#!/bin/sh
# Normalization scripts for reproducible builds
# Strips non-deterministic content from build artifacts
#
# Usage: normalize.sh <directory>

set -eu

DIR="${1:-.}"

log() {
    echo "[normalize] $*" >&2
}

# Strip timestamps from __DATE__ and __TIME__ macros
strip_date_time() {
    log "Stripping date/time macros..."
    # Already handled by SOURCE_DATE_EPOCH in modern GCC
}

# Normalize build paths
normalize_paths() {
    log "Normalizing build paths..."
    # Handled by -fdebug-prefix-map
}

# Normalize ar archives for deterministic ordering
normalize_archives() {
    log "Normalizing ar archives..."
    find "$DIR" -name "*.a" -type f | while read -r archive; do
        if ar --version 2>&1 | grep -q "GNU ar"; then
            # GNU ar with deterministic mode
            ar -rcsD "$archive.tmp" "$archive" && mv "$archive.tmp" "$archive" 2>/dev/null || true
        fi
    done
}

# Strip debug sections that contain non-deterministic info
strip_debug_timestamps() {
    log "Stripping debug timestamps..."
    find "$DIR" -type f \( -name "*.o" -o -name "*.so" -o -name "*.so.*" -o -executable \) | while read -r obj; do
        # Check if ELF
        file "$obj" 2>/dev/null | grep -q "ELF" || continue
        
        # Strip build-id if not needed (we regenerate it)
        # objcopy --remove-section=.note.gnu.build-id "$obj" 2>/dev/null || true
        
        # Remove timestamps from DWARF debug info
        # This is typically handled by SOURCE_DATE_EPOCH
    done
}

# Normalize tar archives
normalize_tars() {
    log "Normalizing tar archives..."
    # When creating tars, use:
    # tar --sort=name --mtime="@${SOURCE_DATE_EPOCH}" --owner=0 --group=0 --numeric-owner
}

# Run all normalizations
normalize_paths
normalize_archives
strip_debug_timestamps

log "Normalization complete"