ghostscope_compiler/script/

compiler.rs

use crate::script::ast::{Program, Statement, TracePattern};
use crate::CompileError;
// BinaryAnalyzer is now internal to ghostscope-binary, use DwarfAnalyzer instead
use ghostscope_dwarf::ModuleDefaultPolicy;
use inkwell::context::Context;
use std::borrow::Cow;
use std::collections::hash_map::DefaultHasher;
use std::fmt::Write as _;
use std::hash::{Hash, Hasher};
use tracing::{debug, error, info, warn};

/// Resolved target information from DWARF queries
#[derive(Debug, Clone)]
pub struct ResolvedTarget {
    pub function_name: Option<String>,
    pub function_address: Option<u64>,
    pub binary_path: String,
    pub uprobe_offset: Option<u64>,
    pub pattern: TracePattern,
}

/// Complete uprobe configuration ready for attachment
#[derive(Debug, Clone)]
pub struct UProbeConfig {
    /// The trace pattern this uprobe corresponds to
    pub trace_pattern: TracePattern,

    /// Target binary path
    pub binary_path: String,

    /// Function name (for FunctionName patterns)
    pub function_name: Option<String>,

    /// Resolved function address in the binary
    pub function_address: Option<u64>,

    /// Calculated uprobe offset (for aya uprobe attachment)
    pub uprobe_offset: Option<u64>,

    /// Process ID to attach to (None means attach to all instances)
    pub target_pid: Option<u32>,

    /// eBPF bytecode for this uprobe
    pub ebpf_bytecode: Vec<u8>,

    /// eBPF function name for this uprobe (e.g., "ghostscope_main_0", "ghostscope_printf_1")
    pub ebpf_function_name: String,

    /// Trace ID assigned by compiler (starts from starting_trace_id and increments)
    pub assigned_trace_id: u32,

    /// Trace context containing all strings, types, and variable names used in this uprobe
    pub trace_context: ghostscope_protocol::TraceContext,

    /// BPF-facing compact DWARF CFI rows used by the `bt` unwinder.
    pub backtrace_unwind_rows: Vec<ghostscope_protocol::BacktraceUnwindRow>,

    /// Module cookie to row range entries used by the `bt` unwinder.
    pub backtrace_module_row_ranges: Vec<(u64, ghostscope_protocol::BacktraceModuleRowRange)>,

    /// Optional eBPF tail-call step program used by the `bt` unwinder.
    pub backtrace_tail_call_program: Option<crate::ebpf::context::BacktraceTailCallProgram>,

    /// Global 1-based index of this address within the resolved target list (if applicable)
    pub resolved_address_index: Option<usize>,
}

/// Compilation result containing all uprobe configurations
#[derive(Debug)]
pub struct CompilationResult {
    pub uprobe_configs: Vec<UProbeConfig>,
    pub trace_count: usize,
    pub target_info: String,
    pub failed_targets: Vec<FailedTarget>, // New field for failed compilation info
    pub next_available_trace_id: u32,      // Next trace_id that can be used by trace_manager
}

/// Information about a target that failed to compile
#[derive(Debug, Clone)]
pub struct FailedTarget {
    pub target_name: String,
    pub pc_address: u64,
    pub error_message: String,
}

/// Unified AST compiler that performs DWARF queries and code generation in single pass
pub struct AstCompiler<'a> {
    process_analyzer: Option<&'a ghostscope_dwarf::DwarfAnalyzer>,
    uprobe_configs: Vec<UProbeConfig>,
    failed_targets: Vec<FailedTarget>, // Track failed compilation attempts
    binary_path_hint: Option<String>,
    current_trace_id: u32, // Current trace_id counter (increments for each uprobe)
    compile_options: crate::CompileOptions, // Compilation options (save + eBPF map config)
}

impl<'a> AstCompiler<'a> {
    pub fn new(
        process_analyzer: Option<&'a ghostscope_dwarf::DwarfAnalyzer>,
        binary_path_hint: Option<String>,
        starting_trace_id: u32,
        compile_options: crate::CompileOptions,
    ) -> Self {
        Self {
            process_analyzer,
            uprobe_configs: Vec::new(),
            failed_targets: Vec::new(),
            binary_path_hint,
            current_trace_id: starting_trace_id,
            compile_options,
        }
    }

    /// Main entry point: compile AST with integrated DWARF queries and code generation
    pub fn compile_program(
        &mut self,
        program: &Program,
        pid: Option<u32>,
    ) -> Result<CompilationResult, CompileError> {
        info!(
            "Starting unified AST compilation with {} statements",
            program.statements.len()
        );

        // AST will be saved immediately when we know the target details in generate_ebpf_for_target

        if program.statements.is_empty() {
            return Err(CompileError::Other(
                "script must contain at least one top-level trace statement".to_string(),
            ));
        }

        // Single-pass traversal: process each statement immediately
        // Continue processing even if some trace points fail
        let mut successful_trace_points = 0;
        let mut failed_trace_points = 0;
        let mut first_error: Option<String> = None;

        for (index, stmt) in program.statements.iter().enumerate() {
            match stmt {
                Statement::TracePoint { pattern, body } => {
                    debug!("Processing trace point {}: {:?}", index, pattern);
                    match self.process_trace_point(pattern, body, pid, index) {
                        Ok(_) => {
                            successful_trace_points += 1;
                            info!(
                                "✓ Successfully processed trace point {}: {:?}",
                                index, pattern
                            );
                        }
                        Err(e) => {
                            failed_trace_points += 1;
                            let error_msg = e.user_message().into_owned();
                            error!(
                                "❌ Failed to process trace point {}: {:?} - Error: {}",
                                index, pattern, error_msg
                            );

                            // Save first error for detailed error message
                            if first_error.is_none() {
                                first_error = Some(error_msg.clone());
                            }

                            // Check if failed_targets was already populated by process_trace_point
                            // (e.g., when all addresses failed for a function)
                            // If not, add a general failed target entry
                            let has_failed_for_this_pattern =
                                self.failed_targets.iter().any(|ft| match pattern {
                                    TracePattern::FunctionName(name) => ft.target_name == *name,
                                    TracePattern::SourceLine {
                                        file_path,
                                        line_number,
                                    } => ft.target_name == format!("{file_path}:{line_number}"),
                                    TracePattern::Address(addr) => {
                                        ft.target_name == format!("0x{addr:x}")
                                            && ft.pc_address == *addr
                                    }
                                    TracePattern::AddressInModule { module, address } => {
                                        ft.target_name == format!("{module}:0x{address:x}")
                                            && ft.pc_address == *address
                                    }
                                    _ => false,
                                });

                            if !has_failed_for_this_pattern {
                                let target_name = match pattern {
                                    TracePattern::FunctionName(name) => name.clone(),
                                    TracePattern::SourceLine {
                                        file_path,
                                        line_number,
                                    } => format!("{file_path}:{line_number}"),
                                    TracePattern::Address(addr) => format!("0x{addr:x}"),
                                    TracePattern::AddressInModule { module, address } => {
                                        format!("{module}:0x{address:x}")
                                    }
                                    _ => format!("trace_point_{index}"),
                                };
                                let pc_address = match pattern {
                                    TracePattern::Address(addr) => *addr,
                                    TracePattern::AddressInModule { address, .. } => *address,
                                    _ => 0,
                                };

                                self.failed_targets.push(FailedTarget {
                                    target_name,
                                    pc_address,
                                    error_message: error_msg,
                                });
                            }
                        }
                    }
                }
                _ => {
                    let message = Self::top_level_statement_error(stmt);
                    error!("{message}");
                    return Err(CompileError::Other(message));
                }
            }
        }

        if successful_trace_points > 0 && failed_trace_points == 0 {
            info!(
                "All {} trace points processed successfully",
                successful_trace_points
            );
        } else if successful_trace_points > 0 && failed_trace_points > 0 {
            warn!(
                "Partial success: {} trace points successful, {} failed",
                successful_trace_points, failed_trace_points
            );
        } else if failed_trace_points > 0 {
            // All trace points failed - return error with first failure reason
            error!("All {} trace points failed to process", failed_trace_points);
            return Err(CompileError::Other(
                self.format_all_trace_points_failed_error(first_error),
            ));
        }

        // Generate target info summary
        let target_info = self.generate_target_info_summary();

        info!(
            "Compilation completed: {} uprobe configs generated",
            self.uprobe_configs.len()
        );

        let trace_count = self.uprobe_configs.len();
        Ok(CompilationResult {
            uprobe_configs: std::mem::take(&mut self.uprobe_configs),
            failed_targets: std::mem::take(&mut self.failed_targets),
            trace_count,
            target_info,
            next_available_trace_id: self.current_trace_id,
        })
    }

    fn format_all_trace_points_failed_error(&self, first_error: Option<String>) -> String {
        let mut message = first_error.unwrap_or_else(|| "All trace points failed".to_string());
        if self.failed_targets.is_empty() {
            return message;
        }

        message.push_str("\n\nFailed targets:\n");
        for failed in &self.failed_targets {
            let _ = writeln!(
                message,
                "  - {} at 0x{:x}: {}",
                failed.target_name, failed.pc_address, failed.error_message
            );
        }
        message.push_str("\nTip: fix the reported compile-time errors above.");
        message
    }

    fn configured_target_path(&self) -> Option<&str> {
        self.compile_options
            .target_binary_path
            .as_deref()
            .map(str::trim)
            .filter(|path| !path.is_empty())
    }

    fn top_level_statement_error(statement: &Statement) -> String {
        let kind = match statement {
            Statement::Print(_) => "print",
            Statement::Backtrace(_) => "backtrace",
            Statement::Expr(_) => "expression",
            Statement::VarDeclaration { .. } | Statement::AliasDeclaration { .. } => "let",
            Statement::If { .. } => "if",
            Statement::Block(_) => "block",
            Statement::TracePoint { .. } => "trace",
        };
        format!(
            "top-level {kind} statement is not allowed in a script file; put executable statements inside a trace block, for example: trace <target> {{ ... }}"
        )
    }

    /// Process a trace point: resolve target + generate eBPF in one step
    fn process_trace_point(
        &mut self,
        pattern: &TracePattern,
        statements: &[Statement],
        pid: Option<u32>,
        index: usize,
    ) -> Result<(), CompileError> {
        match pattern {
            TracePattern::SourceLine {
                file_path,
                line_number,
            } => {
                let analyzer = self.process_analyzer.ok_or_else(|| {
                    CompileError::Other(
                        "No process analyzer available to resolve source line".to_string(),
                    )
                })?;
                let target_path = self.configured_target_path();
                let source_line = analyzer
                    .resolve_source_line_addresses_best_effort(
                        analyzer.source_line_candidates(file_path),
                        *line_number,
                        target_path,
                    )
                    .map_err(|e| CompileError::Other(e.to_string()))?;
                let module_addresses = source_line.addresses;

                if source_line.raw_address_count > 0 && module_addresses.is_empty() {
                    let target = target_path.unwrap_or("<unknown>");
                    return Err(CompileError::Other(format!(
                        "No addresses resolved for source line {file_path}:{line_number} in -t target '{target}'. When -t and -p are combined, -t takes precedence for trace target resolution."
                    )));
                }
                if module_addresses.is_empty() {
                    let detailed = analyzer.describe_source_line_failure(file_path, *line_number);
                    return Err(CompileError::Other(detailed));
                }

                debug!(
                    "Resolved {}:{} to {} address(es) for trace point {}",
                    file_path,
                    line_number,
                    module_addresses.len(),
                    index
                );

                // Validate optional single-index selection (1-based)
                if let Some(idx) = self.compile_options.selected_index {
                    if idx == 0 || idx > module_addresses.len() {
                        return Err(CompileError::Other(format!(
                            "Selected index {idx} is out of range for {file_path}:{line_number} (valid 1..={}). Use 'info' to view indices.",
                            module_addresses.len()
                        )));
                    }
                }

                // Optional single-index filter (1-based); otherwise process all
                let mut successful_addresses = 0;
                let mut failed_addresses = 0;
                // Iterate with indices (1-based) so we can propagate the global address index
                let iterator: Box<dyn Iterator<Item = (usize, &ghostscope_dwarf::ModuleAddress)>> =
                    if let Some(idx) = self.compile_options.selected_index {
                        let i = idx - 1; // safe due to validation above
                        Box::new(std::iter::once((idx, &module_addresses[i])))
                    } else {
                        Box::new(module_addresses.iter().enumerate().map(|(i, m)| (i + 1, m)))
                    };

                for (global_idx, module_address) in iterator {
                    // Convert DWARF PC (vaddr) to ELF file offset for uprobe
                    let file_off = self.process_analyzer.as_ref().and_then(|an| {
                        an.vaddr_to_file_offset(&module_address.module_path, module_address.address)
                    });

                    let target_info = ResolvedTarget {
                        function_name: Some(format!("{file_path}:{line_number}")),
                        // Keep function_address as DWARF PC for compile-time DWARF queries
                        function_address: Some(module_address.address),
                        binary_path: module_address.module_path.to_string_lossy().to_string(),
                        // Attach with absolute file offset if conversion succeeded
                        uprobe_offset: file_off,
                        pattern: pattern.clone(),
                    };

                    match self.generate_ebpf_for_target(
                        &target_info,
                        statements,
                        pid,
                        Some(global_idx),
                    ) {
                        Ok(uprobe_config) => {
                            self.uprobe_configs.push(uprobe_config);
                            successful_addresses += 1;
                            info!(
                                "✓ Successfully generated eBPF for {}:{} at 0x{:x}",
                                file_path, line_number, module_address.address
                            );
                        }
                        Err(e) => {
                            failed_addresses += 1;
                            error!(
                                "❌ Failed to generate eBPF for {}:{} at 0x{:x}: {}",
                                file_path, line_number, module_address.address, e
                            );

                            // Record this failed target
                            self.failed_targets.push(FailedTarget {
                                target_name: format!("{file_path}:{line_number}"),
                                pc_address: module_address.address,
                                error_message: e.user_message().into_owned(),
                            });

                            // Continue processing other addresses
                        }
                    }
                }

                // Log summary for this trace point
                if successful_addresses > 0 && failed_addresses == 0 {
                    info!(
                        "All {} addresses for {}:{} processed successfully",
                        successful_addresses, file_path, line_number
                    );
                } else if successful_addresses > 0 && failed_addresses > 0 {
                    warn!(
                        "Partial success for {}:{}: {} successful, {} failed addresses",
                        file_path, line_number, successful_addresses, failed_addresses
                    );
                } else {
                    error!(
                        "All {} addresses for {}:{} failed to process",
                        failed_addresses, file_path, line_number
                    );
                    // Don't return error here - let the caller decide based on overall results
                }
                Ok(())
            }
            TracePattern::Address(addr) => {
                let analyzer = self.process_analyzer.ok_or_else(|| {
                    CompileError::Other(
                        "No process analyzer available to resolve address".to_string(),
                    )
                })?;
                let module_path = analyzer
                    .resolve_address_module(
                        None,
                        self.configured_target_path(),
                        ModuleDefaultPolicy::MainExecutableOrSingleSharedLibrary,
                    )
                    .map_err(|e| CompileError::Other(e.to_string()))?;

                // Convert DWARF PC (vaddr) to ELF file offset for uprobe
                let file_off = analyzer.vaddr_to_file_offset(&module_path, *addr);
                let module_path = module_path.to_string_lossy().to_string();

                if file_off.is_none() {
                    return Err(CompileError::Other(format!(
                        "Address 0x{addr:x} is not within a loadable segment of '{module_path}' (cannot compute file offset)"
                    )));
                }

                let target_info = ResolvedTarget {
                    function_name: None,
                    function_address: Some(*addr),
                    binary_path: module_path,
                    uprobe_offset: file_off,
                    pattern: pattern.clone(),
                };

                match self.generate_ebpf_for_target(&target_info, statements, pid, None) {
                    Ok(uprobe_config) => {
                        self.uprobe_configs.push(uprobe_config);
                        info!("✓ Successfully generated eBPF for address 0x{:x}", addr);
                        Ok(())
                    }
                    Err(e) => {
                        let error_msg = e.user_message().into_owned();
                        error!(
                            "❌ Failed to generate eBPF for address 0x{:x}: {}",
                            addr, error_msg
                        );
                        self.failed_targets.push(FailedTarget {
                            target_name: format!("0x{addr:x}"),
                            pc_address: *addr,
                            error_message: error_msg,
                        });
                        Err(e)
                    }
                }
            }
            TracePattern::AddressInModule { module, address } => {
                let analyzer = self.process_analyzer.ok_or_else(|| {
                    CompileError::Other(
                        "No process analyzer available to resolve module".to_string(),
                    )
                })?;
                let module_path = analyzer
                    .resolve_address_module(
                        Some(module),
                        self.configured_target_path(),
                        ModuleDefaultPolicy::MainExecutableOrSingleSharedLibrary,
                    )
                    .map_err(|e| CompileError::Other(e.to_string()))?;

                // Convert DWARF PC (vaddr) to ELF file offset for uprobe
                let file_off = analyzer.vaddr_to_file_offset(&module_path, *address);
                let module_path = module_path.to_string_lossy().to_string();

                if file_off.is_none() {
                    return Err(CompileError::Other(format!(
                        "Address 0x{address:x} is not within a loadable segment of '{module_path}' (cannot compute file offset)"
                    )));
                }

                let target_info = ResolvedTarget {
                    function_name: None,
                    function_address: Some(*address),
                    binary_path: module_path,
                    uprobe_offset: file_off,
                    pattern: pattern.clone(),
                };

                match self.generate_ebpf_for_target(&target_info, statements, pid, None) {
                    Ok(uprobe_config) => {
                        self.uprobe_configs.push(uprobe_config);
                        info!(
                            "✓ Successfully generated eBPF for module-qualified address {}:0x{:x}",
                            module, address
                        );
                        Ok(())
                    }
                    Err(e) => {
                        let error_msg = e.user_message().into_owned();
                        error!(
                            "❌ Failed to generate eBPF for module-qualified address {}:0x{:x}: {}",
                            module, address, error_msg
                        );
                        self.failed_targets.push(FailedTarget {
                            target_name: format!("{module}:0x{address:x}"),
                            pc_address: *address,
                            error_message: error_msg,
                        });
                        Err(e)
                    }
                }
            }
            TracePattern::FunctionName(func_name) => {
                // Resolve all addresses for the function name and generate per-PC programs
                let module_addresses = if let Some(analyzer) = self.process_analyzer {
                    analyzer.lookup_function_addresses(func_name)
                } else {
                    Vec::new()
                };

                if module_addresses.is_empty() {
                    // Strict behavior: fail this trace point immediately instead of skipping silently
                    return Err(CompileError::Other(format!(
                        "No addresses resolved for function '{func_name}' - function not found in debug symbols"
                    )));
                }

                let original_address_count = module_addresses.len();
                let target_path = self.configured_target_path();
                let module_addresses = self
                    .process_analyzer
                    .ok_or_else(|| {
                        CompileError::Other(
                            "No process analyzer available to resolve -t target".to_string(),
                        )
                    })?
                    .filter_module_addresses_to_target(module_addresses, target_path)
                    .map_err(|e| CompileError::Other(e.to_string()))?;
                if original_address_count > 0 && module_addresses.is_empty() {
                    let target = target_path.unwrap_or("<unknown>");
                    return Err(CompileError::Other(format!(
                        "No addresses resolved for function '{func_name}' in -t target '{target}'. When -t and -p are combined, -t takes precedence for trace target resolution."
                    )));
                }

                let total_addresses: usize = module_addresses.len();
                debug!(
                    "Resolved function '{}' to {} address(es) across {} modules",
                    func_name,
                    total_addresses,
                    module_addresses.len()
                );

                // Validate optional single-index selection (1-based)
                if let Some(idx) = self.compile_options.selected_index {
                    if idx == 0 || idx > module_addresses.len() {
                        return Err(CompileError::Other(format!(
                            "Selected index {idx} is out of range for function '{func_name}' (valid 1..={}). Use 'info function {func_name}' to view indices.",
                            module_addresses.len()
                        )));
                    }
                }

                // We may need analyzer again to compute precise uprobe offsets
                // Optional single-index filter (1-based); otherwise process all addresses
                let mut successful_addresses = 0;
                let mut failed_addresses = 0;

                // Iterate with indices (1-based) so we can propagate the global address index
                let iterator: Box<dyn Iterator<Item = (usize, &ghostscope_dwarf::ModuleAddress)>> =
                    if let Some(idx) = self.compile_options.selected_index {
                        let i = idx - 1; // safe due to validation above
                        Box::new(std::iter::once((idx, &module_addresses[i])))
                    } else {
                        Box::new(module_addresses.iter().enumerate().map(|(i, m)| (i + 1, m)))
                    };

                for (global_idx, module_address) in iterator {
                    // Convert DWARF function address (vaddr) to ELF file offset for uprobe attach
                    let file_off = self.process_analyzer.as_ref().and_then(|an| {
                        an.vaddr_to_file_offset(&module_address.module_path, module_address.address)
                    });

                    let target_info = ResolvedTarget {
                        function_name: Some(func_name.clone()),
                        function_address: Some(module_address.address),
                        binary_path: module_address.module_path.to_string_lossy().to_string(),
                        uprobe_offset: file_off,
                        pattern: pattern.clone(),
                    };

                    match self.generate_ebpf_for_target(
                        &target_info,
                        statements,
                        pid,
                        Some(global_idx),
                    ) {
                        Ok(uprobe_config) => {
                            self.uprobe_configs.push(uprobe_config);
                            successful_addresses += 1;
                            info!(
                                "✓ Successfully generated eBPF for function '{}' at 0x{:x}",
                                func_name, module_address.address
                            );
                        }
                        Err(e) => {
                            failed_addresses += 1;
                            error!(
                                "❌ Failed to generate eBPF for function '{}' at 0x{:x}: {}",
                                func_name, module_address.address, e
                            );

                            // Record this failed target
                            self.failed_targets.push(FailedTarget {
                                target_name: func_name.clone(),
                                pc_address: module_address.address,
                                error_message: e.user_message().into_owned(),
                            });

                            // Continue processing other addresses
                        }
                    }
                }

                // Log summary for this trace point
                if successful_addresses > 0 && failed_addresses == 0 {
                    info!(
                        "All {} addresses for function '{}' processed successfully",
                        successful_addresses, func_name
                    );
                    Ok(())
                } else if successful_addresses > 0 && failed_addresses > 0 {
                    warn!(
                        "Partial success for function '{}': {} successful, {} failed addresses",
                        func_name, successful_addresses, failed_addresses
                    );
                    Ok(())
                } else {
                    // All addresses failed to process — record failures already captured above
                    // Defer final error shaping to the caller based on aggregated results
                    error!(
                        "All {} addresses for function '{}' failed to process",
                        failed_addresses, func_name
                    );
                    Ok(())
                }
            }
            _ => {
                unimplemented!();
            }
        }
    }

    /// Generate eBPF bytecode for resolved target
    fn generate_ebpf_for_target(
        &mut self,
        target: &ResolvedTarget,
        statements: &[Statement],
        pid: Option<u32>,
        resolved_address_index: Option<usize>,
    ) -> Result<UProbeConfig, CompileError> {
        let context = Context::create();

        // Allocate trace_id for this uprobe
        let assigned_trace_id = self.current_trace_id;
        self.current_trace_id += 1;

        // Generate unified eBPF function name using the assigned trace_id
        let ebpf_function_name = self.generate_unified_function_name(target, assigned_trace_id);
        let compile_options = self.compile_options.clone();
        let binary_path_hint = self.binary_path_hint.clone();

        info!(
            "Generating eBPF code for '{}' (function: {})",
            target.function_name.as_deref().unwrap_or("unknown"),
            ebpf_function_name
        );

        // Save AST immediately when we know the target details (before generating LLVM IR)
        if let Some(compile_options) = self.get_compile_options() {
            if compile_options.save_ast {
                let ast_filename = self.generate_filename(target, assigned_trace_id, "txt");
                // Create a Program from statements to save
                let program = Program {
                    statements: statements.to_vec(),
                };
                if let Err(e) = self.save_ast_to_file(&program, &ast_filename) {
                    warn!("Failed to save AST to {}: {}", ast_filename, e);
                } else {
                    info!("Saved AST to: {}", ast_filename);
                }
            }
        }

        // Use the eBPF context implementation with full AST compilation.
        let mut codegen = crate::ebpf::context::EbpfContext::new_with_process_analyzer(
            &context,
            &ebpf_function_name,
            self.process_analyzer,
            Some(assigned_trace_id),
            &self.compile_options,
        )
        .map_err(|e| CompileError::LLVM(format!("Failed to create new codegen: {e}")))?;

        // Set compile-time context for DWARF queries
        if let Some(function_address) = target.function_address {
            codegen.set_compile_time_context(function_address, target.binary_path.clone());
        }

        info!(
            "Compiling full AST program with {} statements",
            statements.len()
        );

        // Use full AST compilation
        let (_main_function, trace_context) = codegen
            .compile_program(
                &crate::script::ast::Program { statements: vec![] }, // Empty program - statements passed separately
                &ebpf_function_name,
                statements,
                pid,
                target.function_address,
                Some(&target.binary_path),
            )
            .map_err(CompileError::CodeGen)?;

        info!(
            "Generated TraceContext for '{}' with {} strings and {} variables",
            ebpf_function_name,
            trace_context.string_count(),
            trace_context.variable_name_count()
        );

        let module = codegen.get_module();

        // Generate eBPF bytecode from LLVM module
        let ebpf_bytecode = Self::generate_ebpf_bytecode(
            module,
            &ebpf_function_name,
            target,
            assigned_trace_id,
            &compile_options,
            binary_path_hint.as_deref(),
        )?;

        // Use the TraceContext returned from compile_program (no need to get it again)

        Ok(UProbeConfig {
            trace_pattern: target.pattern.clone(),
            binary_path: target.binary_path.clone(),
            function_name: target.function_name.clone(),
            function_address: target.function_address,
            uprobe_offset: target.uprobe_offset,
            target_pid: pid,
            ebpf_bytecode,
            ebpf_function_name,
            assigned_trace_id,
            trace_context,
            backtrace_unwind_rows: codegen.backtrace_unwind_rows.clone(),
            backtrace_module_row_ranges: codegen
                .backtrace_module_row_ranges
                .iter()
                .map(|entry| (entry.cookie, entry.range))
                .collect(),
            backtrace_tail_call_program: codegen.backtrace_tail_call_program(),
            resolved_address_index,
        })
    }

    /// Generate summary of all targets for reporting
    fn generate_target_info_summary(&self) -> String {
        if self.uprobe_configs.is_empty() {
            return "no_targets".to_string();
        }

        let first_target = &self.uprobe_configs[0];
        match &first_target.function_name {
            Some(name) => name.clone(),
            None => format!("addr_0x{:x}", first_target.function_address.unwrap_or(0)),
        }
    }

    /// Generate unified eBPF function name for all contexts
    ///
    /// This is the SINGLE source of truth for eBPF function naming.
    /// All other naming logic should use this method to ensure consistency.
    /// Calculate 8-digit hex hash for module path with logging
    fn calculate_module_hash(&self, module_path: &str) -> String {
        let effective_path = self.effective_binary_path(module_path);
        let mut hasher = DefaultHasher::new();
        effective_path.hash(&mut hasher);
        let hash = hasher.finish();
        let truncated = (hash & 0xFFFF_FFFF) as u32;
        let hash_hex = format!("{truncated:08x}");

        info!("Module hash calculated: {} -> {}", effective_path, hash_hex);
        hash_hex
    }

    /// Generate unified function name with format: ghostscope_{module_hash}_{address_hex}_{trace_id}
    fn generate_unified_function_name(&self, target: &ResolvedTarget, trace_id: u32) -> String {
        let module_hash = self.calculate_module_hash(&target.binary_path);
        let effective_path = self.effective_binary_path(&target.binary_path);
        let address_hex = if let Some(addr) = target.function_address {
            format!("{addr:x}")
        } else {
            "unknown".to_string()
        };

        let function_name = format!("ghostscope_{module_hash}_{address_hex}_trace{trace_id}");
        info!(
            "Generated eBPF function name: {} (module: {}, address: 0x{}, trace_id: {})",
            function_name, effective_path, address_hex, trace_id
        );

        function_name
    }

    /// Get save options (helper method)
    fn get_compile_options(&self) -> Option<&crate::CompileOptions> {
        Some(&self.compile_options)
    }

    /// Pick a binary path, falling back to compiler hint when the resolved target is empty
    fn effective_binary_path<'b>(&'b self, target_path: &'b str) -> Cow<'b, str> {
        if target_path.is_empty() {
            if let Some(hint) = &self.binary_path_hint {
                Cow::Owned(hint.clone())
            } else {
                Cow::Borrowed("unknown")
            }
        } else {
            Cow::Borrowed(target_path)
        }
    }

    /// Generate filename for output files
    fn generate_filename(&self, target: &ResolvedTarget, trace_id: u32, extension: &str) -> String {
        let module_hash = self.calculate_module_hash(&target.binary_path);
        let address_hex = if let Some(addr) = target.function_address {
            format!("{addr:x}")
        } else {
            "unknown".to_string()
        };

        format!("gs_{module_hash}_{address_hex}_trace{trace_id}.{extension}")
    }

    fn generate_filename_with_hint(
        target: &ResolvedTarget,
        trace_id: u32,
        extension: &str,
        binary_path_hint: Option<&str>,
    ) -> String {
        let effective_path = if target.binary_path.is_empty() {
            binary_path_hint.unwrap_or("unknown")
        } else {
            target.binary_path.as_str()
        };
        let mut hasher = DefaultHasher::new();
        effective_path.hash(&mut hasher);
        let module_hash = format!("{:08x}", (hasher.finish() & 0xFFFF_FFFF) as u32);
        let address_hex = if let Some(addr) = target.function_address {
            format!("{addr:x}")
        } else {
            "unknown".to_string()
        };

        format!("gs_{module_hash}_{address_hex}_trace{trace_id}.{extension}")
    }

    /// Generate eBPF bytecode from LLVM module
    fn generate_ebpf_bytecode(
        module: &inkwell::module::Module,
        function_name: &str,
        target: &ResolvedTarget,
        assigned_trace_id: u32,
        compile_options: &crate::CompileOptions,
        binary_path_hint: Option<&str>,
    ) -> Result<Vec<u8>, CompileError> {
        use inkwell::targets::{FileType, Target, TargetTriple};
        use inkwell::OptimizationLevel;

        if compile_options.save_llvm_ir {
            let filename = Self::generate_filename_with_hint(
                target,
                assigned_trace_id,
                "ll",
                binary_path_hint,
            );
            if let Err(e) = module.print_to_file(&filename) {
                warn!("Failed to save LLVM IR to {}: {}", filename, e);
            } else {
                info!("Saved LLVM IR to: {}", filename);
            }
        }
        info!("Successfully generated LLVM module for {}", function_name);

        // Get target triple
        let triple = TargetTriple::create("bpf-pc-linux");
        info!("Created target triple: bpf-pc-linux for {}", function_name);

        // Get BPF target
        let llvm_target = Target::from_triple(&triple).map_err(|e| {
            error!("Failed to get target for {}: {}", function_name, e);
            CompileError::LLVM(format!("Failed to get target for {function_name}: {e}"))
        })?;
        info!("Successfully got LLVM target for {}", function_name);

        // Create target machine
        let target_machine = llvm_target
            .create_target_machine(
                &triple,
                "generic", // CPU
                "+alu32",  // Enable BPF ALU32 instructions
                OptimizationLevel::Default,
                inkwell::targets::RelocMode::PIC,
                inkwell::targets::CodeModel::Small,
            )
            .ok_or_else(|| {
                error!("Failed to create target machine for {}", function_name);
                CompileError::LLVM(format!(
                    "Failed to create target machine for {function_name}"
                ))
            })?;
        info!("Successfully created target machine for {}", function_name);

        // Validate module before generating object code
        info!("Validating LLVM module for {}...", function_name);
        if let Err(llvm_errors) = module.verify() {
            error!(
                "LLVM module validation failed for {}: {}",
                function_name, llvm_errors
            );
            return Err(CompileError::LLVM(format!(
                "Module validation failed for {function_name}: {llvm_errors}"
            )));
        }
        info!("Module validation passed for {}", function_name);

        // Generate eBPF object file
        info!("Generating eBPF object file for {}...", function_name);
        info!("About to call LLVM write_to_memory_buffer...");

        let object_code = {
            // Add a flush to ensure logs are written before potential crash
            use std::io::Write;
            let _ = std::io::stderr().flush();
            let _ = std::io::stdout().flush();

            info!("Calling target_machine.write_to_memory_buffer...");
            match target_machine.write_to_memory_buffer(module, FileType::Object) {
                Ok(code) => {
                    info!("Successfully generated object code for {}", function_name);
                    code
                }
                Err(e) => {
                    error!("LLVM compilation failed for {}: {}", function_name, e);
                    error!("This might be due to unsupported eBPF instructions or invalid LLVM IR");

                    return Err(CompileError::LLVM(format!(
                        "eBPF compilation failed for {function_name}: {e}. This often indicates unsupported instructions or invalid IR."
                    )));
                }
            }
        };

        info!(
            "Successfully generated object code for {}! Size: {}",
            function_name,
            object_code.get_size()
        );

        // Convert to Vec<u8>
        let bytecode = object_code.as_slice().to_vec();

        // Save eBPF object file and AST if requested
        if compile_options.save_ebpf {
            let filename =
                Self::generate_filename_with_hint(target, assigned_trace_id, "o", binary_path_hint);
            if let Err(e) = std::fs::write(&filename, &bytecode) {
                warn!("Failed to save eBPF object to {}: {}", filename, e);
            } else {
                info!("Saved eBPF object to: {}", filename);
            }
        }

        // AST has already been saved earlier in generate_ebpf_for_target
        Ok(bytecode)
    }

    /// Save AST to file
    fn save_ast_to_file(
        &mut self,
        program: &crate::script::ast::Program,
        filename: &str,
    ) -> Result<(), CompileError> {
        let mut ast_content = String::new();
        ast_content.push_str("=== AST Tree ===\n");
        ast_content.push_str("Program:\n");
        for (i, stmt) in program.statements.iter().enumerate() {
            ast_content.push_str(&format!("  Statement {i}: {stmt:?}\n"));
        }
        ast_content.push_str("=== End AST Tree ===\n");

        std::fs::write(filename, ast_content).map_err(|e| {
            CompileError::Other(format!("Failed to save AST file '{filename}': {e}"))
        })?;

        Ok(())
    }
}