util.py

# This Source Code Form is subject to the terms of the Mozilla Public
# License, v. 2.0. If a copy of the MPL was not distributed with this file,
# You can obtain one at http://mozilla.org/MPL/2.0/.

# This file contains miscellaneous utility functions that don't belong anywhere
# in particular.

from __future__ import absolute_import, print_function, unicode_literals

import argparse
import collections
import ctypes
import difflib
import errno
import functools
import hashlib
import os
import pprint
import re
import stat
import subprocess
import sys
import time
from collections import (
    OrderedDict,
)
from io import (BytesIO, StringIO)

import six

if sys.platform == 'win32':
    _kernel32 = ctypes.windll.kernel32
    _FILE_ATTRIBUTE_NOT_CONTENT_INDEXED = 0x2000
    system_encoding = 'mbcs'
else:
    system_encoding = 'utf-8'


def exec_(object, globals=None, locals=None):
    """Wrapper around the exec statement to avoid bogus errors like:

    SyntaxError: unqualified exec is not allowed in function ...
    it is a nested function.

    or

    SyntaxError: unqualified exec is not allowed in function ...
    it contains a nested function with free variable

    which happen with older versions of python 2.7.
    """
    exec(object, globals, locals)


def hash_file(path, hasher=None):
    """Hashes a file specified by the path given and returns the hex digest."""

    # If the default hashing function changes, this may invalidate
    # lots of cached data.  Don't change it lightly.
    h = hasher or hashlib.sha1()

    with open(path, 'rb') as fh:
        while True:
            data = fh.read(8192)

            if not len(data):
                break

            h.update(data)

    return h.hexdigest()


class EmptyValue(six.text_type):
    """A dummy type that behaves like an empty string and sequence.

    This type exists in order to support
    :py:class:`mozbuild.frontend.reader.EmptyConfig`. It should likely not be
    used elsewhere.
    """

    def __init__(self):
        super(EmptyValue, self).__init__()


class ReadOnlyNamespace(object):
    """A class for objects with immutable attributes set at initialization."""

    def __init__(self, **kwargs):
        for k, v in six.iteritems(kwargs):
            super(ReadOnlyNamespace, self).__setattr__(k, v)

    def __delattr__(self, key):
        raise Exception('Object does not support deletion.')

    def __setattr__(self, key, value):
        raise Exception('Object does not support assignment.')

    def __ne__(self, other):
        return not (self == other)

    def __eq__(self, other):
        return self is other or (
            hasattr(other, '__dict__') and self.__dict__ == other.__dict__)

    def __repr__(self):
        return '<%s %r>' % (self.__class__.__name__, self.__dict__)


class ReadOnlyDict(dict):
    """A read-only dictionary."""

    def __init__(self, *args, **kwargs):
        dict.__init__(self, *args, **kwargs)

    def __delitem__(self, key):
        raise Exception('Object does not support deletion.')

    def __setitem__(self, key, value):
        raise Exception('Object does not support assignment.')

    def update(self, *args, **kwargs):
        raise Exception('Object does not support update.')


class undefined_default(object):
    """Represents an undefined argument value that isn't None."""


undefined = undefined_default()


class ReadOnlyDefaultDict(ReadOnlyDict):
    """A read-only dictionary that supports default values on retrieval."""

    def __init__(self, default_factory, *args, **kwargs):
        ReadOnlyDict.__init__(self, *args, **kwargs)
        self._default_factory = default_factory

    def __missing__(self, key):
        value = self._default_factory()
        dict.__setitem__(self, key, value)
        return value


def ensureParentDir(path):
    """Ensures the directory parent to the given file exists."""
    d = os.path.dirname(path)
    if d and not os.path.exists(path):
        try:
            os.makedirs(d)
        except OSError as error:
            if error.errno != errno.EEXIST:
                raise


def mkdir(path, not_indexed=False):
    """Ensure a directory exists.

    If ``not_indexed`` is True, an attribute is set that disables content
    indexing on the directory.
    """
    try:
        os.makedirs(path)
    except OSError as e:
        if e.errno != errno.EEXIST:
            raise

    if not_indexed:
        if sys.platform == 'win32':
            if isinstance(path, six.string_types):
                fn = _kernel32.SetFileAttributesW
            else:
                fn = _kernel32.SetFileAttributesA

            fn(path, _FILE_ATTRIBUTE_NOT_CONTENT_INDEXED)
        elif sys.platform == 'darwin':
            with open(os.path.join(path, '.metadata_never_index'), 'a'):
                pass


def simple_diff(filename, old_lines, new_lines):
    """Returns the diff between old_lines and new_lines, in unified diff form,
    as a list of lines.

    old_lines and new_lines are lists of non-newline terminated lines to
    compare.
    old_lines can be None, indicating a file creation.
    new_lines can be None, indicating a file deletion.
    """

    old_name = '/dev/null' if old_lines is None else filename
    new_name = '/dev/null' if new_lines is None else filename

    return difflib.unified_diff(old_lines or [], new_lines or [],
                                old_name, new_name, n=4, lineterm='')


class FileAvoidWrite(BytesIO):
    """File-like object that buffers output and only writes if content changed.

    We create an instance from an existing filename. New content is written to
    it. When we close the file object, if the content in the in-memory buffer
    differs from what is on disk, then we write out the new content. Otherwise,
    the original file is untouched.

    Instances can optionally capture diffs of file changes. This feature is not
    enabled by default because it a) doesn't make sense for binary files b)
    could add unwanted overhead to calls.

    Additionally, there is dry run mode where the file is not actually written
    out, but reports whether the file was existing and would have been updated
    still occur, as well as diff capture if requested.
    """

    def __init__(self, filename, capture_diff=False, dry_run=False, readmode='rU'):
        BytesIO.__init__(self)
        self.name = filename
        assert type(capture_diff) == bool
        assert type(dry_run) == bool
        assert 'r' in readmode
        self._capture_diff = capture_diff
        self._write_to_file = not dry_run
        self.diff = None
        self.mode = readmode
        self._binary_mode = 'b' in readmode

    def write(self, buf):
        if isinstance(buf, six.text_type):
            buf = buf.encode('utf-8')
        BytesIO.write(self, buf)

    def avoid_writing_to_file(self):
        self._write_to_file = False

    def close(self):
        """Stop accepting writes, compare file contents, and rewrite if needed.

        Returns a tuple of bools indicating what action was performed:

            (file existed, file updated)

        If ``capture_diff`` was specified at construction time and the
        underlying file was changed, ``.diff`` will be populated with the diff
        of the result.
        """
        if self._binary_mode or six.PY2:
            # Use binary data under Python 2 because it can be written to files
            # opened with either open(mode='w') or open(mode='wb') without raising
            # unicode errors. Also use binary data if the caller explicitly asked for
            # it.
            buf = self.getvalue()
        else:
            # Use strings in Python 3 unless the caller explicitly asked for binary
            # data.
            buf = self.getvalue().decode('utf-8')

        BytesIO.close(self)
        existed = False
        old_content = None

        try:
            existing = open(self.name, self.mode)
            existed = True
        except IOError:
            pass
        else:
            try:
                old_content = existing.read()
                if old_content == buf:
                    return True, False
            except IOError:
                pass
            finally:
                existing.close()

        if self._write_to_file:
            ensureParentDir(self.name)
            # Maintain 'b' if specified.  'U' only applies to modes starting with
            # 'r', so it is dropped.
            writemode = 'w'
            if self._binary_mode:
                writemode += 'b'
            with open(self.name, writemode) as file:
                file.write(buf)

        self._generate_diff(buf, old_content)

        return existed, True

    def _generate_diff(self, new_content, old_content):
        """Generate a diff for the changed contents if `capture_diff` is True.

        If the changed contents could not be decoded as utf-8 then generate a
        placeholder message instead of a diff.

        Args:
            new_content: Str or bytes holding the new file contents.
            old_content: Str or bytes holding the original file contents. Should be
                None if no old content is being overwritten.
        """
        if not self._capture_diff:
            return

        try:
            if old_content is None:
                old_lines = None
            else:
                if self._binary_mode:
                    # difflib doesn't work with bytes.
                    old_content = old_content.decode('utf-8')

                old_lines = old_content.splitlines()

            if self._binary_mode:
                # difflib doesn't work with bytes.
                new_content = new_content.decode('utf-8')

            new_lines = new_content.splitlines()

            self.diff = simple_diff(self.name, old_lines, new_lines)
        # FileAvoidWrite isn't unicode/bytes safe. So, files with non-ascii
        # content or opened and written in different modes may involve
        # implicit conversion and this will make Python unhappy. Since
        # diffing isn't a critical feature, we just ignore the failure.
        # This can go away once FileAvoidWrite uses io.BytesIO and
        # io.StringIO. But that will require a lot of work.
        except (UnicodeDecodeError, UnicodeEncodeError):
            self.diff = ['Binary or non-ascii file changed: %s' %
                         self.name]

    def __enter__(self):
        return self

    def __exit__(self, type, value, traceback):
        if not self.closed:
            self.close()


def resolve_target_to_make(topobjdir, target):
    r'''
    Resolve `target` (a target, directory, or file) to a make target.

    `topobjdir` is the object directory; all make targets will be
    rooted at or below the top-level Makefile in this directory.

    Returns a pair `(reldir, target)` where `reldir` is a directory
    relative to `topobjdir` containing a Makefile and `target` is a
    make target (possibly `None`).

    A directory resolves to the nearest directory at or above
    containing a Makefile, and target `None`.

    A regular (non-Makefile) file resolves to the nearest directory at
    or above the file containing a Makefile, and an appropriate
    target.

    A Makefile resolves to the nearest parent strictly above the
    Makefile containing a different Makefile, and an appropriate
    target.
    '''

    target = target.replace(os.sep, '/').lstrip('/')
    abs_target = os.path.join(topobjdir, target)

    # For directories, run |make -C dir|. If the directory does not
    # contain a Makefile, check parents until we find one. At worst,
    # this will terminate at the root.
    if os.path.isdir(abs_target):
        current = abs_target

        while True:
            make_path = os.path.join(current, 'Makefile')
            if os.path.exists(make_path):
                return (current[len(topobjdir) + 1:], None)

            current = os.path.dirname(current)

    # If it's not in a directory, this is probably a top-level make
    # target. Treat it as such.
    if '/' not in target:
        return (None, target)

    # We have a relative path within the tree. We look for a Makefile
    # as far into the path as possible. Then, we compute the make
    # target as relative to that directory.
    reldir = os.path.dirname(target)
    target = os.path.basename(target)

    while True:
        make_path = os.path.join(topobjdir, reldir, 'Makefile')

        # We append to target every iteration, so the check below
        # happens exactly once.
        if target != 'Makefile' and os.path.exists(make_path):
            return (reldir, target)

        target = os.path.join(os.path.basename(reldir), target)
        reldir = os.path.dirname(reldir)


class List(list):
    """A list specialized for moz.build environments.

    We overload the assignment and append operations to require that the
    appended thing is a list. This avoids bad surprises coming from appending
    a string to a list, which would just add each letter of the string.
    """

    def __init__(self, iterable=None, **kwargs):
        if iterable is None:
            iterable = []
        if not isinstance(iterable, list):
            raise ValueError('List can only be created from other list instances.')

        self._kwargs = kwargs
        return super(List, self).__init__(iterable)

    def extend(self, l):
        if not isinstance(l, list):
            raise ValueError('List can only be extended with other list instances.')

        return super(List, self).extend(l)

    def __setitem__(self, key, val):
        if isinstance(key, slice):
            if not isinstance(val, list):
                raise ValueError('List can only be sliced with other list '
                                 'instances.')
            if key.step:
                raise ValueError('List cannot be sliced with a nonzero step '
                                 'value')
            # Python 2 and Python 3 do this differently for some reason.
            if six.PY2:
                return super(List, self).__setslice__(key.start, key.stop,
                                                      val)
            else:
                return super(List, self).__setitem__(key, val)
        return super(List, self).__setitem__(key, val)

    def __setslice__(self, i, j, sequence):
        return self.__setitem__(slice(i, j), sequence)

    def __add__(self, other):
        # Allow None and EmptyValue is a special case because it makes undefined
        # variable references in moz.build behave better.
        other = [] if isinstance(other, (type(None), EmptyValue)) else other
        if not isinstance(other, list):
            raise ValueError('Only lists can be appended to lists.')

        new_list = self.__class__(self, **self._kwargs)
        new_list.extend(other)
        return new_list

    def __iadd__(self, other):
        other = [] if isinstance(other, (type(None), EmptyValue)) else other
        if not isinstance(other, list):
            raise ValueError('Only lists can be appended to lists.')

        return super(List, self).__iadd__(other)


class UnsortedError(Exception):
    def __init__(self, srtd, original):
        assert len(srtd) == len(original)

        self.sorted = srtd
        self.original = original

        for i, orig in enumerate(original):
            s = srtd[i]

            if orig != s:
                self.i = i
                break

    def __str__(self):
        s = StringIO()

        s.write('An attempt was made to add an unsorted sequence to a list. ')
        s.write('The incoming list is unsorted starting at element %d. ' %
                self.i)
        s.write('We expected "%s" but got "%s"' % (
            self.sorted[self.i], self.original[self.i]))

        return s.getvalue()


class StrictOrderingOnAppendList(List):
    """A list specialized for moz.build environments.

    We overload the assignment and append operations to require that incoming
    elements be ordered. This enforces cleaner style in moz.build files.
    """

    @staticmethod
    def ensure_sorted(l):
        if isinstance(l, StrictOrderingOnAppendList):
            return

        def _first_element(e):
            # If the list entry is a tuple, we sort based on the first element
            # in the tuple.
            return e[0] if isinstance(e, tuple) else e
        srtd = sorted(l, key=lambda x: _first_element(x).lower())

        if srtd != l:
            raise UnsortedError(srtd, l)

    def __init__(self, iterable=None, **kwargs):
        if iterable is None:
            iterable = []

        StrictOrderingOnAppendList.ensure_sorted(iterable)

        super(StrictOrderingOnAppendList, self).__init__(iterable, **kwargs)

    def extend(self, l):
        StrictOrderingOnAppendList.ensure_sorted(l)

        return super(StrictOrderingOnAppendList, self).extend(l)

    def __setitem__(self, key, val):
        if isinstance(key, slice):
            StrictOrderingOnAppendList.ensure_sorted(val)
        return super(StrictOrderingOnAppendList, self).__setitem__(key, val)

    def __add__(self, other):
        StrictOrderingOnAppendList.ensure_sorted(other)

        return super(StrictOrderingOnAppendList, self).__add__(other)

    def __iadd__(self, other):
        StrictOrderingOnAppendList.ensure_sorted(other)

        return super(StrictOrderingOnAppendList, self).__iadd__(other)


class ImmutableStrictOrderingOnAppendList(StrictOrderingOnAppendList):
    """Like StrictOrderingOnAppendList, but not allowing mutations of the value.
    """

    def append(self, elt):
        raise Exception("cannot use append on this type")

    def extend(self, iterable):
        raise Exception("cannot use extend on this type")

    def __setslice__(self, i, j, iterable):
        raise Exception("cannot assign to slices on this type")

    def __setitem__(self, i, elt):
        raise Exception("cannot assign to indexes on this type")

    def __iadd__(self, other):
        raise Exception("cannot use += on this type")


class StrictOrderingOnAppendListWithAction(StrictOrderingOnAppendList):
    """An ordered list that accepts a callable to be applied to each item.

    A callable (action) passed to the constructor is run on each item of input.
    The result of running the callable on each item will be stored in place of
    the original input, but the original item must be used to enforce sortedness.
    """

    def __init__(self, iterable=(), action=None):
        if not callable(action):
            raise ValueError('A callable action is required to construct '
                             'a StrictOrderingOnAppendListWithAction')

        self._action = action
        if not isinstance(iterable, (tuple, list)):
            raise ValueError(
                'StrictOrderingOnAppendListWithAction can only be initialized '
                'with another list')
        iterable = [self._action(i) for i in iterable]
        super(StrictOrderingOnAppendListWithAction, self).__init__(
            iterable, action=action)

    def extend(self, l):
        if not isinstance(l, list):
            raise ValueError(
                'StrictOrderingOnAppendListWithAction can only be extended '
                'with another list')
        l = [self._action(i) for i in l]
        return super(StrictOrderingOnAppendListWithAction, self).extend(l)

    def __setitem__(self, key, val):
        if isinstance(key, slice):
            if not isinstance(val, list):
                raise ValueError(
                    'StrictOrderingOnAppendListWithAction can only be sliced '
                    'with another list')
            val = [self._action(item) for item in val]
        return super(StrictOrderingOnAppendListWithAction, self).__setitem__(
            key, val)

    def __add__(self, other):
        if not isinstance(other, list):
            raise ValueError(
                'StrictOrderingOnAppendListWithAction can only be added with '
                'another list')
        return super(StrictOrderingOnAppendListWithAction, self).__add__(other)

    def __iadd__(self, other):
        if not isinstance(other, list):
            raise ValueError(
                'StrictOrderingOnAppendListWithAction can only be added with '
                'another list')
        other = [self._action(i) for i in other]
        return super(StrictOrderingOnAppendListWithAction, self).__iadd__(other)


class MozbuildDeletionError(Exception):
    pass


def FlagsFactory(flags):
    """Returns a class which holds optional flags for an item in a list.

    The flags are defined in the dict given as argument, where keys are
    the flag names, and values the type used for the value of that flag.

    The resulting class is used by the various <TypeName>WithFlagsFactory
    functions below.
    """
    assert isinstance(flags, dict)
    assert all(isinstance(v, type) for v in flags.values())

    class Flags(object):
        __slots__ = flags.keys()
        _flags = flags

        def update(self, **kwargs):
            for k, v in six.iteritems(kwargs):
                setattr(self, k, v)

        def __getattr__(self, name):
            if name not in self.__slots__:
                raise AttributeError("'%s' object has no attribute '%s'" %
                                     (self.__class__.__name__, name))
            try:
                return object.__getattr__(self, name)
            except AttributeError:
                value = self._flags[name]()
                self.__setattr__(name, value)
                return value

        def __setattr__(self, name, value):
            if name not in self.__slots__:
                raise AttributeError("'%s' object has no attribute '%s'" %
                                     (self.__class__.__name__, name))
            if not isinstance(value, self._flags[name]):
                raise TypeError("'%s' attribute of class '%s' must be '%s'" %
                                (name, self.__class__.__name__,
                                 self._flags[name].__name__))
            return object.__setattr__(self, name, value)

        def __delattr__(self, name):
            raise MozbuildDeletionError('Unable to delete attributes for this object')

    return Flags


class StrictOrderingOnAppendListWithFlags(StrictOrderingOnAppendList):
    """A list with flags specialized for moz.build environments.

    Each subclass has a set of typed flags; this class lets us use `isinstance`
    for natural testing.
    """


def StrictOrderingOnAppendListWithFlagsFactory(flags):
    """Returns a StrictOrderingOnAppendList-like object, with optional
    flags on each item.

    The flags are defined in the dict given as argument, where keys are
    the flag names, and values the type used for the value of that flag.

    Example:
        FooList = StrictOrderingOnAppendListWithFlagsFactory({
            'foo': bool, 'bar': unicode
        })
        foo = FooList(['a', 'b', 'c'])
        foo['a'].foo = True
        foo['b'].bar = 'bar'
    """
    class StrictOrderingOnAppendListWithFlagsSpecialization(StrictOrderingOnAppendListWithFlags):
        def __init__(self, iterable=None):
            if iterable is None:
                iterable = []
            StrictOrderingOnAppendListWithFlags.__init__(self, iterable)
            self._flags_type = FlagsFactory(flags)
            self._flags = dict()

        def __getitem__(self, name):
            if name not in self._flags:
                if name not in self:
                    raise KeyError("'%s'" % name)
                self._flags[name] = self._flags_type()
            return self._flags[name]

        def __setitem__(self, name, value):
            if not isinstance(name, slice):
                raise TypeError("'%s' object does not support item assignment" %
                                self.__class__.__name__)
            result = super(StrictOrderingOnAppendListWithFlagsSpecialization,
                           self).__setitem__(name, value)
            # We may have removed items.
            for k in set(self._flags.keys()) - set(self):
                del self._flags[k]
            if isinstance(value, StrictOrderingOnAppendListWithFlags):
                self._update_flags(value)
            return result

        def _update_flags(self, other):
            if self._flags_type._flags != other._flags_type._flags:
                raise ValueError('Expected a list of strings with flags like %s, not like %s' %
                                 (self._flags_type._flags, other._flags_type._flags))
            intersection = set(self._flags.keys()) & set(other._flags.keys())
            if intersection:
                raise ValueError(
                    'Cannot update flags: both lists of strings with flags configure %s' %
                    intersection
                    )
            self._flags.update(other._flags)

        def extend(self, l):
            result = super(StrictOrderingOnAppendListWithFlagsSpecialization,
                           self).extend(l)
            if isinstance(l, StrictOrderingOnAppendListWithFlags):
                self._update_flags(l)
            return result

        def __add__(self, other):
            result = super(StrictOrderingOnAppendListWithFlagsSpecialization,
                           self).__add__(other)
            if isinstance(other, StrictOrderingOnAppendListWithFlags):
                # Result has flags from other but not from self, since
                # internally we duplicate self and then extend with other, and
                # only extend knows about flags.  Since we don't allow updating
                # when the set of flag keys intersect, which we instance we pass
                # to _update_flags here matters.  This needs to be correct but
                # is an implementation detail.
                result._update_flags(self)
            return result

        def __iadd__(self, other):
            result = super(StrictOrderingOnAppendListWithFlagsSpecialization,
                           self).__iadd__(other)
            if isinstance(other, StrictOrderingOnAppendListWithFlags):
                self._update_flags(other)
            return result

    return StrictOrderingOnAppendListWithFlagsSpecialization


class HierarchicalStringList(object):
    """A hierarchy of lists of strings.

    Each instance of this object contains a list of strings, which can be set or
    appended to. A sub-level of the hierarchy is also an instance of this class,
    can be added by appending to an attribute instead.

    For example, the moz.build variable EXPORTS is an instance of this class. We
    can do:

    EXPORTS += ['foo.h']
    EXPORTS.mozilla.dom += ['bar.h']

    In this case, we have 3 instances (EXPORTS, EXPORTS.mozilla, and
    EXPORTS.mozilla.dom), and the first and last each have one element in their
    list.
    """
    __slots__ = ('_strings', '_children')

    def __init__(self):
        # Please change ContextDerivedTypedHierarchicalStringList in context.py
        # if you make changes here.
        self._strings = StrictOrderingOnAppendList()
        self._children = {}

    class StringListAdaptor(collections.Sequence):
        def __init__(self, hsl):
            self._hsl = hsl

        def __getitem__(self, index):
            return self._hsl._strings[index]

        def __len__(self):
            return len(self._hsl._strings)

    def walk(self):
        """Walk over all HierarchicalStringLists in the hierarchy.

        This is a generator of (path, sequence).

        The path is '' for the root level and '/'-delimited strings for
        any descendants.  The sequence is a read-only sequence of the
        strings contained at that level.
        """

        if self._strings:
            path_to_here = ''
            yield path_to_here, self.StringListAdaptor(self)

        for k, l in sorted(self._children.items()):
            for p, v in l.walk():
                path_to_there = '%s/%s' % (k, p)
                yield path_to_there.strip('/'), v

    def __setattr__(self, name, value):
        if name in self.__slots__:
            return object.__setattr__(self, name, value)

        # __setattr__ can be called with a list when a simple assignment is
        # used:
        #
        # EXPORTS.foo = ['file.h']
        #
        # In this case, we need to overwrite foo's current list of strings.
        #
        # However, __setattr__ is also called with a HierarchicalStringList
        # to try to actually set the attribute. We want to ignore this case,
        # since we don't actually create an attribute called 'foo', but just add
        # it to our list of children (using _get_exportvariable()).
        self._set_exportvariable(name, value)

    def __getattr__(self, name):
        if name.startswith('__'):
            return object.__getattr__(self, name)
        return self._get_exportvariable(name)

    def __delattr__(self, name):
        raise MozbuildDeletionError('Unable to delete attributes for this object')

    def __iadd__(self, other):
        if isinstance(other, HierarchicalStringList):
            self._strings += other._strings
            for c in other._children:
                self[c] += other[c]
        else:
            self._check_list(other)
            self._strings += other
        return self

    def __getitem__(self, name):
        return self._get_exportvariable(name)

    def __setitem__(self, name, value):
        self._set_exportvariable(name, value)

    def _get_exportvariable(self, name):
        # Please change ContextDerivedTypedHierarchicalStringList in context.py
        # if you make changes here.
        child = self._children.get(name)
        if not child:
            child = self._children[name] = HierarchicalStringList()
        return child

    def _set_exportvariable(self, name, value):
        if name in self._children:
            if value is self._get_exportvariable(name):
                return
            raise KeyError('global_ns', 'reassign',
                           '<some variable>.%s' % name)

        exports = self._get_exportvariable(name)
        exports._check_list(value)
        exports._strings += value

    def _check_list(self, value):
        if not isinstance(value, list):
            raise ValueError('Expected a list of strings, not %s' % type(value))
        for v in value:
            if not isinstance(v, six.string_types):
                raise ValueError(
                    'Expected a list of strings, not an element of %s' % type(v))


class LockFile(object):
    """LockFile is used by the lock_file method to hold the lock.

    This object should not be used directly, but only through
    the lock_file method below.
    """

    def __init__(self, lockfile):
        self.lockfile = lockfile

    def __del__(self):
        while True:
            try:
                os.remove(self.lockfile)
                break
            except OSError as e:
                if e.errno == errno.EACCES:
                    # Another process probably has the file open, we'll retry.
                    # Just a short sleep since we want to drop the lock ASAP
                    # (but we need to let some other process close the file
                    # first).
                    time.sleep(0.1)
                else:
                    # Re-raise unknown errors
                    raise


def lock_file(lockfile, max_wait=600):
    """Create and hold a lockfile of the given name, with the given timeout.

    To release the lock, delete the returned object.
    """

    # FUTURE This function and object could be written as a context manager.

    while True:
        try:
            fd = os.open(lockfile, os.O_EXCL | os.O_RDWR | os.O_CREAT)
            # We created the lockfile, so we're the owner
            break
        except OSError as e:
            if (e.errno == errno.EEXIST or
                (sys.platform == "win32" and e.errno == errno.EACCES)):
                pass
            else:
                # Should not occur
                raise

        try:
            # The lock file exists, try to stat it to get its age
            # and read its contents to report the owner PID
            f = open(lockfile, 'r')
            s = os.stat(lockfile)
        except EnvironmentError as e:
            if e.errno == errno.ENOENT or e.errno == errno.EACCES:
                # We didn't create the lockfile, so it did exist, but it's
                # gone now. Just try again
                continue

            raise Exception('{0} exists but stat() failed: {1}'.format(
                lockfile, e.strerror))

        # We didn't create the lockfile and it's still there, check
        # its age
        now = int(time.time())
        if now - s[stat.ST_MTIME] > max_wait:
            pid = f.readline().rstrip()
            raise Exception('{0} has been locked for more than '
                            '{1} seconds (PID {2})'.format(lockfile, max_wait, pid))

        # It's not been locked too long, wait a while and retry
        f.close()
        time.sleep(1)

    # if we get here. we have the lockfile. Convert the os.open file
    # descriptor into a Python file object and record our PID in it
    f = os.fdopen(fd, 'w')
    f.write('{0}\n'.format(os.getpid()))
    f.close()

    return LockFile(lockfile)


class OrderedDefaultDict(OrderedDict):
    '''A combination of OrderedDict and defaultdict.'''

    def __init__(self, default_factory, *args, **kwargs):
        OrderedDict.__init__(self, *args, **kwargs)
        self._default_factory = default_factory

    def __missing__(self, key):
        value = self[key] = self._default_factory()
        return value


class KeyedDefaultDict(dict):
    '''Like a defaultdict, but the default_factory function takes the key as
    argument'''

    def __init__(self, default_factory, *args, **kwargs):
        dict.__init__(self, *args, **kwargs)
        self._default_factory = default_factory

    def __missing__(self, key):
        value = self._default_factory(key)
        dict.__setitem__(self, key, value)
        return value


class ReadOnlyKeyedDefaultDict(KeyedDefaultDict, ReadOnlyDict):
    '''Like KeyedDefaultDict, but read-only.'''


class memoize(dict):
    '''A decorator to memoize the results of function calls depending
    on its arguments.
    Both functions and instance methods are handled, although in the
    instance method case, the results are cache in the instance itself.
    '''

    def __init__(self, func):
        self.func = func
        functools.update_wrapper(self, func)

    def __call__(self, *args):
        if args not in self:
            self[args] = self.func(*args)
        return self[args]

    def method_call(self, instance, *args):
        name = '_%s' % self.func.__name__
        if not hasattr(instance, name):
            setattr(instance, name, {})
        cache = getattr(instance, name)
        if args not in cache:
            cache[args] = self.func(instance, *args)
        return cache[args]

    def __get__(self, instance, cls):
        return functools.update_wrapper(
            functools.partial(self.method_call, instance), self.func)


class memoized_property(object):
    '''A specialized version of the memoize decorator that works for
    class instance properties.
    '''

    def __init__(self, func):
        self.func = func

    def __get__(self, instance, cls):
        name = '_%s' % self.func.__name__
        if not hasattr(instance, name):
            setattr(instance, name, self.func(instance))
        return getattr(instance, name)


def TypedNamedTuple(name, fields):
    """Factory for named tuple types with strong typing.

    Arguments are an iterable of 2-tuples. The first member is the
    the field name. The second member is a type the field will be validated
    to be.

    Construction of instances varies from ``collections.namedtuple``.

    First, if a single tuple argument is given to the constructor, this is
    treated as the equivalent of passing each tuple value as a separate
    argument into __init__. e.g.::

        t = (1, 2)
        TypedTuple(t) == TypedTuple(1, 2)

    This behavior is meant for moz.build files, so vanilla tuples are
    automatically cast to typed tuple instances.

    Second, fields in the tuple are validated to be instances of the specified
    type. This is done via an ``isinstance()`` check. To allow multiple types,
    pass a tuple as the allowed types field.
    """
    cls = collections.namedtuple(name, (name for name, typ in fields))

    class TypedTuple(cls):
        __slots__ = ()

        def __new__(klass, *args, **kwargs):
            if len(args) == 1 and not kwargs and isinstance(args[0], tuple):
                args = args[0]

            return super(TypedTuple, klass).__new__(klass, *args, **kwargs)

        def __init__(self, *args, **kwargs):
            for i, (fname, ftype) in enumerate(self._fields):
                value = self[i]

                if not isinstance(value, ftype):
                    raise TypeError('field in tuple not of proper type: %s; '
                                    'got %s, expected %s' % (fname,
                                                             type(value), ftype))

    TypedTuple._fields = fields

    return TypedTuple


@memoize
def TypedList(type, base_class=List):
    '''A list with type coercion.

    The given ``type`` is what list elements are being coerced to. It may do
    strict validation, throwing ValueError exceptions.

    A ``base_class`` type can be given for more specific uses than a List. For
    example, a Typed StrictOrderingOnAppendList can be created with:

       TypedList(unicode, StrictOrderingOnAppendList)
    '''
    class _TypedList(base_class):
        @staticmethod
        def normalize(e):
            if not isinstance(e, type):
                e = type(e)
            return e

        def _ensure_type(self, l):
            if isinstance(l, self.__class__):
                return l

            return [self.normalize(e) for e in l]

        def __init__(self, iterable=None, **kwargs):
            if iterable is None:
                iterable = []
            iterable = self._ensure_type(iterable)

            super(_TypedList, self).__init__(iterable, **kwargs)

        def extend(self, l):
            l = self._ensure_type(l)

            return super(_TypedList, self).extend(l)

        def __setitem__(self, key, val):
            val = self._ensure_type(val)

            return super(_TypedList, self).__setitem__(key, val)

        def __add__(self, other):
            other = self._ensure_type(other)

            return super(_TypedList, self).__add__(other)

        def __iadd__(self, other):
            other = self._ensure_type(other)

            return super(_TypedList, self).__iadd__(other)

        def append(self, other):
            self += [other]

    return _TypedList


def group_unified_files(files, unified_prefix, unified_suffix,
                        files_per_unified_file):
    """Return an iterator of (unified_filename, source_filenames) tuples.

    We compile most C and C++ files in "unified mode"; instead of compiling
    ``a.cpp``, ``b.cpp``, and ``c.cpp`` separately, we compile a single file
    that looks approximately like::

       #include "a.cpp"
       #include "b.cpp"
       #include "c.cpp"

    This function handles the details of generating names for the unified
    files, and determining which original source files go in which unified
    file."""

    # Make sure the input list is sorted. If it's not, bad things could happen!
    files = sorted(files)

    # Our last returned list of source filenames may be short, and we
    # don't want the fill value inserted by zip_longest to be an
    # issue.  So we do a little dance to filter it out ourselves.
    dummy_fill_value = ("dummy",)

    def filter_out_dummy(iterable):
        return six.moves.filter(lambda x: x != dummy_fill_value,
                                iterable)

    # From the itertools documentation, slightly modified:
    def grouper(n, iterable):
        "grouper(3, 'ABCDEFG', 'x') --> ABC DEF Gxx"
        args = [iter(iterable)] * n
        return six.moves.zip_longest(fillvalue=dummy_fill_value, *args)

    for i, unified_group in enumerate(grouper(files_per_unified_file,
                                              files)):
        just_the_filenames = list(filter_out_dummy(unified_group))
        yield '%s%d.%s' % (unified_prefix, i, unified_suffix), just_the_filenames


def pair(iterable):
    '''Given an iterable, returns an iterable pairing its items.

    For example,
        list(pair([1,2,3,4,5,6]))
    returns
        [(1,2), (3,4), (5,6)]
    '''
    i = iter(iterable)
    return six.moves.zip_longest(i, i)


VARIABLES_RE = re.compile('\$\((\w+)\)')


def expand_variables(s, variables):
    '''Given a string with $(var) variable references, replace those references
    with the corresponding entries from the given `variables` dict.

    If a variable value is not a string, it is iterated and its items are
    joined with a whitespace.'''
    result = ''
    for s, name in pair(VARIABLES_RE.split(s)):
        result += s
        value = variables.get(name)
        if not value:
            continue
        if not isinstance(value, six.string_types):
            value = ' '.join(value)
        result += value
    return result


class DefinesAction(argparse.Action):
    '''An ArgumentParser action to handle -Dvar[=value] type of arguments.'''

    def __call__(self, parser, namespace, values, option_string):
        defines = getattr(namespace, self.dest)
        if defines is None:
            defines = {}
        values = values.split('=', 1)
        if len(values) == 1:
            name, value = values[0], 1
        else:
            name, value = values
            if value.isdigit():
                value = int(value)
        defines[name] = value
        setattr(namespace, self.dest, defines)


class EnumStringComparisonError(Exception):
    pass


class EnumString(six.text_type):
    '''A string type that only can have a limited set of values, similarly to
    an Enum, and can only be compared against that set of values.

    The class is meant to be subclassed, where the subclass defines
    POSSIBLE_VALUES. The `subclass` method is a helper to create such
    subclasses.
    '''
    POSSIBLE_VALUES = ()

    def __init__(self, value):
        if value not in self.POSSIBLE_VALUES:
            raise ValueError("'%s' is not a valid value for %s"
                             % (value, self.__class__.__name__))

    def __eq__(self, other):
        if other not in self.POSSIBLE_VALUES:
            raise EnumStringComparisonError(
                'Can only compare with %s'
                % ', '.join("'%s'" % v for v in self.POSSIBLE_VALUES))
        return super(EnumString, self).__eq__(other)

    def __ne__(self, other):
        return not (self == other)

    def __hash__(self):
        return super(EnumString, self).__hash__()

    @staticmethod
    def subclass(*possible_values):
        class EnumStringSubclass(EnumString):
            POSSIBLE_VALUES = possible_values
        return EnumStringSubclass


def _escape_char(c):
    # str.encode('unicode_espace') doesn't escape quotes, presumably because
    # quoting could be done with either ' or ".
    if c == "'":
        return "\\'"
    return six.text_type(c.encode('unicode_escape'))


# The default PrettyPrinter has some issues with UTF-8, so we need to override
# some stuff here.
class _PrettyPrinter(pprint.PrettyPrinter):
    def format(self, object, context, maxlevels, level):
        if not (isinstance(object, six.text_type) or
                isinstance(object, six.binary_type)):
            return super(_PrettyPrinter, self).format(
                object, context, maxlevels, level)
        # This is super hacky and weird, but the output of 'repr' actually
        # varies based on the default I/O encoding of the process, which isn't
        # necessarily utf-8. Instead we open a new shell and ask what the repr
        # WOULD be assuming the default encoding is utf-8. If you can come up
        # with a better way of doing this without simply re-implementing the
        # logic of "repr", please replace this.
        env = dict(os.environ)
        env['PYTHONIOENCODING'] = 'utf-8'
        ret = six.ensure_text(subprocess.check_output(
            [sys.executable], input='print(repr(%s))' % repr(object),
            universal_newlines=True, env=env, encoding='utf-8')).strip()
        return (ret, True, False)


if six.PY2:  # Delete when we get rid of Python 2.
    # Mapping table between raw characters below \x80 and their escaped
    # counterpart, when they differ
    _INDENTED_REPR_TABLE = {
        c: e
        for c, e in map(lambda x: (x, _escape_char(x)),
                        map(unichr, range(128)))
        if c != e
    }
    # Regexp matching all characters to escape.
    _INDENTED_REPR_RE = re.compile(
        '([' + ''.join(_INDENTED_REPR_TABLE.values()) + ']+)')


def indented_repr(o, indent=4):
    '''Similar to repr(), but returns an indented representation of the object

    One notable difference with repr is that the returned representation
    assumes `from __future__ import unicode_literals`.
    '''
    if six.PY3:
        return _PrettyPrinter(indent=indent).pformat(o)
    # Delete everything below when we get rid of Python 2.
    one_indent = ' ' * indent

    def recurse_indented_repr(o, level):
        if isinstance(o, dict):
            yield '{\n'
            for k, v in sorted(o.items()):
                yield one_indent * (level + 1)
                for d in recurse_indented_repr(k, level + 1):
                    yield d
                yield ': '
                for d in recurse_indented_repr(v, level + 1):
                    yield d
                yield ',\n'
            yield one_indent * level
            yield '}'
        elif isinstance(o, bytes):
            yield 'b'
            yield repr(o)
        elif isinstance(o, six.text_type):
            yield "'"
            # We want a readable string (non escaped unicode), but some
            # special characters need escaping (e.g. \n, \t, etc.)
            for i, s in enumerate(_INDENTED_REPR_RE.split(o)):
                if i % 2:
                    for c in s:
                        yield _INDENTED_REPR_TABLE[c]
                else:
                    yield s
            yield "'"
        elif hasattr(o, '__iter__'):
            yield '[\n'
            for i in o:
                yield one_indent * (level + 1)
                for d in recurse_indented_repr(i, level + 1):
                    yield d
                yield ',\n'
            yield one_indent * level
            yield ']'
        else:
            yield repr(o)
    return ''.join(recurse_indented_repr(o, 0))


def patch_main():
    '''This is a hack to work around the fact that Windows multiprocessing needs
    to import the original main module, and assumes that it corresponds to a file
    ending in .py.

    We do this by a sort of two-level function interposing. The first
    level interposes forking.get_command_line() with our version defined
    in my_get_command_line(). Our version of get_command_line will
    replace the command string with the contents of the fork_interpose()
    function to be used in the subprocess.

    The subprocess then gets an interposed imp.find_module(), which we
    hack up to find the main module name multiprocessing will assume, since we
    know what this will be based on the main module in the parent. If we're not
    looking for our main module, then the original find_module will suffice.

    See also: http://bugs.python.org/issue19946
    And: https://bugzilla.mozilla.org/show_bug.cgi?id=914563
    '''
    # XXX In Python 3.4 the multiprocessing module was re-written and the below
    # code is no longer valid. The Python issue19946 also claims to be fixed in
    # this version. It's not clear whether this hack is still needed in 3.4+ or
    # not, but at least some basic mach commands appear to work without it. So
    # skip it in 3.4+ until we determine it's still needed.
    if sys.platform == 'win32' and sys.version_info < (3, 4):
        import inspect
        import os
        from multiprocessing import forking
        global orig_command_line

        # Figure out what multiprocessing will assume our main module
        # is called (see python/Lib/multiprocessing/forking.py).
        main_path = getattr(sys.modules['__main__'], '__file__', None)
        if main_path is None:
            # If someone deleted or modified __main__, there's nothing left for
            # us to do.
            return
        main_file_name = os.path.basename(main_path)
        main_module_name, ext = os.path.splitext(main_file_name)
        if ext == '.py':
            # If main is a .py file, everything ought to work as expected.
            return

        def fork_interpose():
            import imp
            import os
            import sys
            orig_find_module = imp.find_module

            def my_find_module(name, dirs):
                if name == main_module_name:
                    path = os.path.join(dirs[0], main_file_name)
                    f = open(path)
                    return (f, path, ('', 'r', imp.PY_SOURCE))
                return orig_find_module(name, dirs)

            # Don't allow writing bytecode file for the main module.
            orig_load_module = imp.load_module

            def my_load_module(name, file, path, description):
                # multiprocess.forking invokes imp.load_module manually and
                # hard-codes the name __parents_main__ as the module name.
                if name == '__parents_main__':
                    old_bytecode = sys.dont_write_bytecode
                    sys.dont_write_bytecode = True
                    try:
                        return orig_load_module(name, file, path, description)
                    finally:
                        sys.dont_write_bytecode = old_bytecode

                return orig_load_module(name, file, path, description)

            imp.find_module = my_find_module
            imp.load_module = my_load_module
            from multiprocessing.forking import main
            main()

        def my_get_command_line():
            fork_code, lineno = inspect.getsourcelines(fork_interpose)
            # Remove the first line (for 'def fork_interpose():') and the three
            # levels of indentation (12 spaces), add our relevant globals.
            fork_string = ("main_file_name = '%s'\n" % main_file_name +
                           "main_module_name = '%s'\n" % main_module_name +
                           ''.join(x[12:] for x in fork_code[1:]))
            cmdline = orig_command_line()
            cmdline[2] = fork_string
            return cmdline
        orig_command_line = forking.get_command_line
        forking.get_command_line = my_get_command_line


def ensure_bytes(value, encoding='utf-8'):
    if isinstance(value, six.text_type):
        return value.encode(encoding)
    return value


def ensure_unicode(value, encoding='utf-8'):
    if isinstance(value, six.binary_type):
        return value.decode(encoding)
    return value


def ensure_subprocess_env(env, encoding='utf-8'):
    """Ensure the environment is in the correct format for the `subprocess`
    module.

    This will convert all keys and values to bytes on Python 2, and text on
    Python 3.

    Args:
        env (dict): Environment to ensure.
        encoding (str): Encoding to use when converting to/from bytes/text
                        (default: utf-8).
    """
    ensure = ensure_bytes if sys.version_info[0] < 3 else ensure_unicode
    return {ensure(k, encoding): ensure(v, encoding) for k, v in six.iteritems(env)}