def __init__(self,
project,
binary_images,
referenced_images=None,
cpu_name=None):
self.symsynd_symbolizer = make_symbolizer(
project, binary_images, referenced_images=referenced_images)
# This is a duplication from symsynd. The reason is that symsynd
# will only load images that it can find dsyms for but we also
# have system symbols which there are no dsyms for.
self._image_addresses = []
self.images = {}
for img in binary_images:
img_addr = parse_addr(img['image_addr'])
self._image_addresses.append(img_addr)
self.images[img_addr] = img
self._image_addresses.sort()
# This should always succeed but you never quite know.
self.cpu_name = cpu_name
if self.cpu_name is None:
for img in six.itervalues(self.images):
cpu_name = get_cpu_name(img['cpu_type'], img['cpu_subtype'])
if self.cpu_name is None:
self.cpu_name = cpu_name
elif self.cpu_name != cpu_name:
self.cpu_name = None
break
def find_system_symbol(img, instruction_addr, system_info=None):
"""Finds a system symbol."""
return DSymSymbol.objects.lookup_symbol(instruction_addr=instruction_addr,
image_addr=img['image_addr'],
uuid=img['uuid'],
cpu_name=get_cpu_name(
img['cpu_type'],
img['cpu_subtype']),
object_path=img['name'],
system_info=system_info)
def symbolize_system_frame(self,
frame,
img,
sdk_info,
symbolize_inlined=False):
"""Symbolizes a frame with system symbols only."""
# This is most likely a good enough cache match even though we are
# ignoring the image here since we cache by instruction address.
#
# In some cases old clients might not send an sdk_info with it
# in which case the caching won't work.
if sdk_info is not None:
cache_key = 'ssym:%s:%s:%s:%s:%s:%s:%s' % (
frame['instruction_addr'],
get_cpu_name(img['cpu_type'], img['cpu_subtype']),
sdk_info['sdk_name'],
sdk_info['dsym_type'],
sdk_info['version_major'],
sdk_info['version_minor'],
sdk_info['version_patchlevel'],
)
symbol = cache.get(cache_key)
else:
cache_key = None
symbol = None
if symbol is None:
symbol = find_system_symbol(img, frame['instruction_addr'],
sdk_info, self.cpu_name)
if symbol is None:
# Simulator frames cannot be symbolicated
if self._is_simulator_frame(frame, img):
type = EventError.NATIVE_SIMULATOR_FRAME
else:
type = EventError.NATIVE_MISSING_SYSTEM_DSYM
raise SymbolicationFailed(type=type, image=img)
elif cache_key is not None:
cache.set(cache_key, symbol, 3600)
rv = self._process_frame(
dict(frame,
symbol_name=symbol,
filename=None,
line=0,
column=0,
uuid=img['uuid'],
object_name=img['name']), img)
# We actually do not support inline symbolication for system
# frames, so we just only ever return a single frame here. Maybe
# we can improve this in the future.
if symbolize_inlined:
return [rv]
return rv
def find_system_symbol(img, instruction_addr, system_info=None):
"""Finds a system symbol."""
return DSymSymbol.objects.lookup_symbol(
instruction_addr=instruction_addr,
image_addr=img['image_addr'],
uuid=img['uuid'],
cpu_name=get_cpu_name(img['cpu_type'],
img['cpu_subtype']),
object_path=img['name'],
system_info=system_info
)
def symbolize_system_frame(self, frame, img, sdk_info,
symbolize_inlined=False):
"""Symbolizes a frame with system symbols only."""
# This is most likely a good enough cache match even though we are
# ignoring the image here since we cache by instruction address.
#
# In some cases old clients might not send an sdk_info with it
# in which case the caching won't work.
if sdk_info is not None:
cache_key = 'ssym:%s:%s:%s:%s:%s:%s:%s' % (
frame['instruction_addr'],
get_cpu_name(img['cpu_type'], img['cpu_subtype']),
sdk_info['sdk_name'],
sdk_info['dsym_type'],
sdk_info['version_major'],
sdk_info['version_minor'],
sdk_info['version_patchlevel'],
)
symbol = cache.get(cache_key)
else:
cache_key = None
symbol = None
if symbol is None:
symbol = find_system_symbol(
img, frame['instruction_addr'], sdk_info, self.cpu_name)
if symbol is None:
# Simulator frames cannot be symbolicated
if self._is_simulator_frame(frame, img):
type = EventError.NATIVE_SIMULATOR_FRAME
else:
type = EventError.NATIVE_MISSING_SYSTEM_DSYM
raise SymbolicationFailed(
type=type,
image=img
)
elif cache_key is not None:
cache.set(cache_key, symbol, 3600)
rv = self._process_frame(dict(frame,
symbol_name=symbol, filename=None, line=0, column=0,
object_name=img['name']), img)
# We actually do not support inline symbolication for system
# frames, so we just only ever return a single frame here. Maybe
# we can improve this in the future.
if symbolize_inlined:
return [rv]
return rv
def __init__(self, message=None, type=None, image=None):
Exception.__init__(self)
self.message = six.text_type(message)
self.type = type
if image is not None:
self.image_uuid = image['uuid'].lower()
self.image_path = image['name']
self.image_name = image['name'].rsplit('/', 1)[-1]
self.image_arch = get_cpu_name(image['cpu_type'],
image['cpu_subtype'])
else:
self.image_uuid = None
self.image_name = None
self.image_path = None
self.image_arch = None
def __init__(self, message=None, type=None, image=None):
Exception.__init__(self)
self.message = six.text_type(message)
self.type = type
if image is not None:
self.image_uuid = image['uuid'].lower()
self.image_path = image['name']
self.image_name = image['name'].rsplit('/', 1)[-1]
self.image_arch = get_cpu_name(image['cpu_type'],
image['cpu_subtype'])
else:
self.image_uuid = None
self.image_name = None
self.image_path = None
self.image_arch = None
def cpu_name_from_data(data):
"""Returns the CPU name from the given data if it exists."""
device = DeviceContextType.primary_value_for_data(data)
if device:
arch = device.get('arch')
if isinstance(arch, six.string_types):
return arch
# TODO: kill this here. we want to not support that going forward
unique_cpu_name = None
images = (data.get('debug_meta') or {}).get('images') or []
for img in images:
cpu_name = get_cpu_name(img['cpu_type'], img['cpu_subtype'])
if unique_cpu_name is None:
unique_cpu_name = cpu_name
elif unique_cpu_name != cpu_name:
unique_cpu_name = None
break
return unique_cpu_name
def cpu_name_from_data(data):
"""Returns the CPU name from the given data if it exists."""
device = DeviceContextType.primary_value_for_data(data)
if device:
arch = device.get('arch')
if isinstance(arch, six.string_types):
return arch
# TODO: kill this here. we want to not support that going forward
unique_cpu_name = None
images = (data.get('debug_meta') or {}).get('images') or []
for img in images:
cpu_name = get_cpu_name(img['cpu_type'],
img['cpu_subtype'])
if unique_cpu_name is None:
unique_cpu_name = cpu_name
elif unique_cpu_name != cpu_name:
unique_cpu_name = None
break
return unique_cpu_name
def find_debug_images(dsym_paths, binary_images):
images_to_load = set()
with timedsection('iterimages0'):
for image in binary_images:
cpu_name = get_cpu_name(image['cpu_type'], image['cpu_subtype'])
if cpu_name is not None:
images_to_load.add(image['uuid'].lower())
images = {}
# Step one: load images that are named by their UUID
with timedsection('loadimages-fast'):
for uuid in list(images_to_load):
for dsym_path in dsym_paths:
fn = os.path.join(dsym_path, uuid)
if os.path.isfile(fn):
images[uuid] = fn
images_to_load.discard(uuid)
break
# Otherwise fall back to loading images from the dsym bundle. Because
# this loading strategy is pretty slow we do't actually want to use it
# unless we have a path that looks like a bundle. As a result we
# find all the paths which are bundles and then only process those.
if images_to_load:
slow_paths = []
for dsym_path in dsym_paths:
if os.path.isdir(os.path.join(dsym_path, 'Contents')):
slow_paths.append(dsym_path)
with timedsection('loadimages-slow'):
for dsym_path in slow_paths:
dwarf_base = os.path.join(dsym_path, 'Contents', 'Resources',
'DWARF')
if os.path.isdir(dwarf_base):
for fn in os.listdir(dwarf_base):
# Looks like a UUID we loaded, skip it
if fn in images:
continue
full_fn = os.path.join(dwarf_base, fn)
uuids = get_macho_uuids(full_fn)
for _, uuid in uuids:
if uuid in images_to_load:
images[uuid] = full_fn
images_to_load.discard(uuid)
rv = {}
# Now resolve all the images.
with timedsection('resolveimages'):
for image in binary_images:
cpu_name = get_cpu_name(image['cpu_type'], image['cpu_subtype'])
if cpu_name is None:
continue
uid = image['uuid'].lower()
if uid not in images:
continue
rv[image['image_addr']] = {
'uuid': uid,
'image_addr': image['image_addr'],
'dsym_path': images[uid],
'image_vmaddr': image['image_vmaddr'],
'cpu_name': cpu_name,
}
return rv
def test_cpu_names():
assert arch.get_cpu_name(12, 9) == 'armv7'
def find_debug_images(dsym_paths, binary_images):
images_to_load = set()
with timedsection('iterimages0'):
for image in binary_images:
cpu_name = get_cpu_name(image['cpu_type'],
image['cpu_subtype'])
if cpu_name is not None:
images_to_load.add(image['uuid'].lower())
images = {}
# Step one: load images that are named by their UUID
with timedsection('loadimages-fast'):
for uuid in list(images_to_load):
for dsym_path in dsym_paths:
fn = os.path.join(dsym_path, uuid)
if os.path.isfile(fn):
images[uuid] = fn
images_to_load.discard(uuid)
break
# Otherwise fall back to loading images from the dsym bundle. Because
# this loading strategy is pretty slow we do't actually want to use it
# unless we have a path that looks like a bundle. As a result we
# find all the paths which are bundles and then only process those.
if images_to_load:
slow_paths = []
for dsym_path in dsym_paths:
if os.path.isdir(os.path.join(dsym_path, 'Contents')):
slow_paths.append(dsym_path)
with timedsection('loadimages-slow'):
for dsym_path in slow_paths:
dwarf_base = os.path.join(dsym_path, 'Contents',
'Resources', 'DWARF')
if os.path.isdir(dwarf_base):
for fn in os.listdir(dwarf_base):
# Looks like a UUID we loaded, skip it
if fn in images:
continue
full_fn = os.path.join(dwarf_base, fn)
uuids = get_macho_uuids(full_fn)
for _, uuid in uuids:
if uuid in images_to_load:
images[uuid] = full_fn
images_to_load.discard(uuid)
rv = {}
# Now resolve all the images.
with timedsection('resolveimages'):
for image in binary_images:
cpu_name = get_cpu_name(image['cpu_type'],
image['cpu_subtype'])
if cpu_name is None:
continue
uid = image['uuid'].lower()
if uid not in images:
continue
rv[image['image_addr']] = {
'uuid': uid,
'image_addr': image['image_addr'],
'dsym_path': images[uid],
'image_vmaddr': image['image_vmaddr'],
'cpu_name': cpu_name,
}
return rv
