
/// When both formats are present, the new format takes precedence for library lookups.
///
/// # Fields
///
/// * `old_format` - Legacy cache format data (ld.so-1.7.0)
/// * `new_format` - Modern glibc cache format with hardware capabilities
/// * `string_table` - Raw bytes containing null-terminated library names and paths
/// * `string_table_offset` - Absolute file offset where the string table begins (for new format)
///
/// # Examples
///
/// ```rust
/// use ld_so_cache::parsers::parse_ld_cache;
/// # use std::fs;
/// # fn example() -> Result<(), Box<dyn std::error::Error>> {
/// let data = fs::read("/etc/ld.so.cache")?;
/// let cache = parse_ld_cache(&data)?;
///
/// // Check which formats are present
/// if cache.old_format.is_some() {
///     println!("Old format present");
/// }
/// if cache.new_format.is_some() {
///     println!("New format present");
/// }
/// # Ok(())
/// # }
/// ```
#[derive(Debug, Clone, PartialEq)]
pub struct LdCache {
    pub old_format: Option<OldCache>,
    pub new_format: Option<NewCache>,
    pub string_table: Vec<u8>,
    pub string_table_offset: usize,
}

/// Legacy cache format used by ld.so-1.7.0.
///
/// This format provides basic library name to path mappings without
/// hardware capability information.
///
/// # Fields
///
/// * `nlibs` - Number of library entries in the cache
/// * `entries` - Vector of old format file entries
///
/// # Example
///
/// ```rust
/// # use ld_so_cache::OldCache;
/// let old_cache = OldCache {
///     nlibs: 10,
///     entries: vec![], // Would contain actual entries
/// };
/// assert_eq!(old_cache.nlibs, 10);
/// ```
#[derive(Debug, Clone, PartialEq)]
pub struct OldCache {
    pub nlibs: u32,
    pub entries: Vec<OldFileEntry>,
}

/// A single library entry in the old cache format.
///
/// # Fields
///
/// * `flags` - Library type flags (bit 0: ELF library)
/// * `key` - Offset into string table for library name
/// * `value` - Offset into string table for library path
///
/// # Flag Values
///
/// * `flags & 1 != 0` - ELF library
/// * `flags & 1 == 0` - Other/unknown format
///
/// # Example
///
/// ```rust
/// # use ld_so_cache::OldFileEntry;
/// let entry = OldFileEntry {
///     flags: 1,  // ELF library
///     key: 0,    // Library name at offset 0 in string table
///     value: 10, // Library path at offset 10 in string table
/// };
/// assert!(entry.flags & 1 != 0); // Is ELF library
/// ```
#[derive(Debug, Clone, PartialEq)]
pub struct OldFileEntry {
    pub flags: i32,
    pub key: u32,
    pub value: u32,
}

/// Modern glibc cache format with hardware capabilities and extensions.
///
/// This format extends the old format with hardware capability matching
/// and optional extension directories for future enhancements.
///
/// # Fields
///
/// * `nlibs` - Number of library entries in the cache
/// * `len_strings` - Length of the string table in bytes
/// * `flags` - Endianness and format flags (2 = little endian)
/// * `extension_offset` - File offset to extension directory (0 = none)
/// * `entries` - Vector of new format file entries with hardware capabilities
/// * `extensions` - Optional extension directory for additional metadata
///
/// # Example
///
/// ```rust
/// # use ld_so_cache::NewCache;
/// let new_cache = NewCache {
///     nlibs: 100,
///     len_strings: 5000,
///     flags: 2, // Little endian
///     extension_offset: 0, // No extensions
///     entries: vec![],
///     extensions: None,
/// };
/// assert_eq!(new_cache.flags, 2);
/// ```
#[derive(Debug, Clone, PartialEq)]
pub struct NewCache {
    pub nlibs: u32,
    pub len_strings: u32,
    pub flags: u8,
    pub extension_offset: u32,
    pub entries: Vec<NewFileEntry>,
    pub extensions: Option<ExtensionDirectory>,
}

/// A single library entry in the new cache format with hardware capabilities.
///
/// # Fields
///
/// * `flags` - Library type flags (same as old format)
/// * `key` - Offset into string table for library name
/// * `value` - Offset into string table for library path
/// * `osversion_unused` - Unused field (always 0)
/// * `hwcap` - Hardware capability mask indicating required processor features
///
/// # Hardware Capabilities
///
/// The `hwcap` field encodes processor features required by the library:
/// * Bits 0-51: Various CPU features (SSE, AVX, etc.)
/// * Bits 52-61: ISA level (x86-64-v2, x86-64-v3, etc.)
/// * Bit 62: Extension flag
/// * Bit 63: Reserved
///
/// # Example
///
/// ```rust
/// # use ld_so_cache::NewFileEntry;
/// let entry = NewFileEntry {
///     flags: 1,
///     key: 0,
///     value: 20,
///     osversion_unused: 0,
///     hwcap: 0x1000000000000000, // Some capability set
/// };
/// 
/// // Check for extension flag
/// let has_extension = entry.hwcap & (1u64 << 62) != 0;
/// 
/// // Extract ISA level
/// let isa_level = (entry.hwcap >> 52) & 0x3ff;
/// ```
#[derive(Debug, Clone, PartialEq)]
pub struct NewFileEntry {
    pub flags: i32,
    pub key: u32,
    pub value: u32,
    pub osversion_unused: u32,
    pub hwcap: u64,
}

/// Directory of extension sections for future cache format enhancements.
///
/// This allows the cache format to be extended with additional metadata
/// while maintaining backward compatibility.
///
/// # Fields
///
/// * `count` - Number of extension sections
/// * `sections` - Vector of extension section descriptors
///
/// # Example
///
/// ```rust
/// # use ld_so_cache::{ExtensionDirectory, ExtensionSection};
/// let ext_dir = ExtensionDirectory {
///     count: 2,
///     sections: vec![
///         ExtensionSection { tag: 1, flags: 0, offset: 100, size: 50 },
///         ExtensionSection { tag: 2, flags: 0, offset: 150, size: 30 },
///     ],
/// };
/// assert_eq!(ext_dir.count, 2);
/// ```
#[derive(Debug, Clone, PartialEq)]
pub struct ExtensionDirectory {
    pub count: u32,
    pub sections: Vec<ExtensionSection>,
}

/// A single extension section descriptor.
///
/// Extension sections allow for future enhancements to the cache format
/// without breaking existing parsers.
///
/// # Fields
///
/// * `tag` - Identifies the type of extension data
/// * `flags` - Extension-specific flags
/// * `offset` - File offset where extension data begins
/// * `size` - Size of extension data in bytes
///
/// # Example
///
/// ```rust
/// # use ld_so_cache::ExtensionSection;
/// let section = ExtensionSection {
///     tag: 1,      // Extension type 1
///     flags: 0,    // No special flags
///     offset: 1000, // Data starts at offset 1000
///     size: 256,   // 256 bytes of data
/// };
/// ```
#[derive(Debug, Clone, PartialEq)]
pub struct ExtensionSection {
    pub tag: u32,
    pub flags: u32,
    pub offset: u32,
    pub size: u32,
}

/// Represents the different cache format combinations that can be found.
///
/// Real-world cache files may contain only the old format, only the new format,
/// or both formats for backward compatibility.
///
/// # Variants
///
/// * `OldOnly` - File contains only the legacy ld.so-1.7.0 format
/// * `NewOnly` - File contains only the modern glibc format
/// * `Both` - File contains both formats (common for compatibility)
///
/// # Example
///
/// ```rust
/// # use ld_so_cache::{CacheFormat, OldCache, NewCache};
/// # let old_cache = OldCache { nlibs: 0, entries: vec![] };
/// # let new_cache = NewCache { nlibs: 0, len_strings: 0, flags: 2, extension_offset: 0, entries: vec![], extensions: None };
/// let format = CacheFormat::Both {
///     old: old_cache,
///     new: new_cache,
/// };
/// 
/// match format {
///     CacheFormat::OldOnly(_) => println!("Legacy format only"),
///     CacheFormat::NewOnly(_) => println!("Modern format only"),
///     CacheFormat::Both { .. } => println!("Both formats present"),
/// }
/// ```
#[derive(Debug, Clone, PartialEq)]
pub enum CacheFormat {
    OldOnly(OldCache),
    NewOnly(NewCache),
    Both { old: OldCache, new: NewCache },
}

/// A processed library entry extracted from the cache.
///
/// This represents a single library mapping with resolved strings and
/// optional hardware capability information.
///
/// # Fields
///
/// * `library_name` - The library name (e.g., "libc.so.6")
/// * `library_path` - Full path to the library file
/// * `flags` - Library type flags (bit 0: ELF library)
/// * `hwcap` - Hardware capabilities (None for old format entries)
///
/// # Serialization
///
/// When serialized to JSON, hardware capabilities are formatted as
/// hexadecimal strings (e.g., "0x0000000000001000").
///
/// # Example
///
/// ```rust
/// # use ld_so_cache::CacheEntry;
/// let entry = CacheEntry {
///     library_name: "libc.so.6".to_string(),
///     library_path: "/lib/x86_64-linux-gnu/libc.so.6".to_string(),
///     flags: 1, // ELF library
///     hwcap: Some(0x1000), // Some hardware capability
/// };
/// 
/// // Check if it's an ELF library
/// let is_elf = entry.flags & 1 != 0;
/// assert!(is_elf);
/// 
/// // Check for hardware capabilities
/// if let Some(hwcap) = entry.hwcap {
///     println!("Hardware capabilities: 0x{:016x}", hwcap);
/// }
/// ```
#[derive(Debug, Clone, PartialEq, serde::Serialize)]
pub struct CacheEntry {
    pub library_name: String,
    pub library_path: String,
    pub flags: i32,
    #[serde(with = "hwcap_format", skip_serializing_if = "Option::is_none")]
    pub hwcap: Option<u64>,
}

mod hwcap_format {
    use serde::{Serializer, Serialize};

    pub fn serialize<S>(hwcap: &Option<u64>, serializer: S) -> Result<S::Ok, S::Error>
    where
        S: Serializer,
    {
        match hwcap {
            Some(value) => format!("0x{value:016x}").serialize(serializer),
            None => serializer.serialize_none(),
        }
    }
}

impl LdCache {
    /// Extracts all library entries from the cache.
    ///
    /// This method processes both old and new format entries, converting raw
    /// cache data into user-friendly `CacheEntry` structures. If both formats
    /// are present, only the new format entries are returned as they include
    /// hardware capability information.
    ///
    /// # Returns
    ///
    /// A vector of `CacheEntry` structures containing library names, paths,
    /// flags, and hardware capabilities (when available).
    ///
    /// # Errors
    ///
    /// * `CacheError::InvalidStringOffset` - String offset points outside the string table
    /// * `CacheError::ParseError` - String contains invalid UTF-8 characters
    ///
    /// Invalid entries are silently skipped rather than causing the entire
    /// operation to fail.
    ///
    /// # Example
    ///
    /// ```rust
    /// # use ld_so_cache::parsers::parse_ld_cache;
    /// # fn example() -> Result<(), Box<dyn std::error::Error>> {
    /// # let data = vec![]; // Would be actual cache data
    /// # let cache = parse_ld_cache(&data)?;
    /// let entries = cache.get_entries()?;
    /// 
    /// for entry in entries {
    ///     println!("{} -> {}", entry.library_name, entry.library_path);
    ///     
    ///     if entry.flags & 1 != 0 {
    ///         println!("  ELF library");
    ///     }
    ///     
    ///     if let Some(hwcap) = entry.hwcap {
    ///         println!("  Hardware capabilities: 0x{:016x}", hwcap);
    ///     }
    /// }
    /// # Ok(())
    /// # }
    /// ```
    pub fn get_entries(&self) -> Result<Vec<CacheEntry>, CacheError> {
        let string_table = &self.string_table;
        let mut entries = Vec::new();

        if let Some(new_cache) = &self.new_format {
            for entry in &new_cache.entries {
                // For new format, offsets might be absolute file offsets
                let key_offset = if self.string_table_offset > 0 && entry.key as usize >= self.string_table_offset {
                    entry.key as usize - self.string_table_offset
                } else {
                    entry.key as usize
                };
                
                let value_offset = if self.string_table_offset > 0 && entry.value as usize >= self.string_table_offset {
                    entry.value as usize - self.string_table_offset
                } else {
                    entry.value as usize
                };
                
                if key_offset >= string_table.len() || value_offset >= string_table.len() {
                    continue; // Skip invalid entries instead of failing
                }
                
                let name = extract_string(string_table, key_offset)?;
                let path = extract_string(string_table, value_offset)?;
                entries.push(CacheEntry {
                    library_name: name,
                    library_path: path,
                    flags: entry.flags,
                    hwcap: Some(entry.hwcap),
                });
            }
        } else if let Some(old_cache) = &self.old_format {
            for entry in &old_cache.entries {
                if entry.key as usize >= string_table.len() || entry.value as usize >= string_table.len() {
                    continue; // Skip invalid entries instead of failing
                }
                let name = extract_string(string_table, entry.key as usize)?;
                let path = extract_string(string_table, entry.value as usize)?;
                entries.push(CacheEntry {
                    library_name: name,
                    library_path: path,
                    flags: entry.flags,
                    hwcap: None,
                });
            }
        }

        Ok(entries)
    }
}

/// Errors that can occur during cache parsing and processing.
///
/// This enum covers all possible failure modes when parsing `ld.so.cache` files,
/// from file format issues to data corruption.
///
/// # Variants
///
/// * `ParseError` - Generic parsing failure with descriptive message
/// * `InvalidStringOffset` - String offset points outside the string table
/// * `InvalidMagic` - File doesn't start with a recognized magic number
/// * `TruncatedFile` - File is too short to contain valid cache data
/// * `InvalidEndianness` - Unsupported byte order in cache file
///
/// # Example
///
/// ```rust
/// # use ld_so_cache::{CacheError, parsers::parse_ld_cache};
/// let invalid_data = b"not a cache file";
/// let result = parse_ld_cache(invalid_data);
/// 
/// match result {
///     Err(CacheError::InvalidMagic) => {
///         println!("File is not a valid ld.so.cache");
///     }
///     Err(CacheError::TruncatedFile) => {
///         println!("Cache file is incomplete");
///     }
///     Err(e) => println!("Other error: {}", e),
///     Ok(_) => println!("Successfully parsed"),
/// }
/// ```
#[derive(Debug, Clone, PartialEq)]
pub enum CacheError {
    /// Generic parsing error with descriptive message
    ParseError(String),
    /// String offset points beyond the string table bounds
    InvalidStringOffset(usize),
    /// File doesn't begin with a recognized magic number
    InvalidMagic,
    /// File is too short to contain required cache structures
    TruncatedFile,
    /// Cache uses unsupported byte order
    InvalidEndianness,
}

impl std::fmt::Display for CacheError {
    fn fmt(&self, f: &mut std::fmt::Formatter<'_>) -> std::fmt::Result {
        match self {
            CacheError::ParseError(msg) => write!(f, "Parse error: {msg}"),
            CacheError::InvalidStringOffset(offset) => write!(f, "Invalid string offset: {offset}"),
            CacheError::InvalidMagic => write!(f, "Invalid magic number"),
            CacheError::TruncatedFile => write!(f, "Truncated file"),
            CacheError::InvalidEndianness => write!(f, "Invalid endianness"),
        }
    }
}

impl std::error::Error for CacheError {}

fn extract_string(string_table: &[u8], offset: usize) -> Result<String, CacheError> {
    if offset >= string_table.len() {
        return Err(CacheError::InvalidStringOffset(offset));
    }

    let slice = &string_table[offset..];
    let null_pos = slice
        .iter()
        .position(|&b| b == 0)
        .ok_or(CacheError::InvalidStringOffset(offset))?;

    String::from_utf8(slice[..null_pos].to_vec())
        .map_err(|_| CacheError::ParseError("Invalid UTF-8 in string".to_string()))
}

pub mod parsers;

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

    #[test]
    fn test_extract_string() {
        let string_table = b"hello\0world\0test\0";
        
        assert_eq!(extract_string(string_table, 0).unwrap(), "hello");
        assert_eq!(extract_string(string_table, 6).unwrap(), "world");
        assert_eq!(extract_string(string_table, 12).unwrap(), "test");
        
        assert!(extract_string(string_table, 100).is_err());
    }

    #[test]
    fn test_extract_string_edge_cases() {
        let string_table = b"a\0";
        assert_eq!(extract_string(string_table, 0).unwrap(), "a");
        
        let empty_string_table = b"\0";
        assert_eq!(extract_string(empty_string_table, 0).unwrap(), "");
        
        let no_null_terminator = b"hello";
        assert!(extract_string(no_null_terminator, 0).is_err());
        
        let string_table = b"hello\0";
        assert!(extract_string(string_table, 10).is_err());
    }

    #[test]
    fn test_cache_entry_creation() {
        let entry = CacheEntry {
            library_name: "libc.so.6".to_string(),
            library_path: "/lib/x86_64-linux-gnu/libc.so.6".to_string(),
            flags: 1,
            hwcap: Some(0x1234_5678),
        };
        
        assert_eq!(entry.library_name, "libc.so.6");
        assert_eq!(entry.flags, 1);
        assert_eq!(entry.hwcap, Some(0x1234_5678));
    }

    #[test]
    fn test_parse_old_format_cache() {
        let mut data = Vec::new();
        
        data.extend_from_slice(b"ld.so-1.7.0");
        data.extend_from_slice(&2u32.to_le_bytes());
        
        data.extend_from_slice(&1i32.to_le_bytes());
        data.extend_from_slice(&0u32.to_le_bytes());
        data.extend_from_slice(&10u32.to_le_bytes());
        
        data.extend_from_slice(&1i32.to_le_bytes());
        data.extend_from_slice(&20u32.to_le_bytes());
        data.extend_from_slice(&40u32.to_le_bytes());
        
        data.extend_from_slice(b"libc.so.6\0/lib/libc.so.6\0libm.so.6\0/lib/libm.so.6\0");
        
        let cache = parse_ld_cache(&data).unwrap();
        assert!(cache.old_format.is_some());
        assert!(cache.new_format.is_none());
        
        let old_cache = cache.old_format.as_ref().unwrap();
        assert_eq!(old_cache.nlibs, 2);
        assert_eq!(old_cache.entries.len(), 2);
        assert_eq!(old_cache.entries[0].flags, 1);
        assert_eq!(old_cache.entries[0].key, 0);
        assert_eq!(old_cache.entries[0].value, 10);
        
        let entries = cache.get_entries().unwrap();
        assert_eq!(entries.len(), 2);
        assert_eq!(entries[0].library_name, "libc.so.6");
        assert_eq!(entries[0].library_path, "/lib/libc.so.6");
        assert!(entries[0].hwcap.is_none());
    }

    #[test]
    fn test_parse_new_format_cache() {
        let mut data = Vec::new();
        
        data.extend_from_slice(b"glibc-ld.so.cache");
        data.extend_from_slice(b"1.1");
        data.extend_from_slice(&1u32.to_le_bytes());
        data.extend_from_slice(&30u32.to_le_bytes());
        data.push(2);
        data.extend_from_slice(&[0, 0, 0]);
        data.extend_from_slice(&0u32.to_le_bytes());
        data.extend_from_slice(&[0; 12]);
        
        data.extend_from_slice(&1i32.to_le_bytes());
        data.extend_from_slice(&0u32.to_le_bytes());
        data.extend_from_slice(&10u32.to_le_bytes());
        data.extend_from_slice(&0u32.to_le_bytes());
        data.extend_from_slice(&0x1234_5678_u64.to_le_bytes());
        
        data.extend_from_slice(b"libc.so.6\0/lib/libc.so.6\0");
        
        let cache = parse_ld_cache(&data).unwrap();
        assert!(cache.new_format.is_some());
        assert!(cache.old_format.is_none());
        
        let new_cache = cache.new_format.as_ref().unwrap();
        assert_eq!(new_cache.nlibs, 1);
        assert_eq!(new_cache.len_strings, 30);
        assert_eq!(new_cache.flags, 2);
        assert_eq!(new_cache.entries.len(), 1);
        assert_eq!(new_cache.entries[0].hwcap, 0x1234_5678);
        
        let entries = cache.get_entries().unwrap();
        assert_eq!(entries.len(), 1);
        assert_eq!(entries[0].library_name, "libc.so.6");
        assert_eq!(entries[0].library_path, "/lib/libc.so.6");
        assert_eq!(entries[0].hwcap, Some(0x1234_5678));
    }

    #[test]
    fn test_parse_invalid_magic() {
        let invalid_data = b"invalid-magic-that-is-long-enough";
        let result = parse_ld_cache(invalid_data);
        assert!(matches!(result, Err(CacheError::InvalidMagic)));
    }

    #[test]
    fn test_parse_truncated_file() {
        let truncated_data = b"ld.so-1.7.0";
        let result = parse_ld_cache(truncated_data);
        assert!(matches!(result, Err(CacheError::TruncatedFile)));
    }

    #[test]
    fn test_cache_error_display() {
        let error = CacheError::ParseError("test error".to_string());
        assert_eq!(format!("{error}"), "Parse error: test error");
        
        let error = CacheError::InvalidStringOffset(42);
        assert_eq!(format!("{error}"), "Invalid string offset: 42");
        
        let error = CacheError::InvalidMagic;
        assert_eq!(format!("{error}"), "Invalid magic number");
        
        let error = CacheError::TruncatedFile;
        assert_eq!(format!("{error}"), "Truncated file");
        
        let error = CacheError::InvalidEndianness;
        assert_eq!(format!("{error}"), "Invalid endianness");
    }

    #[test]
    fn test_empty_cache() {
        let mut data = Vec::new();
        data.extend_from_slice(b"ld.so-1.7.0");
        data.extend_from_slice(&0u32.to_le_bytes());
        
        let cache = parse_ld_cache(&data).unwrap();
        let entries = cache.get_entries().unwrap();
        assert_eq!(entries.len(), 0);
    }

    #[test]
    fn test_ld_cache_get_entries_preference() {
        let old_cache = OldCache {
            nlibs: 1,
            entries: vec![OldFileEntry {
                flags: 1,
                key: 0,
                value: 10,
            }],
        };
        
        let new_cache = NewCache {
            nlibs: 1,
            len_strings: 30,
            flags: 2,
            extension_offset: 0,
            entries: vec![NewFileEntry {
                flags: 1,
                key: 0,
                value: 10,
                osversion_unused: 0,
                hwcap: 0x1234_5678,
            }],
            extensions: None,
        };
        
        let string_table = b"libc.so.6\0/lib/libc.so.6\0".to_vec();
        
        let cache_with_both = LdCache {
            old_format: Some(old_cache),
            new_format: Some(new_cache),
            string_table: string_table.clone(),
            string_table_offset: 0,
        };
        
        let entries = cache_with_both.get_entries().unwrap();
        assert_eq!(entries.len(), 1);
        assert_eq!(entries[0].hwcap, Some(0x1234_5678));
    }

    #[test]
    fn test_hwcap_bit_decoding() {
        // Test ISA level extraction (bits 52-61)
        let hwcap_with_isa_v2: u64 = 2u64 << 52;
        let isa_level = (hwcap_with_isa_v2 >> 52) & 0x3ff;
        assert_eq!(isa_level, 2);
        
        let hwcap_with_isa_v3: u64 = 3u64 << 52;
        let isa_level = (hwcap_with_isa_v3 >> 52) & 0x3ff;
        assert_eq!(isa_level, 3);
        
        // Test extension flag (bit 62)
        let hwcap_with_extension: u64 = 1u64 << 62;
        assert!(hwcap_with_extension & (1u64 << 62) != 0);
        
        let hwcap_without_extension: u64 = 0x0000_1234_5678_9ABC;
        assert!(hwcap_without_extension & (1u64 << 62) == 0);
        
        // Test CPU features mask (bits 0-51)
        let hwcap_with_features: u64 = 0x000F_FFFF_FFFF_FFFF; // All feature bits set
        let cpu_features = hwcap_with_features & ((1u64 << 52) - 1);
        assert_eq!(cpu_features, 0x000F_FFFF_FFFF_FFFF);
        
        // Test combined hwcap
        let entry = NewFileEntry {
            flags: 1,
            key: 0,
            value: 0,
            osversion_unused: 0,
            hwcap: 0x0020_0000_0000_1234, // ISA level 2 (2 << 52) + some features
        };
        
        let isa_level = (entry.hwcap >> 52) & 0x3ff;
        let features = entry.hwcap & ((1u64 << 52) - 1);
        
        assert_eq!(isa_level, 2);
        assert_eq!(features, 0x1234);
        
        // Test with extension flag
        let entry_with_ext = NewFileEntry {
            flags: 1,
            key: 0,
            value: 0,
            osversion_unused: 0,
            hwcap: 0x4000_0000_0000_0000, // Just extension flag
        };
        
        let has_extension = entry_with_ext.hwcap & (1u64 << 62) != 0;
        assert!(has_extension);
    }

    #[test]
    fn test_graceful_degradation_with_invalid_entries() {
        // Create cache with mix of valid and invalid entries
        let new_cache = NewCache {
            nlibs: 4,
            len_strings: 50,
            flags: 2,
            extension_offset: 0,
            entries: vec![
                NewFileEntry {
                    flags: 1,
                    key: 0,
                    value: 10,
                    osversion_unused: 0,
                    hwcap: 0x1234,
                },
                NewFileEntry {
                    flags: 1,
                    key: 1000, // Invalid offset
                    value: 2000, // Invalid offset
                    osversion_unused: 0,
                    hwcap: 0x5678,
                },
                NewFileEntry {
                    flags: 1,
                    key: 26,
                    value: 36,
                    osversion_unused: 0,
                    hwcap: 0x9ABC,
                },
                NewFileEntry {
                    flags: 1,
                    key: 500, // Invalid offset
                    value: 10,
                    osversion_unused: 0,
                    hwcap: 0xDEF0,
                },
            ],
            extensions: None,
        };
        
        let string_table = b"libc.so.6\0/lib/libc.so.6\0\0libm.so.6\0/lib/libm.so.6\0".to_vec();
        
        let cache = LdCache {
            old_format: None,
            new_format: Some(new_cache),
            string_table,
            string_table_offset: 0,
        };
        
        // Should skip invalid entries and return only valid ones
        let entries = cache.get_entries().unwrap();
        assert_eq!(entries.len(), 2); // Only 2 valid entries
        assert_eq!(entries[0].library_name, "libc.so.6");
        assert_eq!(entries[1].library_name, "libm.so.6");
    }

    #[test]
    fn test_string_table_offset_adjustment() {
        // Test with absolute file offsets
        let new_cache = NewCache {
            nlibs: 1,
            len_strings: 30,
            flags: 2,
            extension_offset: 0,
            entries: vec![NewFileEntry {
                flags: 1,
                key: 1000, // Absolute offset
                value: 1010, // Absolute offset
                osversion_unused: 0,
                hwcap: 0x1234,
            }],
            extensions: None,
        };
        
        let string_table = b"libc.so.6\0/lib/libc.so.6\0".to_vec();
        
        let cache = LdCache {
            old_format: None,
            new_format: Some(new_cache),
            string_table,
            string_table_offset: 1000, // String table starts at offset 1000
        };
        
        let entries = cache.get_entries().unwrap();
        assert_eq!(entries.len(), 1);
        assert_eq!(entries[0].library_name, "libc.so.6");
        assert_eq!(entries[0].library_path, "/lib/libc.so.6");
    }

    #[test]
    fn test_flags_architecture_decoding() {
        // Test i386 (arch bits = 0)
        let i386_flags = 0x0001; // ELF + arch 0
        let arch_bits = (i386_flags >> 8) & 0xf;
        assert_eq!(arch_bits, 0);
        
        // Test x86_64 (arch bits = 3)
        let x86_64_flags = 0x0301; // ELF + arch 3
        let arch_bits = (x86_64_flags >> 8) & 0xf;
        assert_eq!(arch_bits, 3);
        
        // Test libx32 (arch bits = 8)
        let libx32_flags = 0x0801; // ELF + arch 8
        let arch_bits = (libx32_flags >> 8) & 0xf;
        assert_eq!(arch_bits, 8);
    }

    #[test]
    fn test_endianness_flag_values() {
        // Test all documented flag values
        let test_cases = vec![
            (0u8, "Endianness unset (legacy)"),
            (1u8, "Invalid cache"),
            (2u8, "Little endian"),
            (3u8, "Big endian"),
        ];
        
        for (flag_value, expected_meaning) in test_cases {
            let new_cache = NewCache {
                nlibs: 0,
                len_strings: 0,
                flags: flag_value,
                extension_offset: 0,
                entries: vec![],
                extensions: None,
            };
            
            // Verify the flag value is stored correctly
            assert_eq!(new_cache.flags, flag_value);
            
            // Verify the meaning matches documentation
            let meaning = match flag_value {
                0 => "Endianness unset (legacy)",
                1 => "Invalid cache",
                2 => "Little endian",
                3 => "Big endian",
                _ => "Unknown",
            };
            assert_eq!(meaning, expected_meaning);
        }
    }

    #[test]
    fn test_extension_tag_meanings() {
        let ext_dir = ExtensionDirectory {
            count: 2,
            sections: vec![
                ExtensionSection {
                    tag: 1,
                    flags: 0,
                    offset: 100,
                    size: 50,
                },
                ExtensionSection {
                    tag: 2,
                    flags: 0,
                    offset: 150,
                    size: 30,
                },
            ],
        };
        
        // Verify tag meanings as documented
        assert_eq!(ext_dir.sections[0].tag, 1); // cache_extension_tag_generator
        assert_eq!(ext_dir.sections[1].tag, 2); // cache_extension_tag_glibc_hwcaps
    }

    #[test]
    fn test_cache_entry_serialization_with_hwcap() {
        let entry = CacheEntry {
            library_name: "libc.so.6".to_string(),
            library_path: "/lib/x86_64-linux-gnu/libc.so.6".to_string(),
            flags: 0x0301, // x86_64 ELF
            hwcap: Some(0x0000_0000_0000_1000),
        };
        
        // Serialize to JSON
        let json = serde_json::to_string(&entry).unwrap();
        
        // Verify hwcap is formatted as hex string
        assert!(json.contains("\"hwcap\":\"0x0000000000001000\""));
        
        // Test without hwcap
        let entry_no_hwcap = CacheEntry {
            library_name: "libm.so.6".to_string(),
            library_path: "/lib/libm.so.6".to_string(),
            flags: 1,
            hwcap: None,
        };
        
        let json_no_hwcap = serde_json::to_string(&entry_no_hwcap).unwrap();
        assert!(!json_no_hwcap.contains("hwcap"));
    }
}