#!/bin/sh # Build a libFuzzer seed corpus from the existing tests/ data and # the bundled hyph_en_US.dic. Each seed file has the format: # # [word bytes][NUL][dictionary bytes] # # matching what fuzz_hyphen.c expects. set -eu cd "$(dirname "$0")/.." CORPUS_DIR=${1:-fuzz/corpus_seed} mkdir -p "$CORPUS_DIR" count=0 # tests/.pat + tests/.word -> one seed per word line for pat in tests/*.pat; do name=$(basename "$pat" .pat) wordfile="tests/${name}.word" [ -f "$wordfile" ] || continue line_no=0 while IFS= read -r word || [ -n "$word" ]; do line_no=$((line_no + 1)) # strip CR if present word=$(printf '%s' "$word" | tr -d '\r') out="$CORPUS_DIR/${name}_${line_no}" # concat: word, NUL, dictionary { printf '%s' "$word"; printf '\0'; cat "$pat"; } > "$out" count=$((count + 1)) done < "$wordfile" done # A few generic seeds against hyph_en_US.dic (much larger dictionary, # helps the FSM walk lots of states). if [ -f hyph_en_US.dic ]; then for w in hyphenation example creating absolutely beautifully \ antidisestablishmentarianism photograph supercalifragilistic \ "" "a" "ab" "abc" ".a" "a." ".." "a-b" "12345"; do out="$CORPUS_DIR/en_$(printf '%s' "$w" | tr -c 'a-zA-Z0-9' '_')" { printf '%s' "$w"; printf '\0'; cat hyph_en_US.dic; } > "$out" count=$((count + 1)) done fi # A pure-parser seed: empty word, full hyph_en_US dictionary. if [ -f hyph_en_US.dic ]; then cat hyph_en_US.dic > "$CORPUS_DIR/parser_only_en" count=$((count + 1)) fi # Tiny edge cases for the parser path { printf '%s' "x"; printf '\0'; printf 'UTF-8\n1.a\n'; } > "$CORPUS_DIR/tiny_boundary" { printf '%s' "x"; printf '\0'; printf 'UTF-8\n'; } > "$CORPUS_DIR/tiny_empty_dict" { printf '%s' "x"; printf '\0'; printf 'UTF-8\nNOHYPHEN \n1a1\n'; } > "$CORPUS_DIR/tiny_nohyphen_empty" { printf '%s' "x"; printf '\0'; printf 'UTF-8\nNEXTLEVEL\nUTF-8\n1a1\n'; } > "$CORPUS_DIR/tiny_nextlevel" count=$((count + 4)) echo "[corpus] wrote $count seed files to $CORPUS_DIR" du -sh "$CORPUS_DIR"