def cu_offset_to_path_map(config: Config, dwarf_info):
"""Return a map from Dwarf compilation unit offsets to source paths."""
prefixes = config.get_re('collect.prefix')
address_map = {}
for compilation_unit in dwarf_info.iter_CUs():
path = pathlib.Path(compilation_unit.get_top_DIE().get_full_path())
source = simplify_source(str(path.resolve()), prefixes)
address_map[compilation_unit.cu_offset] = source
return address_map
def postprocess_symbols(config: Config, symbols: SymbolDF) -> SymbolDF:
"""Postprocess a symbol table after collecting from one source.
If the symbol table contains FILE symbols, they will be removed and
replaced by a 'file' column on other symbols.
If the symbol table contains ARM mode symbols, they will be removed
and replaced by an 'arm' column on other symbols.
"""
files = []
arms = []
arm_symbols = {}
current_file = ''
current_arm = ''
has_file = False
if config['collect.prefix-file']:
prefixes = config.get_re('collect.prefix')
else:
prefixes = None
if 'type' in symbols.columns:
for symbol in symbols.itertuples():
if symbol.type == 'FILE':
has_file = True
current_file = symbol.symbol
if prefixes:
current_file = simplify_source(current_file, prefixes)
elif symbol.type == 'NOTYPE':
if symbol.symbol.startswith('$'):
if current_arm or symbol.symbol in ARM_SPECIAL_SYMBOLS:
current_arm = symbol.symbol
arm_symbols[current_arm] = True
files.append(current_file)
arms.append(current_arm)
if has_file:
symbols['file'] = files
if current_arm:
symbols['arm'] = arms
if has_file:
symbols = symbols[symbols['type'] != 'FILE']
if current_arm:
syms = arm_symbols.keys()
symbols = symbols[~symbols.symbol.isin(syms)]
return symbols
def read_symbols(config: Config, filename: str) -> SymbolDF:
"""Read a binary's symbol map using bloaty."""
column_map = {
'compileunits': 'cu',
'sections': 'section',
'symbols': 'symbol',
'vmsize': 'size',
}
process = memdf.util.subprocess.run_tool_pipe(config, [
'bloaty', '--tsv', '--demangle=none', '-n', '0', '-d',
'compileunits,sections,symbols', filename
])
if not process or not process.stdout:
return SymbolDF()
df = pd.read_table(io.TextIOWrapper(process.stdout, newline=os.linesep),
usecols=list(column_map.keys()),
dtype=SymbolDF.dtype,
na_filter=False)
df.rename(inplace=True, columns=column_map)
prefixes = config.get_re('collect.prefix')
df['cu'] = df['cu'].apply(lambda s: simplify_source(s, prefixes))
return df
\s(?P<kind>\S)
\s(?P<symbol>\S+)
(\t(?P<source>\S+):(?P<line>\d+))?
""", re.VERBOSE)
columns = ['address', 'kind', 'symbol', 'cu']
rows = []
while line := text.readline():
if not (match := decoder.match(line.rstrip())):
continue
if a := match.group('address'):
address = int(a, 16)
else:
address = 0
source = match.group('source') or ''
if source:
source = simplify_source(source, prefixes)
rows.append(
[address,
match.group('kind'),
match.group('symbol'), source])
return SymbolSourceDF(rows, columns=columns)
def add_cu(config: Config, filename: str, symbols: SymbolDF) -> SymbolDF:
"""Add a 'cu' (compilation unit) column to a symbol table."""
sources = read_sources(config, filename).set_index(['symbol', 'address'])
symbols = pd.merge(symbols, sources, on=('symbol', 'address'), how='left')
symbols.fillna({'cu': ''}, inplace=True)
return symbols