| # SPDX-License-Identifier: GPL-2.0+ |
| # Copyright (c) 2016 Google, Inc |
| # |
| # Base class for all entries |
| # |
| |
| from collections import namedtuple |
| import importlib |
| import os |
| import sys |
| |
| import fdt_util |
| import tools |
| from tools import ToHex, ToHexSize |
| import tout |
| |
| modules = {} |
| |
| our_path = os.path.dirname(os.path.realpath(__file__)) |
| |
| |
| # An argument which can be passed to entries on the command line, in lieu of |
| # device-tree properties. |
| EntryArg = namedtuple('EntryArg', ['name', 'datatype']) |
| |
| # Information about an entry for use when displaying summaries |
| EntryInfo = namedtuple('EntryInfo', ['indent', 'name', 'etype', 'size', |
| 'image_pos', 'uncomp_size', 'offset', |
| 'entry']) |
| |
| class Entry(object): |
| """An Entry in the section |
| |
| An entry corresponds to a single node in the device-tree description |
| of the section. Each entry ends up being a part of the final section. |
| Entries can be placed either right next to each other, or with padding |
| between them. The type of the entry determines the data that is in it. |
| |
| This class is not used by itself. All entry objects are subclasses of |
| Entry. |
| |
| Attributes: |
| section: Section object containing this entry |
| node: The node that created this entry |
| offset: Offset of entry within the section, None if not known yet (in |
| which case it will be calculated by Pack()) |
| size: Entry size in bytes, None if not known |
| pre_reset_size: size as it was before ResetForPack(). This allows us to |
| keep track of the size we started with and detect size changes |
| uncomp_size: Size of uncompressed data in bytes, if the entry is |
| compressed, else None |
| contents_size: Size of contents in bytes, 0 by default |
| align: Entry start offset alignment, or None |
| align_size: Entry size alignment, or None |
| align_end: Entry end offset alignment, or None |
| pad_before: Number of pad bytes before the contents, 0 if none |
| pad_after: Number of pad bytes after the contents, 0 if none |
| data: Contents of entry (string of bytes) |
| compress: Compression algoithm used (e.g. 'lz4'), 'none' if none |
| orig_offset: Original offset value read from node |
| orig_size: Original size value read from node |
| """ |
| def __init__(self, section, etype, node, name_prefix=''): |
| # Put this here to allow entry-docs and help to work without libfdt |
| global state |
| import state |
| |
| self.section = section |
| self.etype = etype |
| self._node = node |
| self.name = node and (name_prefix + node.name) or 'none' |
| self.offset = None |
| self.size = None |
| self.pre_reset_size = None |
| self.uncomp_size = None |
| self.data = None |
| self.contents_size = 0 |
| self.align = None |
| self.align_size = None |
| self.align_end = None |
| self.pad_before = 0 |
| self.pad_after = 0 |
| self.offset_unset = False |
| self.image_pos = None |
| self._expand_size = False |
| self.compress = 'none' |
| |
| @staticmethod |
| def Lookup(node_path, etype): |
| """Look up the entry class for a node. |
| |
| Args: |
| node_node: Path name of Node object containing information about |
| the entry to create (used for errors) |
| etype: Entry type to use |
| |
| Returns: |
| The entry class object if found, else None |
| """ |
| # Convert something like 'u-boot@0' to 'u_boot' since we are only |
| # interested in the type. |
| module_name = etype.replace('-', '_') |
| if '@' in module_name: |
| module_name = module_name.split('@')[0] |
| module = modules.get(module_name) |
| |
| # Also allow entry-type modules to be brought in from the etype directory. |
| |
| # Import the module if we have not already done so. |
| if not module: |
| old_path = sys.path |
| sys.path.insert(0, os.path.join(our_path, 'etype')) |
| try: |
| module = importlib.import_module(module_name) |
| except ImportError as e: |
| raise ValueError("Unknown entry type '%s' in node '%s' (expected etype/%s.py, error '%s'" % |
| (etype, node_path, module_name, e)) |
| finally: |
| sys.path = old_path |
| modules[module_name] = module |
| |
| # Look up the expected class name |
| return getattr(module, 'Entry_%s' % module_name) |
| |
| @staticmethod |
| def Create(section, node, etype=None): |
| """Create a new entry for a node. |
| |
| Args: |
| section: Section object containing this node |
| node: Node object containing information about the entry to |
| create |
| etype: Entry type to use, or None to work it out (used for tests) |
| |
| Returns: |
| A new Entry object of the correct type (a subclass of Entry) |
| """ |
| if not etype: |
| etype = fdt_util.GetString(node, 'type', node.name) |
| obj = Entry.Lookup(node.path, etype) |
| |
| # Call its constructor to get the object we want. |
| return obj(section, etype, node) |
| |
| def ReadNode(self): |
| """Read entry information from the node |
| |
| This must be called as the first thing after the Entry is created. |
| |
| This reads all the fields we recognise from the node, ready for use. |
| """ |
| if 'pos' in self._node.props: |
| self.Raise("Please use 'offset' instead of 'pos'") |
| self.offset = fdt_util.GetInt(self._node, 'offset') |
| self.size = fdt_util.GetInt(self._node, 'size') |
| self.orig_offset = fdt_util.GetInt(self._node, 'orig-offset') |
| self.orig_size = fdt_util.GetInt(self._node, 'orig-size') |
| if self.GetImage().copy_to_orig: |
| self.orig_offset = self.offset |
| self.orig_size = self.size |
| |
| # These should not be set in input files, but are set in an FDT map, |
| # which is also read by this code. |
| self.image_pos = fdt_util.GetInt(self._node, 'image-pos') |
| self.uncomp_size = fdt_util.GetInt(self._node, 'uncomp-size') |
| |
| self.align = fdt_util.GetInt(self._node, 'align') |
| if tools.NotPowerOfTwo(self.align): |
| raise ValueError("Node '%s': Alignment %s must be a power of two" % |
| (self._node.path, self.align)) |
| self.pad_before = fdt_util.GetInt(self._node, 'pad-before', 0) |
| self.pad_after = fdt_util.GetInt(self._node, 'pad-after', 0) |
| self.align_size = fdt_util.GetInt(self._node, 'align-size') |
| if tools.NotPowerOfTwo(self.align_size): |
| self.Raise("Alignment size %s must be a power of two" % |
| self.align_size) |
| self.align_end = fdt_util.GetInt(self._node, 'align-end') |
| self.offset_unset = fdt_util.GetBool(self._node, 'offset-unset') |
| self.expand_size = fdt_util.GetBool(self._node, 'expand-size') |
| |
| def GetDefaultFilename(self): |
| return None |
| |
| def GetFdts(self): |
| """Get the device trees used by this entry |
| |
| Returns: |
| Empty dict, if this entry is not a .dtb, otherwise: |
| Dict: |
| key: Filename from this entry (without the path) |
| value: Tuple: |
| Fdt object for this dtb, or None if not available |
| Filename of file containing this dtb |
| """ |
| return {} |
| |
| def ExpandEntries(self): |
| pass |
| |
| def AddMissingProperties(self): |
| """Add new properties to the device tree as needed for this entry""" |
| for prop in ['offset', 'size', 'image-pos']: |
| if not prop in self._node.props: |
| state.AddZeroProp(self._node, prop) |
| if self.GetImage().allow_repack: |
| if self.orig_offset is not None: |
| state.AddZeroProp(self._node, 'orig-offset', True) |
| if self.orig_size is not None: |
| state.AddZeroProp(self._node, 'orig-size', True) |
| |
| if self.compress != 'none': |
| state.AddZeroProp(self._node, 'uncomp-size') |
| err = state.CheckAddHashProp(self._node) |
| if err: |
| self.Raise(err) |
| |
| def SetCalculatedProperties(self): |
| """Set the value of device-tree properties calculated by binman""" |
| state.SetInt(self._node, 'offset', self.offset) |
| state.SetInt(self._node, 'size', self.size) |
| base = self.section.GetRootSkipAtStart() if self.section else 0 |
| state.SetInt(self._node, 'image-pos', self.image_pos - base) |
| if self.GetImage().allow_repack: |
| if self.orig_offset is not None: |
| state.SetInt(self._node, 'orig-offset', self.orig_offset, True) |
| if self.orig_size is not None: |
| state.SetInt(self._node, 'orig-size', self.orig_size, True) |
| if self.uncomp_size is not None: |
| state.SetInt(self._node, 'uncomp-size', self.uncomp_size) |
| state.CheckSetHashValue(self._node, self.GetData) |
| |
| def ProcessFdt(self, fdt): |
| """Allow entries to adjust the device tree |
| |
| Some entries need to adjust the device tree for their purposes. This |
| may involve adding or deleting properties. |
| |
| Returns: |
| True if processing is complete |
| False if processing could not be completed due to a dependency. |
| This will cause the entry to be retried after others have been |
| called |
| """ |
| return True |
| |
| def SetPrefix(self, prefix): |
| """Set the name prefix for a node |
| |
| Args: |
| prefix: Prefix to set, or '' to not use a prefix |
| """ |
| if prefix: |
| self.name = prefix + self.name |
| |
| def SetContents(self, data): |
| """Set the contents of an entry |
| |
| This sets both the data and content_size properties |
| |
| Args: |
| data: Data to set to the contents (bytes) |
| """ |
| self.data = data |
| self.contents_size = len(self.data) |
| |
| def ProcessContentsUpdate(self, data): |
| """Update the contents of an entry, after the size is fixed |
| |
| This checks that the new data is the same size as the old. If the size |
| has changed, this triggers a re-run of the packing algorithm. |
| |
| Args: |
| data: Data to set to the contents (bytes) |
| |
| Raises: |
| ValueError if the new data size is not the same as the old |
| """ |
| size_ok = True |
| new_size = len(data) |
| if state.AllowEntryExpansion() and new_size > self.contents_size: |
| # self.data will indicate the new size needed |
| size_ok = False |
| elif state.AllowEntryContraction() and new_size < self.contents_size: |
| size_ok = False |
| |
| # If not allowed to change, try to deal with it or give up |
| if size_ok: |
| if new_size > self.contents_size: |
| self.Raise('Cannot update entry size from %d to %d' % |
| (self.contents_size, new_size)) |
| |
| # Don't let the data shrink. Pad it if necessary |
| if size_ok and new_size < self.contents_size: |
| data += tools.GetBytes(0, self.contents_size - new_size) |
| |
| if not size_ok: |
| tout.Debug("Entry '%s' size change from %s to %s" % ( |
| self._node.path, ToHex(self.contents_size), |
| ToHex(new_size))) |
| self.SetContents(data) |
| return size_ok |
| |
| def ObtainContents(self): |
| """Figure out the contents of an entry. |
| |
| Returns: |
| True if the contents were found, False if another call is needed |
| after the other entries are processed. |
| """ |
| # No contents by default: subclasses can implement this |
| return True |
| |
| def ResetForPack(self): |
| """Reset offset/size fields so that packing can be done again""" |
| self.Detail('ResetForPack: offset %s->%s, size %s->%s' % |
| (ToHex(self.offset), ToHex(self.orig_offset), |
| ToHex(self.size), ToHex(self.orig_size))) |
| self.pre_reset_size = self.size |
| self.offset = self.orig_offset |
| self.size = self.orig_size |
| |
| def Pack(self, offset): |
| """Figure out how to pack the entry into the section |
| |
| Most of the time the entries are not fully specified. There may be |
| an alignment but no size. In that case we take the size from the |
| contents of the entry. |
| |
| If an entry has no hard-coded offset, it will be placed at @offset. |
| |
| Once this function is complete, both the offset and size of the |
| entry will be know. |
| |
| Args: |
| Current section offset pointer |
| |
| Returns: |
| New section offset pointer (after this entry) |
| """ |
| self.Detail('Packing: offset=%s, size=%s, content_size=%x' % |
| (ToHex(self.offset), ToHex(self.size), |
| self.contents_size)) |
| if self.offset is None: |
| if self.offset_unset: |
| self.Raise('No offset set with offset-unset: should another ' |
| 'entry provide this correct offset?') |
| self.offset = tools.Align(offset, self.align) |
| needed = self.pad_before + self.contents_size + self.pad_after |
| needed = tools.Align(needed, self.align_size) |
| size = self.size |
| if not size: |
| size = needed |
| new_offset = self.offset + size |
| aligned_offset = tools.Align(new_offset, self.align_end) |
| if aligned_offset != new_offset: |
| size = aligned_offset - self.offset |
| new_offset = aligned_offset |
| |
| if not self.size: |
| self.size = size |
| |
| if self.size < needed: |
| self.Raise("Entry contents size is %#x (%d) but entry size is " |
| "%#x (%d)" % (needed, needed, self.size, self.size)) |
| # Check that the alignment is correct. It could be wrong if the |
| # and offset or size values were provided (i.e. not calculated), but |
| # conflict with the provided alignment values |
| if self.size != tools.Align(self.size, self.align_size): |
| self.Raise("Size %#x (%d) does not match align-size %#x (%d)" % |
| (self.size, self.size, self.align_size, self.align_size)) |
| if self.offset != tools.Align(self.offset, self.align): |
| self.Raise("Offset %#x (%d) does not match align %#x (%d)" % |
| (self.offset, self.offset, self.align, self.align)) |
| self.Detail(' - packed: offset=%#x, size=%#x, content_size=%#x, next_offset=%x' % |
| (self.offset, self.size, self.contents_size, new_offset)) |
| |
| return new_offset |
| |
| def Raise(self, msg): |
| """Convenience function to raise an error referencing a node""" |
| raise ValueError("Node '%s': %s" % (self._node.path, msg)) |
| |
| def Detail(self, msg): |
| """Convenience function to log detail referencing a node""" |
| tag = "Node '%s'" % self._node.path |
| tout.Detail('%30s: %s' % (tag, msg)) |
| |
| def GetEntryArgsOrProps(self, props, required=False): |
| """Return the values of a set of properties |
| |
| Args: |
| props: List of EntryArg objects |
| |
| Raises: |
| ValueError if a property is not found |
| """ |
| values = [] |
| missing = [] |
| for prop in props: |
| python_prop = prop.name.replace('-', '_') |
| if hasattr(self, python_prop): |
| value = getattr(self, python_prop) |
| else: |
| value = None |
| if value is None: |
| value = self.GetArg(prop.name, prop.datatype) |
| if value is None and required: |
| missing.append(prop.name) |
| values.append(value) |
| if missing: |
| self.Raise('Missing required properties/entry args: %s' % |
| (', '.join(missing))) |
| return values |
| |
| def GetPath(self): |
| """Get the path of a node |
| |
| Returns: |
| Full path of the node for this entry |
| """ |
| return self._node.path |
| |
| def GetData(self): |
| self.Detail('GetData: size %s' % ToHexSize(self.data)) |
| return self.data |
| |
| def GetOffsets(self): |
| """Get the offsets for siblings |
| |
| Some entry types can contain information about the position or size of |
| other entries. An example of this is the Intel Flash Descriptor, which |
| knows where the Intel Management Engine section should go. |
| |
| If this entry knows about the position of other entries, it can specify |
| this by returning values here |
| |
| Returns: |
| Dict: |
| key: Entry type |
| value: List containing position and size of the given entry |
| type. Either can be None if not known |
| """ |
| return {} |
| |
| def SetOffsetSize(self, offset, size): |
| """Set the offset and/or size of an entry |
| |
| Args: |
| offset: New offset, or None to leave alone |
| size: New size, or None to leave alone |
| """ |
| if offset is not None: |
| self.offset = offset |
| if size is not None: |
| self.size = size |
| |
| def SetImagePos(self, image_pos): |
| """Set the position in the image |
| |
| Args: |
| image_pos: Position of this entry in the image |
| """ |
| self.image_pos = image_pos + self.offset |
| |
| def ProcessContents(self): |
| """Do any post-packing updates of entry contents |
| |
| This function should call ProcessContentsUpdate() to update the entry |
| contents, if necessary, returning its return value here. |
| |
| Args: |
| data: Data to set to the contents (bytes) |
| |
| Returns: |
| True if the new data size is OK, False if expansion is needed |
| |
| Raises: |
| ValueError if the new data size is not the same as the old and |
| state.AllowEntryExpansion() is False |
| """ |
| return True |
| |
| def WriteSymbols(self, section): |
| """Write symbol values into binary files for access at run time |
| |
| Args: |
| section: Section containing the entry |
| """ |
| pass |
| |
| def CheckOffset(self): |
| """Check that the entry offsets are correct |
| |
| This is used for entries which have extra offset requirements (other |
| than having to be fully inside their section). Sub-classes can implement |
| this function and raise if there is a problem. |
| """ |
| pass |
| |
| @staticmethod |
| def GetStr(value): |
| if value is None: |
| return '<none> ' |
| return '%08x' % value |
| |
| @staticmethod |
| def WriteMapLine(fd, indent, name, offset, size, image_pos): |
| print('%s %s%s %s %s' % (Entry.GetStr(image_pos), ' ' * indent, |
| Entry.GetStr(offset), Entry.GetStr(size), |
| name), file=fd) |
| |
| def WriteMap(self, fd, indent): |
| """Write a map of the entry to a .map file |
| |
| Args: |
| fd: File to write the map to |
| indent: Curent indent level of map (0=none, 1=one level, etc.) |
| """ |
| self.WriteMapLine(fd, indent, self.name, self.offset, self.size, |
| self.image_pos) |
| |
| def GetEntries(self): |
| """Return a list of entries contained by this entry |
| |
| Returns: |
| List of entries, or None if none. A normal entry has no entries |
| within it so will return None |
| """ |
| return None |
| |
| def GetArg(self, name, datatype=str): |
| """Get the value of an entry argument or device-tree-node property |
| |
| Some node properties can be provided as arguments to binman. First check |
| the entry arguments, and fall back to the device tree if not found |
| |
| Args: |
| name: Argument name |
| datatype: Data type (str or int) |
| |
| Returns: |
| Value of argument as a string or int, or None if no value |
| |
| Raises: |
| ValueError if the argument cannot be converted to in |
| """ |
| value = state.GetEntryArg(name) |
| if value is not None: |
| if datatype == int: |
| try: |
| value = int(value) |
| except ValueError: |
| self.Raise("Cannot convert entry arg '%s' (value '%s') to integer" % |
| (name, value)) |
| elif datatype == str: |
| pass |
| else: |
| raise ValueError("GetArg() internal error: Unknown data type '%s'" % |
| datatype) |
| else: |
| value = fdt_util.GetDatatype(self._node, name, datatype) |
| return value |
| |
| @staticmethod |
| def WriteDocs(modules, test_missing=None): |
| """Write out documentation about the various entry types to stdout |
| |
| Args: |
| modules: List of modules to include |
| test_missing: Used for testing. This is a module to report |
| as missing |
| """ |
| print('''Binman Entry Documentation |
| =========================== |
| |
| This file describes the entry types supported by binman. These entry types can |
| be placed in an image one by one to build up a final firmware image. It is |
| fairly easy to create new entry types. Just add a new file to the 'etype' |
| directory. You can use the existing entries as examples. |
| |
| Note that some entries are subclasses of others, using and extending their |
| features to produce new behaviours. |
| |
| |
| ''') |
| modules = sorted(modules) |
| |
| # Don't show the test entry |
| if '_testing' in modules: |
| modules.remove('_testing') |
| missing = [] |
| for name in modules: |
| if name.startswith('__'): |
| continue |
| module = Entry.Lookup(name, name) |
| docs = getattr(module, '__doc__') |
| if test_missing == name: |
| docs = None |
| if docs: |
| lines = docs.splitlines() |
| first_line = lines[0] |
| rest = [line[4:] for line in lines[1:]] |
| hdr = 'Entry: %s: %s' % (name.replace('_', '-'), first_line) |
| print(hdr) |
| print('-' * len(hdr)) |
| print('\n'.join(rest)) |
| print() |
| print() |
| else: |
| missing.append(name) |
| |
| if missing: |
| raise ValueError('Documentation is missing for modules: %s' % |
| ', '.join(missing)) |
| |
| def GetUniqueName(self): |
| """Get a unique name for a node |
| |
| Returns: |
| String containing a unique name for a node, consisting of the name |
| of all ancestors (starting from within the 'binman' node) separated |
| by a dot ('.'). This can be useful for generating unique filesnames |
| in the output directory. |
| """ |
| name = self.name |
| node = self._node |
| while node.parent: |
| node = node.parent |
| if node.name == 'binman': |
| break |
| name = '%s.%s' % (node.name, name) |
| return name |
| |
| def ExpandToLimit(self, limit): |
| """Expand an entry so that it ends at the given offset limit""" |
| if self.offset + self.size < limit: |
| self.size = limit - self.offset |
| # Request the contents again, since changing the size requires that |
| # the data grows. This should not fail, but check it to be sure. |
| if not self.ObtainContents(): |
| self.Raise('Cannot obtain contents when expanding entry') |
| |
| def HasSibling(self, name): |
| """Check if there is a sibling of a given name |
| |
| Returns: |
| True if there is an entry with this name in the the same section, |
| else False |
| """ |
| return name in self.section.GetEntries() |
| |
| def GetSiblingImagePos(self, name): |
| """Return the image position of the given sibling |
| |
| Returns: |
| Image position of sibling, or None if the sibling has no position, |
| or False if there is no such sibling |
| """ |
| if not self.HasSibling(name): |
| return False |
| return self.section.GetEntries()[name].image_pos |
| |
| @staticmethod |
| def AddEntryInfo(entries, indent, name, etype, size, image_pos, |
| uncomp_size, offset, entry): |
| """Add a new entry to the entries list |
| |
| Args: |
| entries: List (of EntryInfo objects) to add to |
| indent: Current indent level to add to list |
| name: Entry name (string) |
| etype: Entry type (string) |
| size: Entry size in bytes (int) |
| image_pos: Position within image in bytes (int) |
| uncomp_size: Uncompressed size if the entry uses compression, else |
| None |
| offset: Entry offset within parent in bytes (int) |
| entry: Entry object |
| """ |
| entries.append(EntryInfo(indent, name, etype, size, image_pos, |
| uncomp_size, offset, entry)) |
| |
| def ListEntries(self, entries, indent): |
| """Add files in this entry to the list of entries |
| |
| This can be overridden by subclasses which need different behaviour. |
| |
| Args: |
| entries: List (of EntryInfo objects) to add to |
| indent: Current indent level to add to list |
| """ |
| self.AddEntryInfo(entries, indent, self.name, self.etype, self.size, |
| self.image_pos, self.uncomp_size, self.offset, self) |
| |
| def ReadData(self, decomp=True): |
| """Read the data for an entry from the image |
| |
| This is used when the image has been read in and we want to extract the |
| data for a particular entry from that image. |
| |
| Args: |
| decomp: True to decompress any compressed data before returning it; |
| False to return the raw, uncompressed data |
| |
| Returns: |
| Entry data (bytes) |
| """ |
| # Use True here so that we get an uncompressed section to work from, |
| # although compressed sections are currently not supported |
| tout.Debug("ReadChildData section '%s', entry '%s'" % |
| (self.section.GetPath(), self.GetPath())) |
| data = self.section.ReadChildData(self, decomp) |
| return data |
| |
| def ReadChildData(self, child, decomp=True): |
| """Read the data for a particular child entry |
| |
| This reads data from the parent and extracts the piece that relates to |
| the given child. |
| |
| Args: |
| child: Child entry to read data for (must be valid) |
| decomp: True to decompress any compressed data before returning it; |
| False to return the raw, uncompressed data |
| |
| Returns: |
| Data for the child (bytes) |
| """ |
| pass |
| |
| def LoadData(self, decomp=True): |
| data = self.ReadData(decomp) |
| self.contents_size = len(data) |
| self.ProcessContentsUpdate(data) |
| self.Detail('Loaded data size %x' % len(data)) |
| |
| def GetImage(self): |
| """Get the image containing this entry |
| |
| Returns: |
| Image object containing this entry |
| """ |
| return self.section.GetImage() |
| |
| def WriteData(self, data, decomp=True): |
| """Write the data to an entry in the image |
| |
| This is used when the image has been read in and we want to replace the |
| data for a particular entry in that image. |
| |
| The image must be re-packed and written out afterwards. |
| |
| Args: |
| data: Data to replace it with |
| decomp: True to compress the data if needed, False if data is |
| already compressed so should be used as is |
| |
| Returns: |
| True if the data did not result in a resize of this entry, False if |
| the entry must be resized |
| """ |
| if self.size is not None: |
| self.contents_size = self.size |
| else: |
| self.contents_size = self.pre_reset_size |
| ok = self.ProcessContentsUpdate(data) |
| self.Detail('WriteData: size=%x, ok=%s' % (len(data), ok)) |
| section_ok = self.section.WriteChildData(self) |
| return ok and section_ok |
| |
| def WriteChildData(self, child): |
| """Handle writing the data in a child entry |
| |
| This should be called on the child's parent section after the child's |
| data has been updated. It |
| |
| This base-class implementation does nothing, since the base Entry object |
| does not have any children. |
| |
| Args: |
| child: Child Entry that was written |
| |
| Returns: |
| True if the section could be updated successfully, False if the |
| data is such that the section could not updat |
| """ |
| return True |
| |
| def GetSiblingOrder(self): |
| """Get the relative order of an entry amoung its siblings |
| |
| Returns: |
| 'start' if this entry is first among siblings, 'end' if last, |
| otherwise None |
| """ |
| entries = list(self.section.GetEntries().values()) |
| if entries: |
| if self == entries[0]: |
| return 'start' |
| elif self == entries[-1]: |
| return 'end' |
| return 'middle' |