path: root/lib/spack/llnl/util/filesystem.py

# Copyright 2013-2024 Lawrence Livermore National Security, LLC and other
# Spack Project Developers. See the top-level COPYRIGHT file for details.
#
# SPDX-License-Identifier: (Apache-2.0 OR MIT)
import collections
import collections.abc
import errno
import fnmatch
import glob
import hashlib
import itertools
import numbers
import os
import pathlib
import posixpath
import re
import shutil
import stat
import sys
import tempfile
from contextlib import contextmanager
from itertools import accumulate
from typing import Callable, Iterable, List, Match, Optional, Tuple, Union

import llnl.util.symlink
from llnl.util import tty
from llnl.util.lang import dedupe, memoized
from llnl.util.symlink import islink, readlink, resolve_link_target_relative_to_the_link, symlink

from spack.util.executable import Executable, which

from ..path import path_to_os_path, system_path_filter

if sys.platform != "win32":
    import grp
    import pwd
else:
    import win32security


__all__ = [
    "FileFilter",
    "FileList",
    "HeaderList",
    "LibraryList",
    "ancestor",
    "can_access",
    "change_sed_delimiter",
    "copy_mode",
    "filter_file",
    "find",
    "find_headers",
    "find_all_headers",
    "find_libraries",
    "find_system_libraries",
    "fix_darwin_install_name",
    "force_remove",
    "force_symlink",
    "getuid",
    "chgrp",
    "chmod_x",
    "copy",
    "install",
    "copy_tree",
    "install_tree",
    "is_exe",
    "join_path",
    "last_modification_time_recursive",
    "library_extensions",
    "mkdirp",
    "partition_path",
    "prefixes",
    "remove_dead_links",
    "remove_directory_contents",
    "remove_if_dead_link",
    "remove_linked_tree",
    "rename",
    "set_executable",
    "set_install_permissions",
    "touch",
    "touchp",
    "traverse_tree",
    "unset_executable_mode",
    "working_dir",
    "keep_modification_time",
    "BaseDirectoryVisitor",
    "visit_directory_tree",
]

if sys.version_info < (3, 7, 4):
    # monkeypatch shutil.copystat to fix PermissionError when copying read-only
    # files on Lustre when using Python < 3.7.4

    def copystat(src, dst, follow_symlinks=True):
        """Copy file metadata
        Copy the permission bits, last access time, last modification time, and
        flags from `src` to `dst`. On Linux, copystat() also copies the "extended
        attributes" where possible. The file contents, owner, and group are
        unaffected. `src` and `dst` are path names given as strings.
        If the optional flag `follow_symlinks` is not set, symlinks aren't
        followed if and only if both `src` and `dst` are symlinks.
        """

        def _nop(args, ns=None, follow_symlinks=None):
            pass

        # follow symlinks (aka don't not follow symlinks)
        follow = follow_symlinks or not (islink(src) and islink(dst))
        if follow:
            # use the real function if it exists
            def lookup(name):
                return getattr(os, name, _nop)

        else:
            # use the real function only if it exists
            # *and* it supports follow_symlinks
            def lookup(name):
                fn = getattr(os, name, _nop)
                if sys.version_info >= (3, 3):
                    if fn in os.supports_follow_symlinks:  # novermin
                        return fn
                return _nop

        st = lookup("stat")(src, follow_symlinks=follow)
        mode = stat.S_IMODE(st.st_mode)
        lookup("utime")(dst, ns=(st.st_atime_ns, st.st_mtime_ns), follow_symlinks=follow)

        # We must copy extended attributes before the file is (potentially)
        # chmod()'ed read-only, otherwise setxattr() will error with -EACCES.
        shutil._copyxattr(src, dst, follow_symlinks=follow)

        try:
            lookup("chmod")(dst, mode, follow_symlinks=follow)
        except NotImplementedError:
            # if we got a NotImplementedError, it's because
            #   * follow_symlinks=False,
            #   * lchown() is unavailable, and
            #   * either
            #       * fchownat() is unavailable or
            #       * fchownat() doesn't implement AT_SYMLINK_NOFOLLOW.
            #         (it returned ENOSUP.)
            # therefore we're out of options--we simply cannot chown the
            # symlink.  give up, suppress the error.
            # (which is what shutil always did in this circumstance.)
            pass
        if hasattr(st, "st_flags"):
            try:
                lookup("chflags")(dst, st.st_flags, follow_symlinks=follow)
            except OSError as why:
                for err in "EOPNOTSUPP", "ENOTSUP":
                    if hasattr(errno, err) and why.errno == getattr(errno, err):
                        break
                else:
                    raise

    shutil.copystat = copystat


def polite_path(components: Iterable[str]):
    """
    Given a list of strings which are intended to be path components,
    generate a path, and format each component to avoid generating extra
    path entries.

    For example all "/", "\", and ":" characters will be replaced with
    "_". Other characters like "=" will also be replaced.
    """
    return os.path.join(*[polite_filename(x) for x in components])


@memoized
def _polite_antipattern():
    # A regex of all the characters we don't want in a filename
    return re.compile(r"[^A-Za-z0-9_.-]")


def polite_filename(filename: str) -> str:
    """
    Replace generally problematic filename characters with underscores.

    This differs from sanitize_filename in that it is more aggressive in
    changing characters in the name. For example it removes "=" which can
    confuse path parsing in external tools.
    """
    # This character set applies for both Windows and Linux. It does not
    # account for reserved filenames in Windows.
    return _polite_antipattern().sub("_", filename)


def getuid():
    if sys.platform == "win32":
        import ctypes

        if ctypes.windll.shell32.IsUserAnAdmin() == 0:
            return 1
        return 0
    else:
        return os.getuid()


@system_path_filter
def rename(src, dst):
    # On Windows, os.rename will fail if the destination file already exists
    if sys.platform == "win32":
        # Windows path existence checks will sometimes fail on junctions/links/symlinks
        # so check for that case
        if os.path.exists(dst) or islink(dst):
            os.remove(dst)
    os.rename(src, dst)


@system_path_filter
def path_contains_subdirectory(path, root):
    norm_root = os.path.abspath(root).rstrip(os.path.sep) + os.path.sep
    norm_path = os.path.abspath(path).rstrip(os.path.sep) + os.path.sep
    return norm_path.startswith(norm_root)


@memoized
def file_command(*args):
    """Creates entry point to `file` system command with provided arguments"""
    file_cmd = which("file", required=True)
    for arg in args:
        file_cmd.add_default_arg(arg)
    return file_cmd


@memoized
def _get_mime_type():
    """Generate method to call `file` system command to aquire mime type
    for a specified path
    """
    if sys.platform == "win32":
        # -h option (no-dereference) does not exist in Windows
        return file_command("-b", "--mime-type")
    else:
        return file_command("-b", "-h", "--mime-type")


@memoized
def _get_mime_type_compressed():
    """Same as _get_mime_type but attempts to check for
    compression first
    """
    mime_uncompressed = _get_mime_type()
    mime_uncompressed.add_default_arg("-Z")
    return mime_uncompressed


def mime_type(filename):
    """Returns the mime type and subtype of a file.

    Args:
        filename: file to be analyzed

    Returns:
        Tuple containing the MIME type and subtype
    """
    output = _get_mime_type()(filename, output=str, error=str).strip()
    tty.debug("==> " + output)
    type, _, subtype = output.partition("/")
    return type, subtype


def compressed_mime_type(filename):
    """Same as mime_type but checks for type that has been compressed

    Args:
        filename (str): file to be analyzed

    Returns:
        Tuple containing the MIME type and subtype
    """
    output = _get_mime_type_compressed()(filename, output=str, error=str).strip()
    tty.debug("==> " + output)
    type, _, subtype = output.partition("/")
    return type, subtype


#: This generates the library filenames that may appear on any OS.
library_extensions = ["a", "la", "so", "tbd", "dylib"]


def possible_library_filenames(library_names):
    """Given a collection of library names like 'libfoo', generate the set of
    library filenames that may be found on the system (e.g. libfoo.so).
    """
    lib_extensions = library_extensions
    return set(
        ".".join((lib, extension))
        for lib, extension in itertools.product(library_names, lib_extensions)
    )


def paths_containing_libs(paths, library_names):
    """Given a collection of filesystem paths, return the list of paths that
    which include one or more of the specified libraries.
    """
    required_lib_fnames = possible_library_filenames(library_names)

    rpaths_to_include = []
    paths = path_to_os_path(*paths)
    for path in paths:
        fnames = set(os.listdir(path))
        if fnames & required_lib_fnames:
            rpaths_to_include.append(path)

    return rpaths_to_include


@system_path_filter
def same_path(path1, path2):
    norm1 = os.path.abspath(path1).rstrip(os.path.sep)
    norm2 = os.path.abspath(path2).rstrip(os.path.sep)
    return norm1 == norm2


def filter_file(
    regex: str,
    repl: Union[str, Callable[[Match], str]],
    *filenames: str,
    string: bool = False,
    backup: bool = False,
    ignore_absent: bool = False,
    start_at: Optional[str] = None,
    stop_at: Optional[str] = None,
) -> None:
    r"""Like sed, but uses python regular expressions.

    Filters every line of each file through regex and replaces the file
    with a filtered version.  Preserves mode of filtered files.

    As with re.sub, ``repl`` can be either a string or a callable.
    If it is a callable, it is passed the match object and should
    return a suitable replacement string.  If it is a string, it
    can contain ``\1``, ``\2``, etc. to represent back-substitution
    as sed would allow.

    Args:
        regex (str): The regular expression to search for
        repl (str): The string to replace matches with
        *filenames: One or more files to search and replace
        string (bool): Treat regex as a plain string. Default it False
        backup (bool): Make backup file(s) suffixed with ``~``. Default is False
        ignore_absent (bool): Ignore any files that don't exist.
            Default is False
        start_at (str): Marker used to start applying the replacements. If a
            text line matches this marker filtering is started at the next line.
            All contents before the marker and the marker itself are copied
            verbatim. Default is to start filtering from the first line of the
            file.
        stop_at (str): Marker used to stop scanning the file further. If a text
            line matches this marker filtering is stopped and the rest of the
            file is copied verbatim. Default is to filter until the end of the
            file.
    """
    # Allow strings to use \1, \2, etc. for replacement, like sed
    if not callable(repl):
        unescaped = repl.replace(r"\\", "\\")

        def replace_groups_with_groupid(m: Match) -> str:
            def groupid_to_group(x):
                return m.group(int(x.group(1)))

            return re.sub(r"\\([1-9])", groupid_to_group, unescaped)

        repl = replace_groups_with_groupid

    if string:
        regex = re.escape(regex)
    for filename in path_to_os_path(*filenames):
        msg = 'FILTER FILE: {0} [replacing "{1}"]'
        tty.debug(msg.format(filename, regex))

        backup_filename = filename + "~"
        tmp_filename = filename + ".spack~"

        if ignore_absent and not os.path.exists(filename):
            msg = 'FILTER FILE: file "{0}" not found. Skipping to next file.'
            tty.debug(msg.format(filename))
            continue

        # Create backup file. Don't overwrite an existing backup
        # file in case this file is being filtered multiple times.
        if not os.path.exists(backup_filename):
            shutil.copy(filename, backup_filename)

        # Create a temporary file to read from. We cannot use backup_filename
        # in case filter_file is invoked multiple times on the same file.
        shutil.copy(filename, tmp_filename)

        try:
            # Open as a text file and filter until the end of the file is
            # reached, or we found a marker in the line if it was specified
            #
            # To avoid translating line endings (\n to \r\n and vice-versa)
            # we force os.open to ignore translations and use the line endings
            # the file comes with
            with open(tmp_filename, mode="r", errors="surrogateescape", newline="") as input_file:
                with open(filename, mode="w", errors="surrogateescape", newline="") as output_file:
                    do_filtering = start_at is None
                    # Using iter and readline is a workaround needed not to
                    # disable input_file.tell(), which will happen if we call
                    # input_file.next() implicitly via the for loop
                    for line in iter(input_file.readline, ""):
                        if stop_at is not None:
                            current_position = input_file.tell()
                            if stop_at == line.strip():
                                output_file.write(line)
                                break
                        if do_filtering:
                            filtered_line = re.sub(regex, repl, line)
                            output_file.write(filtered_line)
                        else:
                            do_filtering = start_at == line.strip()
                            output_file.write(line)
                    else:
                        current_position = None

            # If we stopped filtering at some point, reopen the file in
            # binary mode and copy verbatim the remaining part
            if current_position and stop_at:
                with open(tmp_filename, mode="rb") as input_binary_buffer:
                    input_binary_buffer.seek(current_position)
                    with open(filename, mode="ab") as output_binary_buffer:
                        output_binary_buffer.writelines(input_binary_buffer.readlines())

        except BaseException:
            # clean up the original file on failure.
            shutil.move(backup_filename, filename)
            raise

        finally:
            os.remove(tmp_filename)
            if not backup and os.path.exists(backup_filename):
                os.remove(backup_filename)


class FileFilter:
    """Convenience class for calling ``filter_file`` a lot."""

    def __init__(self, *filenames):
        self.filenames = filenames

    def filter(
        self,
        regex: str,
        repl: Union[str, Callable[[Match], str]],
        string: bool = False,
        backup: bool = False,
        ignore_absent: bool = False,
        start_at: Optional[str] = None,
        stop_at: Optional[str] = None,
    ) -> None:
        return filter_file(
            regex,
            repl,
            *self.filenames,
            string=string,
            backup=backup,
            ignore_absent=ignore_absent,
            start_at=start_at,
            stop_at=stop_at,
        )


def change_sed_delimiter(old_delim, new_delim, *filenames):
    """Find all sed search/replace commands and change the delimiter.

    e.g., if the file contains seds that look like ``'s///'``, you can
    call ``change_sed_delimiter('/', '@', file)`` to change the
    delimiter to ``'@'``.

    Note that this routine will fail if the delimiter is ``'`` or ``"``.
    Handling those is left for future work.

    Parameters:
        old_delim (str): The delimiter to search for
        new_delim (str): The delimiter to replace with
        *filenames: One or more files to search and replace
    """
    assert len(old_delim) == 1
    assert len(new_delim) == 1

    # TODO: handle these cases one day?
    assert old_delim != '"'
    assert old_delim != "'"
    assert new_delim != '"'
    assert new_delim != "'"

    whole_lines = "^s@([^@]*)@(.*)@[gIp]$"
    whole_lines = whole_lines.replace("@", old_delim)

    single_quoted = r"'s@((?:\\'|[^@'])*)@((?:\\'|[^'])*)@[gIp]?'"
    single_quoted = single_quoted.replace("@", old_delim)

    double_quoted = r'"s@((?:\\"|[^@"])*)@((?:\\"|[^"])*)@[gIp]?"'
    double_quoted = double_quoted.replace("@", old_delim)

    repl = r"s@\1@\2@g"
    repl = repl.replace("@", new_delim)
    filenames = path_to_os_path(*filenames)
    for f in filenames:
        filter_file(whole_lines, repl, f)
        filter_file(single_quoted, "'%s'" % repl, f)
        filter_file(double_quoted, '"%s"' % repl, f)


@contextmanager
def exploding_archive_catch(stage):
    # Check for an exploding tarball, i.e. one that doesn't expand to
    # a single directory.  If the tarball *didn't* explode, move its
    # contents to the staging source directory & remove the container
    # directory.  If the tarball did explode, just rename the tarball
    # directory to the staging source directory.
    #
    # NOTE: The tar program on Mac OS X will encode HFS metadata in
    # hidden files, which can end up *alongside* a single top-level
    # directory.  We initially ignore presence of hidden files to
    # accomodate these "semi-exploding" tarballs but ensure the files
    # are copied to the source directory.

    # Expand all tarballs in their own directory to contain
    # exploding tarballs.
    tarball_container = os.path.join(stage.path, "spack-expanded-archive")
    mkdirp(tarball_container)
    orig_dir = os.getcwd()
    os.chdir(tarball_container)
    try:
        yield
        # catch an exploding archive on sucessful extraction
        os.chdir(orig_dir)
        exploding_archive_handler(tarball_container, stage)
    except Exception as e:
        # return current directory context to previous on failure
        os.chdir(orig_dir)
        raise e


@system_path_filter
def exploding_archive_handler(tarball_container, stage):
    """
    Args:
        tarball_container: where the archive was expanded to
        stage: Stage object referencing filesystem location
            where archive is being expanded
    """
    files = os.listdir(tarball_container)
    non_hidden = [f for f in files if not f.startswith(".")]
    if len(non_hidden) == 1:
        src = os.path.join(tarball_container, non_hidden[0])
        if os.path.isdir(src):
            stage.srcdir = non_hidden[0]
            shutil.move(src, stage.source_path)
            if len(files) > 1:
                files.remove(non_hidden[0])
                for f in files:
                    src = os.path.join(tarball_container, f)
                    dest = os.path.join(stage.path, f)
                    shutil.move(src, dest)
            os.rmdir(tarball_container)
        else:
            # This is a non-directory entry (e.g., a patch file) so simply
            # rename the tarball container to be the source path.
            shutil.move(tarball_container, stage.source_path)
    else:
        shutil.move(tarball_container, stage.source_path)


@system_path_filter(arg_slice=slice(1))
def get_owner_uid(path, err_msg=None):
    if not os.path.exists(path):
        mkdirp(path, mode=stat.S_IRWXU)

        p_stat = os.stat(path)
        if p_stat.st_mode & stat.S_IRWXU != stat.S_IRWXU:
            tty.error(
                "Expected {0} to support mode {1}, but it is {2}".format(
                    path, stat.S_IRWXU, p_stat.st_mode
                )
            )

            raise OSError(errno.EACCES, err_msg.format(path, path) if err_msg else "")
    else:
        p_stat = os.stat(path)

    if sys.platform != "win32":
        owner_uid = p_stat.st_uid
    else:
        sid = win32security.GetFileSecurity(
            path, win32security.OWNER_SECURITY_INFORMATION
        ).GetSecurityDescriptorOwner()
        owner_uid = win32security.LookupAccountSid(None, sid)[0]
    return owner_uid


@system_path_filter
def set_install_permissions(path):
    """Set appropriate permissions on the installed file."""
    # If this points to a file maintained in a Spack prefix, it is assumed that
    # this function will be invoked on the target. If the file is outside a
    # Spack-maintained prefix, the permissions should not be modified.
    if islink(path):
        return
    if os.path.isdir(path):
        os.chmod(path, 0o755)
    else:
        os.chmod(path, 0o644)


def group_ids(uid=None):
    """Get group ids that a uid is a member of.

    Arguments:
        uid (int): id of user, or None for current user

    Returns:
        (list of int): gids of groups the user is a member of
    """
    if sys.platform == "win32":
        tty.warn("Function is not supported on Windows")
        return []

    if uid is None:
        uid = getuid()

    pwd_entry = pwd.getpwuid(uid)
    user = pwd_entry.pw_name

    # user's primary group id may not be listed in grp (i.e. /etc/group)
    # you have to check pwd for that, so start the list with that
    gids = [pwd_entry.pw_gid]

    return sorted(set(gids + [g.gr_gid for g in grp.getgrall() if user in g.gr_mem]))


@system_path_filter(arg_slice=slice(1))
def chgrp(path, group, follow_symlinks=True):
    """Implement the bash chgrp function on a single path"""
    if sys.platform == "win32":
        raise OSError("Function 'chgrp' is not supported on Windows")

    if isinstance(group, str):
        gid = grp.getgrnam(group).gr_gid
    else:
        gid = group
    if os.stat(path).st_gid == gid:
        return
    if follow_symlinks:
        os.chown(path, -1, gid)
    else:
        os.lchown(path, -1, gid)


@system_path_filter(arg_slice=slice(1))
def chmod_x(entry, perms):
    """Implements chmod, treating all executable bits as set using the chmod
    utility's `+X` option.
    """
    mode = os.stat(entry).st_mode
    if os.path.isfile(entry):
        if not mode & (stat.S_IXUSR | stat.S_IXGRP | stat.S_IXOTH):
            perms &= ~stat.S_IXUSR
            perms &= ~stat.S_IXGRP
            perms &= ~stat.S_IXOTH
    os.chmod(entry, perms)


@system_path_filter
def copy_mode(src, dest):
    """Set the mode of dest to that of src unless it is a link."""
    if islink(dest):
        return
    src_mode = os.stat(src).st_mode
    dest_mode = os.stat(dest).st_mode
    if src_mode & stat.S_IXUSR:
        dest_mode |= stat.S_IXUSR
    if src_mode & stat.S_IXGRP:
        dest_mode |= stat.S_IXGRP
    if src_mode & stat.S_IXOTH:
        dest_mode |= stat.S_IXOTH
    os.chmod(dest, dest_mode)


@system_path_filter
def unset_executable_mode(path):
    mode = os.stat(path).st_mode
    mode &= ~stat.S_IXUSR
    mode &= ~stat.S_IXGRP
    mode &= ~stat.S_IXOTH
    os.chmod(path, mode)


@system_path_filter
def copy(src, dest, _permissions=False):
    """Copy the file(s) *src* to the file or directory *dest*.

    If *dest* specifies a directory, the file will be copied into *dest*
    using the base filename from *src*.

    *src* may contain glob characters.

    Parameters:
        src (str): the file(s) to copy
        dest (str): the destination file or directory
        _permissions (bool): for internal use only

    Raises:
        IOError: if *src* does not match any files or directories
        ValueError: if *src* matches multiple files but *dest* is
            not a directory
    """
    if _permissions:
        tty.debug("Installing {0} to {1}".format(src, dest))
    else:
        tty.debug("Copying {0} to {1}".format(src, dest))

    files = glob.glob(src)
    if not files:
        raise IOError("No such file or directory: '{0}'".format(src))
    if len(files) > 1 and not os.path.isdir(dest):
        raise ValueError(
            "'{0}' matches multiple files but '{1}' is not a directory".format(src, dest)
        )

    for src in files:
        # Expand dest to its eventual full path if it is a directory.
        dst = dest
        if os.path.isdir(dest):
            dst = join_path(dest, os.path.basename(src))

        shutil.copy(src, dst)

        if _permissions:
            set_install_permissions(dst)
            copy_mode(src, dst)


@system_path_filter
def install(src, dest):
    """Install the file(s) *src* to the file or directory *dest*.

    Same as :py:func:`copy` with the addition of setting proper
    permissions on the installed file.

    Parameters:
        src (str): the file(s) to install
        dest (str): the destination file or directory

    Raises:
        IOError: if *src* does not match any files or directories
        ValueError: if *src* matches multiple files but *dest* is
            not a directory
    """
    copy(src, dest, _permissions=True)


@system_path_filter
def copy_tree(
    src: str,
    dest: str,
    symlinks: bool = True,
    allow_broken_symlinks: bool = sys.platform != "win32",
    ignore: Optional[Callable[[str], bool]] = None,
    _permissions: bool = False,
):
    """Recursively copy an entire directory tree rooted at *src*.

    If the destination directory *dest* does not already exist, it will
    be created as well as missing parent directories.

    *src* may contain glob characters.

    If *symlinks* is true, symbolic links in the source tree are represented
    as symbolic links in the new tree and the metadata of the original links
    will be copied as far as the platform allows; if false, the contents and
    metadata of the linked files are copied to the new tree.

    If *ignore* is set, then each path relative to *src* will be passed to
    this function; the function returns whether that path should be skipped.

    Parameters:
        src (str): the directory to copy
        dest (str): the destination directory
        symlinks (bool): whether or not to preserve symlinks
        allow_broken_symlinks (bool): whether or not to allow broken (dangling) symlinks,
            On Windows, setting this to True will raise an exception. Defaults to true on unix.
        ignore (typing.Callable): function indicating which files to ignore
        _permissions (bool): for internal use only

    Raises:
        IOError: if *src* does not match any files or directories
        ValueError: if *src* is a parent directory of *dest*
    """
    if allow_broken_symlinks and sys.platform == "win32":
        raise llnl.util.symlink.SymlinkError("Cannot allow broken symlinks on Windows!")
    if _permissions:
        tty.debug("Installing {0} to {1}".format(src, dest))
    else:
        tty.debug("Copying {0} to {1}".format(src, dest))

    abs_dest = os.path.abspath(dest)
    if not abs_dest.endswith(os.path.sep):
        abs_dest += os.path.sep

    files = glob.glob(src)
    if not files:
        raise IOError("No such file or directory: '{0}'".format(src))

    # For Windows hard-links and junctions, the source path must exist to make a symlink. Add
    # all symlinks to this list while traversing the tree, then when finished, make all
    # symlinks at the end.
    links = []

    for src in files:
        abs_src = os.path.abspath(src)
        if not abs_src.endswith(os.path.sep):
            abs_src += os.path.sep

        # Stop early to avoid unnecessary recursion if being asked to copy
        # from a parent directory.
        if abs_dest.startswith(abs_src):
            raise ValueError(
                "Cannot copy ancestor directory {0} into {1}".format(abs_src, abs_dest)
            )

        mkdirp(abs_dest)

        for s, d in traverse_tree(
            abs_src,
            abs_dest,
            order="pre",
            follow_links=not symlinks,
            ignore=ignore,
            follow_nonexisting=True,
        ):
            if islink(s):
                link_target = resolve_link_target_relative_to_the_link(s)
                if symlinks:
                    target = os.readlink(s)
                    if os.path.isabs(target):

                        def escaped_path(path):
                            return path.replace("\\", r"\\")

                        new_target = re.sub(escaped_path(abs_src), escaped_path(abs_dest), target)
                        if new_target != target:
                            tty.debug("Redirecting link {0} to {1}".format(target, new_target))
                            target = new_target

                    links.append((target, d, s))
                    continue

                elif os.path.isdir(link_target):
                    mkdirp(d)
                else:
                    shutil.copyfile(s, d)
            else:
                if os.path.isdir(s):
                    mkdirp(d)
                else:
                    shutil.copy2(s, d)

            if _permissions:
                set_install_permissions(d)
                copy_mode(s, d)

    for target, d, s in links:
        symlink(target, d, allow_broken_symlinks=allow_broken_symlinks)
        if _permissions:
            set_install_permissions(d)
            copy_mode(s, d)


@system_path_filter
def install_tree(
    src, dest, symlinks=True, ignore=None, allow_broken_symlinks=sys.platform != "win32"
):
    """Recursively install an entire directory tree rooted at *src*.

    Same as :py:func:`copy_tree` with the addition of setting proper
    permissions on the installed files and directories.

    Parameters:
        src (str): the directory to install
        dest (str): the destination directory
        symlinks (bool): whether or not to preserve symlinks
        ignore (typing.Callable): function indicating which files to ignore
        allow_broken_symlinks (bool): whether or not to allow broken (dangling) symlinks,
            On Windows, setting this to True will raise an exception.

    Raises:
        IOError: if *src* does not match any files or directories
        ValueError: if *src* is a parent directory of *dest*
    """
    copy_tree(
        src,
        dest,
        symlinks=symlinks,
        allow_broken_symlinks=allow_broken_symlinks,
        ignore=ignore,
        _permissions=True,
    )


@system_path_filter
def is_exe(path):
    """True if path is an executable file."""
    return os.path.isfile(path) and os.access(path, os.X_OK)


@system_path_filter
def get_filetype(path_name):
    """
    Return the output of file path_name as a string to identify file type.
    """
    file = Executable("file")
    file.add_default_env("LC_ALL", "C")
    output = file("-b", "-h", "%s" % path_name, output=str, error=str)
    return output.strip()


@system_path_filter
def is_nonsymlink_exe_with_shebang(path):
    """
    Returns whether the path is an executable script with a shebang.
    Return False when the path is a *symlink* to an executable script.
    """
    try:
        st = os.lstat(path)
        # Should not be a symlink
        if stat.S_ISLNK(st.st_mode):
            return False

        # Should be executable
        if not st.st_mode & (stat.S_IXUSR | stat.S_IXGRP | stat.S_IXOTH):
            return False

        # Should start with a shebang
        with open(path, "rb") as f:
            return f.read(2) == b"#!"
    except (IOError, OSError):
        return False


@system_path_filter(arg_slice=slice(1))
def chgrp_if_not_world_writable(path, group):
    """chgrp path to group if path is not world writable"""
    mode = os.stat(path).st_mode
    if not mode & stat.S_IWOTH:
        chgrp(path, group)


def mkdirp(
    *paths: str,
    mode: Optional[int] = None,
    group: Optional[Union[str, int]] = None,
    default_perms: Optional[str] = None,
):
    """Creates a directory, as well as parent directories if needed.

    Arguments:
        paths: paths to create with mkdirp
        mode: optional permissions to set on the created directory -- use OS default
            if not provided
        group: optional group for permissions of final created directory -- use OS
            default if not provided. Only used if world write permissions are not set
        default_perms: one of 'parents' or 'args'. The default permissions that are set for
            directories that are not themselves an argument for mkdirp. 'parents' means
            intermediate directories get the permissions of their direct parent directory,
            'args' means intermediate get the same permissions specified in the arguments to
            mkdirp -- default value is 'args'
    """
    default_perms = default_perms or "args"
    paths = path_to_os_path(*paths)
    for path in paths:
        if not os.path.exists(path):
            try:
                last_parent, intermediate_folders = longest_existing_parent(path)

                # create folders
                os.makedirs(path)

                # leaf folder permissions
                if mode is not None:
                    os.chmod(path, mode)
                if group:
                    chgrp_if_not_world_writable(path, group)
                    if mode is not None:
                        os.chmod(path, mode)  # reset sticky grp bit post chgrp

                # for intermediate folders, change mode just for newly created
                # ones and if mode_intermediate has been specified, otherwise
                # intermediate folders list is not populated at all and default
                # OS mode will be used
                if default_perms == "args":
                    intermediate_mode = mode
                    intermediate_group = group
                elif default_perms == "parents":
                    stat_info = os.stat(last_parent)
                    intermediate_mode = stat_info.st_mode
                    intermediate_group = stat_info.st_gid
                else:
                    msg = "Invalid value: '%s'. " % default_perms
                    msg += "Choose from 'args' or 'parents'."
                    raise ValueError(msg)

                for intermediate_path in reversed(intermediate_folders):
                    if intermediate_mode is not None:
                        os.chmod(intermediate_path, intermediate_mode)
                    if intermediate_group is not None:
                        chgrp_if_not_world_writable(intermediate_path, intermediate_group)
                        if intermediate_mode is not None:
                            os.chmod(
                                intermediate_path, intermediate_mode
                            )  # reset sticky bit after

            except OSError as e:
                if e.errno != errno.EEXIST or not os.path.isdir(path):
                    raise e
        elif not os.path.isdir(path):
            raise OSError(errno.EEXIST, "File already exists", path)


def longest_existing_parent(path: str) -> Tuple[str, List[str]]:
    """Return the last existing parent and a list of all intermediate directories
    to be created for the directory passed as input.

    Args:
        path: directory to be created
    """
    # detect missing intermediate folders
    intermediate_folders = []
    last_parent = ""
    intermediate_path = os.path.dirname(path)
    while intermediate_path:
        if os.path.lexists(intermediate_path):
            last_parent = intermediate_path
            break

        intermediate_folders.append(intermediate_path)
        intermediate_path = os.path.dirname(intermediate_path)
    return last_parent, intermediate_folders


@system_path_filter
def force_remove(*paths):
    """Remove files without printing errors.  Like ``rm -f``, does NOT
    remove directories."""
    for path in paths:
        try:
            os.remove(path)
        except OSError:
            pass


@contextmanager
@system_path_filter
def working_dir(dirname: str, *, create: bool = False):
    if create:
        mkdirp(dirname)

    orig_dir = os.getcwd()
    os.chdir(dirname)
    try:
        yield
    finally:
        os.chdir(orig_dir)


class CouldNotRestoreDirectoryBackup(RuntimeError):
    def __init__(self, inner_exception, outer_exception):
        self.inner_exception = inner_exception
        self.outer_exception = outer_exception


@contextmanager
@system_path_filter
def replace_directory_transaction(directory_name):
    """Temporarily renames a directory in the same parent dir. If the operations
    executed within the context manager don't raise an exception, the renamed directory
    is deleted. If there is an exception, the move is undone.

    Args:
        directory_name (path): absolute path of the directory name

    Returns:
        temporary directory where ``directory_name`` has been moved
    """
    # Check the input is indeed a directory with absolute path.
    # Raise before anything is done to avoid moving the wrong directory
    directory_name = os.path.abspath(directory_name)
    assert os.path.isdir(directory_name), "Not a directory: " + directory_name

    # Note: directory_name is normalized here, meaning the trailing slash is dropped,
    # so dirname is the directory's parent not the directory itself.
    tmpdir = tempfile.mkdtemp(dir=os.path.dirname(directory_name), prefix=".backup")

    # We have to jump through hoops to support Windows, since
    # os.rename(directory_name, tmpdir) errors there.
    backup_dir = os.path.join(tmpdir, "backup")
    os.rename(directory_name, backup_dir)
    tty.debug("Directory moved [src={0}, dest={1}]".format(directory_name, backup_dir))

    try:
        yield backup_dir
    except (Exception, KeyboardInterrupt, SystemExit) as inner_exception:
        # Try to recover the original directory, if this fails, raise a
        # composite exception.
        try:
            # Delete what was there, before copying back the original content
            if os.path.exists(directory_name):
                shutil.rmtree(directory_name)
            os.rename(backup_dir, directory_name)
        except Exception as outer_exception:
            raise CouldNotRestoreDirectoryBackup(inner_exception, outer_exception)

        tty.debug("Directory recovered [{0}]".format(directory_name))
        raise
    else:
        # Otherwise delete the temporary directory
        shutil.rmtree(tmpdir, ignore_errors=True)
        tty.debug("Temporary directory deleted [{0}]".format(tmpdir))


@system_path_filter
def hash_directory(directory, ignore=[]):
    """Hashes recursively the content of a directory.

    Args:
        directory (path): path to a directory to be hashed

    Returns:
        hash of the directory content
    """
    assert os.path.isdir(directory), '"directory" must be a directory!'

    md5_hash = hashlib.md5()

    # Adapted from https://stackoverflow.com/a/3431835/771663
    for root, dirs, files in os.walk(directory):
        for name in sorted(files):
            filename = os.path.join(root, name)
            if filename not in ignore:
                # TODO: if caching big files becomes an issue, convert this to
                # TODO: read in chunks. Currently it's used only for testing
                # TODO: purposes.
                with open(filename, "rb") as f:
                    md5_hash.update(f.read())

    return md5_hash.hexdigest()


@contextmanager
@system_path_filter
def write_tmp_and_move(filename):
    """Write to a temporary file, then move into place."""
    dirname = os.path.dirname(filename)
    basename = os.path.basename(filename)
    tmp = os.path.join(dirname, ".%s.tmp" % basename)
    with open(tmp, "w") as f:
        yield f
    shutil.move(tmp, filename)


@contextmanager
@system_path_filter
def open_if_filename(str_or_file, mode="r"):
    """Takes either a path or a file object, and opens it if it is a path.

    If it's a file object, just yields the file object.
    """
    if isinstance(str_or_file, str):
        with open(str_or_file, mode) as f:
            yield f
    else:
        yield str_or_file


@system_path_filter
def touch(path):
    """Creates an empty file at the specified path."""
    if sys.platform == "win32":
        perms = os.O_WRONLY | os.O_CREAT
    else:
        perms = os.O_WRONLY | os.O_CREAT | os.O_NONBLOCK | os.O_NOCTTY
    fd = None
    try:
        fd = os.open(path, perms)
        os.utime(path, None)
    finally:
        if fd is not None:
            os.close(fd)


@system_path_filter
def touchp(path):
    """Like ``touch``, but creates any parent directories needed for the file."""
    mkdirp(os.path.dirname(path))
    touch(path)


@system_path_filter
def force_symlink(src, dest):
    try:
        symlink(src, dest)
    except OSError:
        os.remove(dest)
        symlink(src, dest)


@system_path_filter
def join_path(prefix, *args):
    path = str(prefix)
    for elt in args:
        path = os.path.join(path, str(elt))
    return path


@system_path_filter
def ancestor(dir, n=1):
    """Get the nth ancestor of a directory."""
    parent = os.path.abspath(dir)
    for i in range(n):
        parent = os.path.dirname(parent)
    return parent


@system_path_filter
def get_single_file(directory):
    fnames = os.listdir(directory)
    if len(fnames) != 1:
        raise ValueError("Expected exactly 1 file, got {0}".format(str(len(fnames))))
    return fnames[0]


@contextmanager
def temp_cwd():
    tmp_dir = tempfile.mkdtemp()
    try:
        with working_dir(tmp_dir):
            yield tmp_dir
    finally:
        kwargs = {}
        if sys.platform == "win32":
            kwargs["ignore_errors"] = False
            kwargs["onerror"] = readonly_file_handler(ignore_errors=True)
        shutil.rmtree(tmp_dir, **kwargs)


@contextmanager
@system_path_filter
def temp_rename(orig_path, temp_path):
    same_path = os.path.realpath(orig_path) == os.path.realpath(temp_path)
    if not same_path:
        shutil.move(orig_path, temp_path)
    try:
        yield
    finally:
        if not same_path:
            shutil.move(temp_path, orig_path)


@system_path_filter
def can_access(file_name):
    """True if we have read/write access to the file."""
    return os.access(file_name, os.R_OK | os.W_OK)


@system_path_filter
def traverse_tree(
    source_root: str,
    dest_root: str,
    rel_path: str = "",
    *,
    order: str = "pre",
    ignore: Optional[Callable[[str], bool]] = None,
    follow_nonexisting: bool = True,
    follow_links: bool = False,
):
    """Traverse two filesystem trees simultaneously.

    Walks the LinkTree directory in pre or post order.  Yields each
    file in the source directory with a matching path from the dest
    directory, along with whether the file is a directory.
    e.g., for this tree::

        root/
          a/
            file1
            file2
          b/
            file3

    When called on dest, this yields::

        ('root',         'dest')
        ('root/a',       'dest/a')
        ('root/a/file1', 'dest/a/file1')
        ('root/a/file2', 'dest/a/file2')
        ('root/b',       'dest/b')
        ('root/b/file3', 'dest/b/file3')

    Keyword Arguments:
        order (str): Whether to do pre- or post-order traversal. Accepted
            values are 'pre' and 'post'
        ignore (typing.Callable): function indicating which files to ignore. This will also
            ignore symlinks if they point to an ignored file (regardless of whether the symlink
            is explicitly ignored); note this only supports one layer of indirection (i.e. if
            you have x -> y -> z, and z is ignored but x/y are not, then y would be ignored
            but not x). To avoid this, make sure the ignore function also ignores the symlink
            paths too.
        follow_nonexisting (bool): Whether to descend into directories in
            ``src`` that do not exit in ``dest``. Default is True
        follow_links (bool): Whether to descend into symlinks in ``src``
    """
    if order not in ("pre", "post"):
        raise ValueError("Order must be 'pre' or 'post'.")

    # List of relative paths to ignore under the src root.
    ignore = ignore or (lambda filename: False)

    # Don't descend into ignored directories
    if ignore(rel_path):
        return

    source_path = os.path.join(source_root, rel_path)
    dest_path = os.path.join(dest_root, rel_path)

    # preorder yields directories before children
    if order == "pre":
        yield (source_path, dest_path)

    for f in os.listdir(source_path):
        source_child = os.path.join(source_path, f)
        dest_child = os.path.join(dest_path, f)
        rel_child = os.path.join(rel_path, f)

        # If the source path is a link and the link's source is ignored, then ignore the link too,
        # but only do this if the ignore is defined.
        if ignore is not None:
            if islink(source_child) and not follow_links:
                target = readlink(source_child)
                all_parents = accumulate(target.split(os.sep), lambda x, y: os.path.join(x, y))
                if any(map(ignore, all_parents)):
                    tty.warn(
                        f"Skipping {source_path} because the source or a part of the source's "
                        f"path is included in the ignores."
                    )
                    continue

        # Treat as a directory
        # TODO: for symlinks, os.path.isdir looks for the link target. If the
        # target is relative to the link, then that may not resolve properly
        # relative to our cwd - see resolve_link_target_relative_to_the_link
        if os.path.isdir(source_child) and (follow_links or not islink(source_child)):
            # When follow_nonexisting isn't set, don't descend into dirs
            # in source that do not exist in dest
            if follow_nonexisting or os.path.exists(dest_child):
                tuples = traverse_tree(
                    source_root,
                    dest_root,
                    rel_child,
                    order=order,
                    ignore=ignore,
                    follow_nonexisting=follow_nonexisting,
                    follow_links=follow_links,
                )
                for t in tuples:
                    yield t

        # Treat as a file.
        elif not ignore(os.path.join(rel_path, f)):
            yield (source_child, dest_child)

    if order == "post":
        yield (source_path, dest_path)


def lexists_islink_isdir(path):
    """Computes the tuple (lexists(path), islink(path), isdir(path)) in a minimal
    number of stat calls on unix. Use os.path and symlink.islink methods for windows."""
    if sys.platform == "win32":
        if not os.path.lexists(path):
            return False, False, False
        return os.path.lexists(path), islink(path), os.path.isdir(path)
    # First try to lstat, so we know if it's a link or not.
    try:
        lst = os.lstat(path)
    except (IOError, OSError):
        return False, False, False

    is_link = stat.S_ISLNK(lst.st_mode)

    # Check whether file is a dir.
    if not is_link:
        is_dir = stat.S_ISDIR(lst.st_mode)
        return True, is_link, is_dir

    # Check whether symlink points to a dir.
    try:
        st = os.stat(path)
        is_dir = stat.S_ISDIR(st.st_mode)
    except (IOError, OSError):
        # Dangling symlink (i.e. it lexists but not exists)
        is_dir = False

    return True, is_link, is_dir


class BaseDirectoryVisitor:
    """Base class and interface for :py:func:`visit_directory_tree`."""

    def visit_file(self, root, rel_path, depth):
        """Handle the non-symlink file at ``os.path.join(root, rel_path)``

        Parameters:
            root (str): root directory
            rel_path (str): relative path to current file from ``root``
            depth (int): depth of current file from the ``root`` directory"""
        pass

    def visit_symlinked_file(self, root, rel_path, depth):
        """Handle the symlink to a file at ``os.path.join(root, rel_path)``.
        Note: ``rel_path`` is the location of the symlink, not to what it is
        pointing to. The symlink may be dangling.

        Parameters:
            root (str): root directory
            rel_path (str): relative path to current symlink from ``root``
            depth (int): depth of current symlink from the ``root`` directory"""
        pass

    def before_visit_dir(self, root, rel_path, depth):
        """Return True from this function to recurse into the directory at
        os.path.join(root, rel_path). Return False in order not to recurse further.

        Parameters:
            root (str): root directory
            rel_path (str): relative path to current directory from ``root``
            depth (int): depth of current directory from the ``root`` directory

        Returns:
            bool: ``True`` when the directory should be recursed into. ``False`` when
            not"""
        return False

    def before_visit_symlinked_dir(self, root, rel_path, depth):
        """Return ``True`` to recurse into the symlinked directory and ``False`` in
        order not to. Note: ``rel_path`` is the path to the symlink itself.
        Following symlinked directories blindly can cause infinite recursion due to
        cycles.

        Parameters:
            root (str): root directory
            rel_path (str): relative path to current symlink from ``root``
            depth (int): depth of current symlink from the ``root`` directory

        Returns:
            bool: ``True`` when the directory should be recursed into. ``False`` when
            not"""
        return False

    def after_visit_dir(self, root, rel_path, depth):
        """Called after recursion into ``rel_path`` finished. This function is not
        called when ``rel_path`` was not recursed into.

        Parameters:
            root (str): root directory
            rel_path (str): relative path to current directory from ``root``
            depth (int): depth of current directory from the ``root`` directory"""
        pass

    def after_visit_symlinked_dir(self, root, rel_path, depth):
        """Called after recursion into ``rel_path`` finished. This function is not
        called when ``rel_path`` was not recursed into.

        Parameters:
            root (str): root directory
            rel_path (str): relative path to current symlink from ``root``
            depth (int): depth of current symlink from the ``root`` directory"""
        pass


def visit_directory_tree(root, visitor, rel_path="", depth=0):
    """Recurses the directory root depth-first through a visitor pattern using the
    interface from :py:class:`BaseDirectoryVisitor`

    Parameters:
        root (str): path of directory to recurse into
        visitor (BaseDirectoryVisitor): what visitor to use
        rel_path (str): current relative path from the root
        depth (str): current depth from the root
    """
    dir = os.path.join(root, rel_path)
    dir_entries = sorted(os.scandir(dir), key=lambda d: d.name)

    for f in dir_entries:
        rel_child = os.path.join(rel_path, f.name)
        islink = f.is_symlink()
        # On Windows, symlinks to directories are distinct from
        # symlinks to files, and it is possible to create a
        # broken symlink to a directory (e.g. using os.symlink
        # without `target_is_directory=True`), invoking `isdir`
        # on a symlink on Windows that is broken in this manner
        # will result in an error. In this case we can work around
        # the issue by reading the target and resolving the
        # directory ourselves
        try:
            isdir = f.is_dir()
        except OSError as e:
            if sys.platform == "win32" and hasattr(e, "winerror") and e.winerror == 5 and islink:
                # if path is a symlink, determine destination and
                # evaluate file vs directory
                link_target = resolve_link_target_relative_to_the_link(f)
                # link_target might be relative but
                # resolve_link_target_relative_to_the_link
                # will ensure that if so, that it is relative
                # to the CWD and therefore
                # makes sense
                isdir = os.path.isdir(link_target)
            else:
                raise e

        if not isdir and not islink:
            # handle non-symlink files
            visitor.visit_file(root, rel_child, depth)
        elif not isdir:
            visitor.visit_symlinked_file(root, rel_child, depth)
        elif not islink and visitor.before_visit_dir(root, rel_child, depth):
            # Handle ordinary directories
            visit_directory_tree(root, visitor, rel_child, depth + 1)
            visitor.after_visit_dir(root, rel_child, depth)
        elif islink and visitor.before_visit_symlinked_dir(root, rel_child, depth):
            # Handle symlinked directories
            visit_directory_tree(root, visitor, rel_child, depth + 1)
            visitor.after_visit_symlinked_dir(root, rel_child, depth)


@system_path_filter
def set_executable(path):
    mode = os.stat(path).st_mode
    if mode & stat.S_IRUSR:
        mode |= stat.S_IXUSR
    if mode & stat.S_IRGRP:
        mode |= stat.S_IXGRP
    if mode & stat.S_IROTH:
        mode |= stat.S_IXOTH
    os.chmod(path, mode)


@system_path_filter
def last_modification_time_recursive(path):
    path = os.path.abspath(path)
    times = [os.stat(path).st_mtime]
    times.extend(
        os.lstat(os.path.join(root, name)).st_mtime
        for root, dirs, files in os.walk(path)
        for name in dirs + files
    )
    return max(times)


@system_path_filter
def remove_empty_directories(root):
    """Ascend up from the leaves accessible from `root` and remove empty
    directories.

    Parameters:
        root (str): path where to search for empty directories
    """
    for dirpath, subdirs, files in os.walk(root, topdown=False):
        for sd in subdirs:
            sdp = os.path.join(dirpath, sd)
            try:
                os.rmdir(sdp)
            except OSError:
                pass


@system_path_filter
def remove_dead_links(root):
    """Recursively removes any dead link that is present in root.

    Parameters:
        root (str): path where to search for dead links
    """
    for dirpath, subdirs, files in os.walk(root, topdown=False):
        for f in files:
            path = join_path(dirpath, f)
            remove_if_dead_link(path)


@system_path_filter
def remove_if_dead_link(path):
    """Removes the argument if it is a dead link.

    Parameters:
        path (str): The potential dead link
    """
    if islink(path) and not os.path.exists(path):
        os.unlink(path)


def readonly_file_handler(ignore_errors=False):
    # TODO: generate stages etc. with write permissions wherever
    # so this callback is no-longer required
    """
    Generate callback for shutil.rmtree to handle permissions errors on
    Windows. Some files may unexpectedly lack write permissions even
    though they were generated by Spack on behalf of the user (e.g. the
    stage), so this callback will detect such cases and modify the
    permissions if that is the issue. For other errors, the fallback
    is either to raise (if ignore_errors is False) or ignore (if
    ignore_errors is True). This is only intended for Windows systems
    and will raise a separate error if it is ever invoked (by accident)
    on a non-Windows system.
    """

    def error_remove_readonly(func, path, exc):
        if sys.platform != "win32":
            raise RuntimeError("This method should only be invoked on Windows")
        excvalue = exc[1]
        if (
            sys.platform == "win32"
            and func in (os.rmdir, os.remove, os.unlink)
            and excvalue.errno == errno.EACCES
        ):
            # change the file to be readable,writable,executable: 0777
            os.chmod(path, stat.S_IRWXU | stat.S_IRWXG | stat.S_IRWXO)
            # retry
            func(path)
        elif not ignore_errors:
            raise

    return error_remove_readonly


@system_path_filter
def remove_linked_tree(path):
    """Removes a directory and its contents.

    If the directory is a symlink, follows the link and removes the real
    directory before removing the link.

    This method will force-delete files on Windows

    Parameters:
        path (str): Directory to be removed
    """
    kwargs = {"ignore_errors": True}

    # Windows readonly files cannot be removed by Python
    # directly.
    if sys.platform == "win32":
        kwargs["ignore_errors"] = False
        kwargs["onerror"] = readonly_file_handler(ignore_errors=True)

    if os.path.exists(path):
        if islink(path):
            shutil.rmtree(os.path.realpath(path), **kwargs)
            os.unlink(path)
        else:
            shutil.rmtree(path, **kwargs)


@contextmanager
@system_path_filter
def safe_remove(*files_or_dirs):
    """Context manager to remove the files passed as input, but restore
    them in case any exception is raised in the context block.

    Args:
        *files_or_dirs: glob expressions for files or directories
            to be removed

    Returns:
        Dictionary that maps deleted files to their temporary copy
        within the context block.
    """
    # Find all the files or directories that match
    glob_matches = [glob.glob(x) for x in files_or_dirs]
    # Sort them so that shorter paths like "/foo/bar" come before
    # nested paths like "/foo/bar/baz.yaml". This simplifies the
    # handling of temporary copies below
    sorted_matches = sorted([os.path.abspath(x) for x in itertools.chain(*glob_matches)], key=len)

    # Copy files and directories in a temporary location
    removed, dst_root = {}, tempfile.mkdtemp()
    try:
        for id, file_or_dir in enumerate(sorted_matches):
            # The glob expression at the top ensures that the file/dir exists
            # at the time we enter the loop. Double check here since it might
            # happen that a previous iteration of the loop already removed it.
            # This is the case, for instance, if we remove the directory
            # "/foo/bar" before the file "/foo/bar/baz.yaml".
            if not os.path.exists(file_or_dir):
                continue
            # The monotonic ID is a simple way to make the filename
            # or directory name unique in the temporary folder
            basename = os.path.basename(file_or_dir) + "-{0}".format(id)
            temporary_path = os.path.join(dst_root, basename)
            shutil.move(file_or_dir, temporary_path)
            removed[file_or_dir] = temporary_path
        yield removed
    except BaseException:
        # Restore the files that were removed
        for original_path, temporary_path in removed.items():
            shutil.move(temporary_path, original_path)
        raise


@system_path_filter
def fix_darwin_install_name(path):
    """Fix install name of dynamic libraries on Darwin to have full path.

    There are two parts of this task:

    1. Use ``install_name('-id', ...)`` to change install name of a single lib
    2. Use ``install_name('-change', ...)`` to change the cross linking between
       libs. The function assumes that all libraries are in one folder and
       currently won't follow subfolders.

    Parameters:
        path (str): directory in which .dylib files are located
    """
    libs = glob.glob(join_path(path, "*.dylib"))
    for lib in libs:
        # fix install name first:
        install_name_tool = Executable("install_name_tool")
        install_name_tool("-id", lib, lib)
        otool = Executable("otool")
        long_deps = otool("-L", lib, output=str).split("\n")
        deps = [dep.partition(" ")[0][1::] for dep in long_deps[2:-1]]
        # fix all dependencies:
        for dep in deps:
            for loc in libs:
                # We really want to check for either
                #     dep == os.path.basename(loc)   or
                #     dep == join_path(builddir, os.path.basename(loc)),
                # but we don't know builddir (nor how symbolic links look
                # in builddir). We thus only compare the basenames.
                if os.path.basename(dep) == os.path.basename(loc):
                    install_name_tool("-change", dep, loc, lib)
                    break


def find_first(root: str, files: Union[Iterable[str], str], bfs_depth: int = 2) -> Optional[str]:
    """Find the first file matching a pattern.

    The following

    .. code-block:: console

       $ find /usr -name 'abc*' -o -name 'def*' -quit

    is equivalent to:

    >>> find_first("/usr", ["abc*", "def*"])

    Any glob pattern supported by fnmatch can be used.

    The search order of this method is breadth-first over directories,
    until depth bfs_depth, after which depth-first search is used.

    Parameters:
        root (str): The root directory to start searching from
        files (str or Iterable): File pattern(s) to search for
        bfs_depth (int): (advanced) parameter that specifies at which
            depth to switch to depth-first search.

    Returns:
        str or None: The matching file or None when no file is found.
    """
    if isinstance(files, str):
        files = [files]
    return FindFirstFile(root, *files, bfs_depth=bfs_depth).find()


def find(root, files, recursive=True):
    """Search for ``files`` starting from the ``root`` directory.

    Like GNU/BSD find but written entirely in Python.

    Examples:

    .. code-block:: console

       $ find /usr -name python

    is equivalent to:

    >>> find('/usr', 'python')

    .. code-block:: console

       $ find /usr/local/bin -maxdepth 1 -name python

    is equivalent to:

    >>> find('/usr/local/bin', 'python', recursive=False)

    Accepts any glob characters accepted by fnmatch:

    ==========  ====================================
    Pattern     Meaning
    ==========  ====================================
    ``*``       matches everything
    ``?``       matches any single character
    ``[seq]``   matches any character in ``seq``
    ``[!seq]``  matches any character not in ``seq``
    ==========  ====================================

    Parameters:
        root (str): The root directory to start searching from
        files (str or collections.abc.Sequence): Library name(s) to search for
        recursive (bool): if False search only root folder,
            if True descends top-down from the root. Defaults to True.

    Returns:
        list: The files that have been found
    """
    if isinstance(files, str):
        files = [files]

    if recursive:
        tty.debug(f"Find (recursive): {root} {str(files)}")
        result = _find_recursive(root, files)
    else:
        tty.debug(f"Find (not recursive): {root} {str(files)}")
        result = _find_non_recursive(root, files)

    tty.debug(f"Find complete: {root} {str(files)}")
    return result


@system_path_filter
def _find_recursive(root, search_files):
    # The variable here is **on purpose** a defaultdict. The idea is that
    # we want to poke the filesystem as little as possible, but still maintain
    # stability in the order of the answer. Thus we are recording each library
    # found in a key, and reconstructing the stable order later.
    found_files = collections.defaultdict(list)

    # Make the path absolute to have os.walk also return an absolute path
    root = os.path.abspath(root)
    for path, _, list_files in os.walk(root):
        for search_file in search_files:
            matches = glob.glob(os.path.join(path, search_file))
            matches = [os.path.join(path, x) for x in matches]
            found_files[search_file].extend(matches)

    answer = []
    for search_file in search_files:
        answer.extend(found_files[search_file])

    return answer


@system_path_filter
def _find_non_recursive(root, search_files):
    # The variable here is **on purpose** a defaultdict as os.list_dir
    # can return files in any order (does not preserve stability)
    found_files = collections.defaultdict(list)

    # Make the path absolute to have absolute path returned
    root = os.path.abspath(root)

    for search_file in search_files:
        matches = glob.glob(os.path.join(root, search_file))
        matches = [os.path.join(root, x) for x in matches]
        found_files[search_file].extend(matches)

    answer = []
    for search_file in search_files:
        answer.extend(found_files[search_file])

    return answer


# Utilities for libraries and headers


class FileList(collections.abc.Sequence):
    """Sequence of absolute paths to files.

    Provides a few convenience methods to manipulate file paths.
    """

    def __init__(self, files):
        if isinstance(files, str):
            files = [files]

        self.files = list(dedupe(files))

    @property
    def directories(self):
        """Stable de-duplication of the directories where the files reside.

        >>> l = LibraryList(['/dir1/liba.a', '/dir2/libb.a', '/dir1/libc.a'])
        >>> l.directories
        ['/dir1', '/dir2']
        >>> h = HeaderList(['/dir1/a.h', '/dir1/b.h', '/dir2/c.h'])
        >>> h.directories
        ['/dir1', '/dir2']

        Returns:
            list: A list of directories
        """
        return list(dedupe(os.path.dirname(x) for x in self.files if os.path.dirname(x)))

    @property
    def basenames(self):
        """Stable de-duplication of the base-names in the list

        >>> l = LibraryList(['/dir1/liba.a', '/dir2/libb.a', '/dir3/liba.a'])
        >>> l.basenames
        ['liba.a', 'libb.a']
        >>> h = HeaderList(['/dir1/a.h', '/dir2/b.h', '/dir3/a.h'])
        >>> h.basenames
        ['a.h', 'b.h']

        Returns:
            list: A list of base-names
        """
        return list(dedupe(os.path.basename(x) for x in self.files))

    def __getitem__(self, item):
        cls = type(self)
        if isinstance(item, numbers.Integral):
            return self.files[item]
        return cls(self.files[item])

    def __add__(self, other):
        return self.__class__(dedupe(self.files + list(other)))

    def __radd__(self, other):
        return self.__add__(other)

    def __eq__(self, other):
        return self.files == other.files

    def __len__(self):
        return len(self.files)

    def joined(self, separator=" "):
        return separator.join(self.files)

    def __repr__(self):
        return self.__class__.__name__ + "(" + repr(self.files) + ")"

    def __str__(self):
        return self.joined()


class HeaderList(FileList):
    """Sequence of absolute paths to headers.

    Provides a few convenience methods to manipulate header paths and get
    commonly used compiler flags or names.
    """

    # Make sure to only match complete words, otherwise path components such
    # as "xinclude" will cause false matches.
    # Avoid matching paths such as <prefix>/include/something/detail/include,
    # e.g. in the CUDA Toolkit which ships internal libc++ headers.
    include_regex = re.compile(r"(.*?)(\binclude\b)(.*)")

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

        self._macro_definitions = []
        self._directories = None

    @property
    def directories(self):
        """Directories to be searched for header files."""
        values = self._directories
        if values is None:
            values = self._default_directories()
        return list(dedupe(values))

    @directories.setter
    def directories(self, value):
        value = value or []
        # Accept a single directory as input
        if isinstance(value, str):
            value = [value]

        self._directories = [path_to_os_path(os.path.normpath(x))[0] for x in value]

    def _default_directories(self):
        """Default computation of directories based on the list of
        header files.
        """
        dir_list = super().directories
        values = []
        for d in dir_list:
            # If the path contains a subdirectory named 'include' then stop
            # there and don't add anything else to the path.
            m = self.include_regex.match(d)
            value = os.path.join(*m.group(1, 2)) if m else d
            values.append(value)
        return values

    @property
    def headers(self):
        """Stable de-duplication of the headers.

        Returns:
            list: A list of header files
        """
        return self.files

    @property
    def names(self):
        """Stable de-duplication of header names in the list without extensions

        >>> h = HeaderList(['/dir1/a.h', '/dir2/b.h', '/dir3/a.h'])
        >>> h.names
        ['a', 'b']

        Returns:
            list: A list of files without extensions
        """
        names = []

        for x in self.basenames:
            name = x

            # Valid extensions include: ['.cuh', '.hpp', '.hh', '.h']
            for ext in [".cuh", ".hpp", ".hh", ".h"]:
                i = name.rfind(ext)
                if i != -1:
                    names.append(name[:i])
                    break
            else:
                # No valid extension, should we still include it?
                names.append(name)

        return list(dedupe(names))

    @property
    def include_flags(self):
        """Include flags

        >>> h = HeaderList(['/dir1/a.h', '/dir1/b.h', '/dir2/c.h'])
        >>> h.include_flags
        '-I/dir1 -I/dir2'

        Returns:
            str: A joined list of include flags
        """
        return " ".join(["-I" + x for x in self.directories])

    @property
    def macro_definitions(self):
        """Macro definitions

        >>> h = HeaderList(['/dir1/a.h', '/dir1/b.h', '/dir2/c.h'])
        >>> h.add_macro('-DBOOST_LIB_NAME=boost_regex')
        >>> h.add_macro('-DBOOST_DYN_LINK')
        >>> h.macro_definitions
        '-DBOOST_LIB_NAME=boost_regex -DBOOST_DYN_LINK'

        Returns:
            str: A joined list of macro definitions
        """
        return " ".join(self._macro_definitions)

    @property
    def cpp_flags(self):
        """Include flags + macro definitions

        >>> h = HeaderList(['/dir1/a.h', '/dir1/b.h', '/dir2/c.h'])
        >>> h.cpp_flags
        '-I/dir1 -I/dir2'
        >>> h.add_macro('-DBOOST_DYN_LINK')
        >>> h.cpp_flags
        '-I/dir1 -I/dir2 -DBOOST_DYN_LINK'

        Returns:
            str: A joined list of include flags and macro definitions
        """
        cpp_flags = self.include_flags
        if self.macro_definitions:
            cpp_flags += " " + self.macro_definitions
        return cpp_flags

    def add_macro(self, macro):
        """Add a macro definition

        Parameters:
            macro (str): The macro to add
        """
        self._macro_definitions.append(macro)


def find_headers(headers, root, recursive=False):
    """Returns an iterable object containing a list of full paths to
    headers if found.

    Accepts any glob characters accepted by fnmatch:

    =======  ====================================
    Pattern  Meaning
    =======  ====================================
    *        matches everything
    ?        matches any single character
    [seq]    matches any character in ``seq``
    [!seq]   matches any character not in ``seq``
    =======  ====================================

    Parameters:
        headers (str or list): Header name(s) to search for
        root (str): The root directory to start searching from
        recursive (bool): if False search only root folder,
            if True descends top-down from the root. Defaults to False.

    Returns:
        HeaderList: The headers that have been found
    """
    if isinstance(headers, str):
        headers = [headers]
    elif not isinstance(headers, collections.abc.Sequence):
        message = "{0} expects a string or sequence of strings as the "
        message += "first argument [got {1} instead]"
        message = message.format(find_headers.__name__, type(headers))
        raise TypeError(message)

    # Construct the right suffix for the headers
    suffixes = [
        # C
        "h",
        # C++
        "hpp",
        "hxx",
        "hh",
        "H",
        "txx",
        "tcc",
        "icc",
        # Fortran
        "mod",
        "inc",
    ]

    # List of headers we are searching with suffixes
    headers = ["{0}.{1}".format(header, suffix) for header in headers for suffix in suffixes]

    return HeaderList(find(root, headers, recursive))


@system_path_filter
def find_all_headers(root):
    """Convenience function that returns the list of all headers found
    in the directory passed as argument.

    Args:
        root (str): directory where to look recursively for header files

    Returns:
        List of all headers found in ``root`` and subdirectories.
    """
    return find_headers("*", root=root, recursive=True)


class LibraryList(FileList):
    """Sequence of absolute paths to libraries

    Provides a few convenience methods to manipulate library paths and get
    commonly used compiler flags or names
    """

    @property
    def libraries(self):
        """Stable de-duplication of library files.

        Returns:
            list: A list of library files
        """
        return self.files

    @property
    def names(self):
        """Stable de-duplication of library names in the list

        >>> l = LibraryList(['/dir1/liba.a', '/dir2/libb.a', '/dir3/liba.so'])
        >>> l.names
        ['a', 'b']

        Returns:
            list: A list of library names
        """
        names = []

        for x in self.basenames:
            name = x
            if x.startswith("lib"):
                name = x[3:]

            # Valid extensions include: ['.dylib', '.so', '.a']
            # on non Windows platform
            # Windows valid library extensions are:
            # ['.dll', '.lib']
            valid_exts = [".dll", ".lib"] if sys.platform == "win32" else [".dylib", ".so", ".a"]
            for ext in valid_exts:
                i = name.rfind(ext)
                if i != -1:
                    names.append(name[:i])
                    break
            else:
                # No valid extension, should we still include it?
                names.append(name)

        return list(dedupe(names))

    @property
    def search_flags(self):
        """Search flags for the libraries

        >>> l = LibraryList(['/dir1/liba.a', '/dir2/libb.a', '/dir1/liba.so'])
        >>> l.search_flags
        '-L/dir1 -L/dir2'

        Returns:
            str: A joined list of search flags
        """
        return " ".join(["-L" + x for x in self.directories])

    @property
    def link_flags(self):
        """Link flags for the libraries

        >>> l = LibraryList(['/dir1/liba.a', '/dir2/libb.a', '/dir1/liba.so'])
        >>> l.link_flags
        '-la -lb'

        Returns:
            str: A joined list of link flags
        """
        return " ".join(["-l" + name for name in self.names])

    @property
    def ld_flags(self):
        """Search flags + link flags

        >>> l = LibraryList(['/dir1/liba.a', '/dir2/libb.a', '/dir1/liba.so'])
        >>> l.ld_flags
        '-L/dir1 -L/dir2 -la -lb'

        Returns:
            str: A joined list of search flags and link flags
        """
        return self.search_flags + " " + self.link_flags


def find_system_libraries(libraries, shared=True):
    """Searches the usual system library locations for ``libraries``.

    Search order is as follows:

    1. ``/lib64``
    2. ``/lib``
    3. ``/usr/lib64``
    4. ``/usr/lib``
    5. ``/usr/local/lib64``
    6. ``/usr/local/lib``

    Accepts any glob characters accepted by fnmatch:

    =======  ====================================
    Pattern  Meaning
    =======  ====================================
    *        matches everything
    ?        matches any single character
    [seq]    matches any character in ``seq``
    [!seq]   matches any character not in ``seq``
    =======  ====================================

    Parameters:
        libraries (str or list): Library name(s) to search for
        shared (bool): if True searches for shared libraries,
            otherwise for static. Defaults to True.

    Returns:
        LibraryList: The libraries that have been found
    """
    if isinstance(libraries, str):
        libraries = [libraries]
    elif not isinstance(libraries, collections.abc.Sequence):
        message = "{0} expects a string or sequence of strings as the "
        message += "first argument [got {1} instead]"
        message = message.format(find_system_libraries.__name__, type(libraries))
        raise TypeError(message)

    libraries_found = []
    search_locations = [
        "/lib64",
        "/lib",
        "/usr/lib64",
        "/usr/lib",
        "/usr/local/lib64",
        "/usr/local/lib",
    ]

    for library in libraries:
        for root in search_locations:
            result = find_libraries(library, root, shared, recursive=True)
            if result:
                libraries_found += result
                break

    return libraries_found


def find_libraries(libraries, root, shared=True, recursive=False, runtime=True):
    """Returns an iterable of full paths to libraries found in a root dir.

    Accepts any glob characters accepted by fnmatch:

    =======  ====================================
    Pattern  Meaning
    =======  ====================================
    *        matches everything
    ?        matches any single character
    [seq]    matches any character in ``seq``
    [!seq]   matches any character not in ``seq``
    =======  ====================================

    Parameters:
        libraries (str or list): Library name(s) to search for
        root (str): The root directory to start searching from
        shared (bool): if True searches for shared libraries,
            otherwise for static. Defaults to True.
        recursive (bool): if False search only root folder,
            if True descends top-down from the root. Defaults to False.
        runtime (bool): Windows only option, no-op elsewhere. If true,
            search for runtime shared libs (.DLL), otherwise, search
            for .Lib files. If shared is false, this has no meaning.
            Defaults to True.

    Returns:
        LibraryList: The libraries that have been found
    """
    if isinstance(libraries, str):
        libraries = [libraries]
    elif not isinstance(libraries, collections.abc.Sequence):
        message = "{0} expects a string or sequence of strings as the "
        message += "first argument [got {1} instead]"
        message = message.format(find_libraries.__name__, type(libraries))
        raise TypeError(message)

    if sys.platform == "win32":
        static_ext = "lib"
        # For linking (runtime=False) you need the .lib files regardless of
        # whether you are doing a shared or static link
        shared_ext = "dll" if runtime else "lib"
    else:
        # Used on both Linux and macOS
        static_ext = "a"
        shared_ext = "so"

    # Construct the right suffix for the library
    if shared:
        # Used on both Linux and macOS
        suffixes = [shared_ext]
        if sys.platform == "darwin":
            # Only used on macOS
            suffixes.append("dylib")
    else:
        suffixes = [static_ext]

    # List of libraries we are searching with suffixes
    libraries = ["{0}.{1}".format(lib, suffix) for lib in libraries for suffix in suffixes]

    if not recursive:
        # If not recursive, look for the libraries directly in root
        return LibraryList(find(root, libraries, False))

    # To speedup the search for external packages configured e.g. in /usr,
    # perform first non-recursive search in root/lib then in root/lib64 and
    # finally search all of root recursively. The search stops when the first
    # match is found.
    common_lib_dirs = ["lib", "lib64"]
    if sys.platform == "win32":
        common_lib_dirs.extend(["bin", "Lib"])

    for subdir in common_lib_dirs:
        dirname = join_path(root, subdir)
        if not os.path.isdir(dirname):
            continue
        found_libs = find(dirname, libraries, False)
        if found_libs:
            break
    else:
        found_libs = find(root, libraries, True)

    return LibraryList(found_libs)


def find_all_shared_libraries(root, recursive=False, runtime=True):
    """Convenience function that returns the list of all shared libraries found
    in the directory passed as argument.

    See documentation for `llnl.util.filesystem.find_libraries` for more information
    """
    return find_libraries("*", root=root, shared=True, recursive=recursive, runtime=runtime)


def find_all_static_libraries(root, recursive=False):
    """Convenience function that returns the list of all static libraries found
    in the directory passed as argument.

    See documentation for `llnl.util.filesystem.find_libraries` for more information
    """
    return find_libraries("*", root=root, shared=False, recursive=recursive)


def find_all_libraries(root, recursive=False):
    """Convenience function that returns the list of all libraries found
    in the directory passed as argument.

    See documentation for `llnl.util.filesystem.find_libraries` for more information
    """

    return find_all_shared_libraries(root, recursive=recursive) + find_all_static_libraries(
        root, recursive=recursive
    )


class WindowsSimulatedRPath:
    """Class representing Windows filesystem rpath analog

    One instance of this class is associated with a package (only on Windows)
    For each lib/binary directory in an associated package, this class introduces
    a symlink to any/all dependent libraries/binaries. This includes the packages
    own bin/lib directories, meaning the libraries are linked to the bianry directory
    and vis versa.
    """

    def __init__(self, package, link_install_prefix=True):
        """
        Args:
            package (spack.package_base.PackageBase): Package requiring links
            link_install_prefix (bool): Link against package's own install or stage root.
                Packages that run their own executables during build and require rpaths to
                the build directory during build time require this option. Default: install
                root
        """
        self.pkg = package
        self._addl_rpaths = set()
        self.link_install_prefix = link_install_prefix
        self._additional_library_dependents = set()

    @property
    def library_dependents(self):
        """
        Set of directories where package binaries/libraries are located.
        """
        return set([pathlib.Path(self.pkg.prefix.bin)]) | self._additional_library_dependents

    def add_library_dependent(self, *dest):
        """
        Add paths to directories or libraries/binaries to set of
        common paths that need to link against other libraries

        Specified paths should fall outside of a package's common
        link paths, i.e. the bin
        directories.
        """
        for pth in dest:
            if os.path.isfile(pth):
                self._additional_library_dependents.add(pathlib.Path(pth).parent)
            else:
                self._additional_library_dependents.add(pathlib.Path(pth))

    @property
    def rpaths(self):
        """
        Set of libraries this package needs to link against during runtime
        These packages will each be symlinked into the packages lib and binary dir
        """
        dependent_libs = []
        for path in self.pkg.rpath:
            dependent_libs.extend(list(find_all_shared_libraries(path, recursive=True)))
        for extra_path in self._addl_rpaths:
            dependent_libs.extend(list(find_all_shared_libraries(extra_path, recursive=True)))
        return set([pathlib.Path(x) for x in dependent_libs])

    def add_rpath(self, *paths):
        """
        Add libraries found at the root of provided paths to runtime linking

        These are libraries found outside of the typical scope of rpath linking
        that require manual inclusion in a runtime linking scheme.
        These links are unidirectional, and are only
        intended to bring outside dependencies into this package

        Args:
            *paths (str): arbitrary number of paths to be added to runtime linking
        """
        self._addl_rpaths = self._addl_rpaths | set(paths)

    def _link(self, path: pathlib.Path, dest_dir: pathlib.Path):
        """Perform link step of simulated rpathing, installing
        simlinks of file in path to the dest_dir
        location. This method deliberately prevents
        the case where a path points to a file inside the dest_dir.
        This is because it is both meaningless from an rpath
        perspective, and will cause an error when Developer
        mode is not enabled"""

        def report_already_linked():
            # We have either already symlinked or we are encoutering a naming clash
            # either way, we don't want to overwrite existing libraries
            already_linked = islink(str(dest_file))
            tty.debug(
                "Linking library %s to %s failed, " % (str(path), str(dest_file))
                + "already linked."
                if already_linked
                else "library with name %s already exists at location %s."
                % (str(file_name), str(dest_dir))
            )

        file_name = path.name
        dest_file = dest_dir / file_name
        if not dest_file.exists() and dest_dir.exists() and not dest_file == path:
            try:
                symlink(str(path), str(dest_file))
            # For py2 compatibility, we have to catch the specific Windows error code
            # associate with trying to create a file that already exists (winerror 183)
            # Catch OSErrors missed by the SymlinkError checks
            except OSError as e:
                if sys.platform == "win32" and (e.winerror == 183 or e.errno == errno.EEXIST):
                    report_already_linked()
                else:
                    raise e
            # catch errors we raise ourselves from Spack
            except llnl.util.symlink.AlreadyExistsError:
                report_already_linked()

    def establish_link(self):
        """
        (sym)link packages to runtime dependencies based on RPath configuration for
        Windows heuristics
        """
        # from build_environment.py:463
        # The top-level package is always RPATHed. It hasn't been installed yet
        # so the RPATHs are added unconditionally

        # for each binary install dir in self.pkg (i.e. pkg.prefix.bin, pkg.prefix.lib)
        # install a symlink to each dependent library
        for library, lib_dir in itertools.product(self.rpaths, self.library_dependents):
            self._link(library, lib_dir)


@system_path_filter
@memoized
def can_access_dir(path):
    """Returns True if the argument is an accessible directory.

    Args:
        path: path to be tested

    Returns:
        True if ``path`` is an accessible directory, else False
    """
    return os.path.isdir(path) and os.access(path, os.R_OK | os.X_OK)


@system_path_filter
@memoized
def can_write_to_dir(path):
    """Return True if the argument is a directory in which we can write.

    Args:
        path: path to be tested

    Returns:
        True if ``path`` is an writeable directory, else False
    """
    return os.path.isdir(path) and os.access(path, os.R_OK | os.X_OK | os.W_OK)


@system_path_filter
@memoized
def files_in(*search_paths):
    """Returns all the files in paths passed as arguments.

    Caller must ensure that each path in ``search_paths`` is a directory.

    Args:
        *search_paths: directories to be searched

    Returns:
        List of (file, full_path) tuples with all the files found.
    """
    files = []
    for d in filter(can_access_dir, search_paths):
        files.extend(
            filter(
                lambda x: os.path.isfile(x[1]), [(f, os.path.join(d, f)) for f in os.listdir(d)]
            )
        )
    return files


def is_readable_file(file_path):
    """Return True if the path passed as argument is readable"""
    return os.path.isfile(file_path) and os.access(file_path, os.R_OK)


@system_path_filter
def search_paths_for_executables(*path_hints):
    """Given a list of path hints returns a list of paths where
    to search for an executable.

    Args:
        *path_hints (list of paths): list of paths taken into
            consideration for a search

    Returns:
        A list containing the real path of every existing directory
        in `path_hints` and its `bin` subdirectory if it exists.
    """
    executable_paths = []
    for path in path_hints:
        if not os.path.isdir(path):
            continue

        path = os.path.abspath(path)
        executable_paths.append(path)

        bin_dir = os.path.join(path, "bin")
        if os.path.isdir(bin_dir):
            executable_paths.append(bin_dir)

    return executable_paths


@system_path_filter
def search_paths_for_libraries(*path_hints):
    """Given a list of path hints returns a list of paths where
    to search for a shared library.

    Args:
        *path_hints (list of paths): list of paths taken into
            consideration for a search

    Returns:
        A list containing the real path of every existing directory
        in `path_hints` and its `lib` and `lib64` subdirectory if it exists.
    """
    library_paths = []
    for path in path_hints:
        if not os.path.isdir(path):
            continue

        path = os.path.abspath(path)
        library_paths.append(path)

        lib_dir = os.path.join(path, "lib")
        if os.path.isdir(lib_dir):
            library_paths.append(lib_dir)

        lib64_dir = os.path.join(path, "lib64")
        if os.path.isdir(lib64_dir):
            library_paths.append(lib64_dir)

    return library_paths


@system_path_filter
def partition_path(path, entry=None):
    """
    Split the prefixes of the path at the first occurrence of entry and
    return a 3-tuple containing a list of the prefixes before the entry, a
    string of the prefix ending with the entry, and a list of the prefixes
    after the entry.

    If the entry is not a node in the path, the result will be the prefix list
    followed by an empty string and an empty list.
    """
    paths = prefixes(path)

    if entry is not None:
        # Derive the index of entry within paths, which will correspond to
        # the location of the entry in within the path.
        try:
            sep = os.sep
            entries = path.split(sep)
            if entries[0].endswith(":"):
                # Handle drive letters e.g. C:/ on Windows
                entries[0] = entries[0] + sep
            i = entries.index(entry)
            if "" in entries:
                i -= 1
            return paths[:i], paths[i], paths[i + 1 :]
        except ValueError:
            pass

    return paths, "", []


@system_path_filter
def prefixes(path):
    """
    Returns a list containing the path and its ancestors, top-to-bottom.

    The list for an absolute path will not include an ``os.sep`` entry.
    For example, assuming ``os.sep`` is ``/``, given path ``/ab/cd/efg``
    the resulting paths will be, in order: ``/ab``, ``/ab/cd``, and
    ``/ab/cd/efg``

    The list for a relative path starting ``./`` will not include ``.``.
    For example, path ``./hi/jkl/mn`` results in a list with the following
    paths, in order: ``./hi``, ``./hi/jkl``, and ``./hi/jkl/mn``.

    On Windows, paths will be normalized to use ``/`` and ``/`` will always
    be used as the separator instead of ``os.sep``.

    Parameters:
        path (str): the string used to derive ancestor paths

    Returns:
        A list containing ancestor paths in order and ending with the path
    """
    if not path:
        return []
    sep = os.sep
    parts = path.strip(sep).split(sep)
    if path.startswith(sep):
        parts.insert(0, sep)
    elif parts[0].endswith(":"):
        # Handle drive letters e.g. C:/ on Windows
        parts[0] = parts[0] + sep
    paths = [os.path.join(*parts[: i + 1]) for i in range(len(parts))]

    try:
        paths.remove(sep)
    except ValueError:
        pass

    try:
        paths.remove(".")
    except ValueError:
        pass

    return paths


@system_path_filter
def md5sum(file):
    """Compute the MD5 sum of a file.

    Args:
        file (str): file to be checksummed

    Returns:
        MD5 sum of the file's content
    """
    md5 = hashlib.md5()
    with open(file, "rb") as f:
        md5.update(f.read())
    return md5.digest()


@system_path_filter
def remove_directory_contents(dir):
    """Remove all contents of a directory."""
    if os.path.exists(dir):
        for entry in [os.path.join(dir, entry) for entry in os.listdir(dir)]:
            if os.path.isfile(entry) or islink(entry):
                os.unlink(entry)
            else:
                shutil.rmtree(entry)


@contextmanager
@system_path_filter
def keep_modification_time(*filenames):
    """
    Context manager to keep the modification timestamps of the input files.
    Tolerates and has no effect on non-existent files and files that are
    deleted by the nested code.

    Parameters:
        *filenames: one or more files that must have their modification
            timestamps unchanged
    """
    mtimes = {}
    for f in filenames:
        if os.path.exists(f):
            mtimes[f] = os.path.getmtime(f)
    yield
    for f, mtime in mtimes.items():
        if os.path.exists(f):
            os.utime(f, (os.path.getatime(f), mtime))


@contextmanager
def temporary_dir(
    suffix: Optional[str] = None, prefix: Optional[str] = None, dir: Optional[str] = None
):
    """Create a temporary directory and cd's into it. Delete the directory
    on exit.

    Takes the same arguments as tempfile.mkdtemp()
    """
    tmp_dir = tempfile.mkdtemp(suffix=suffix, prefix=prefix, dir=dir)
    try:
        with working_dir(tmp_dir):
            yield tmp_dir
    finally:
        remove_directory_contents(tmp_dir)


def filesummary(path, print_bytes=16) -> Tuple[int, bytes]:
    """Create a small summary of the given file. Does not error
    when file does not exist.

    Args:
        print_bytes (int): Number of bytes to print from start/end of file

    Returns:
        Tuple of size and byte string containing first n .. last n bytes.
        Size is 0 if file cannot be read."""
    try:
        n = print_bytes
        with open(path, "rb") as f:
            size = os.fstat(f.fileno()).st_size
            if size <= 2 * n:
                short_contents = f.read(2 * n)
            else:
                short_contents = f.read(n)
                f.seek(-n, 2)
                short_contents += b"..." + f.read(n)
        return size, short_contents
    except OSError:
        return 0, b""


class FindFirstFile:
    """Uses hybrid iterative deepening to locate the first matching
    file. Up to depth ``bfs_depth`` it uses iterative deepening, which
    mimics breadth-first with the same memory footprint as depth-first
    search, after which it switches to ordinary depth-first search using
    ``os.walk``."""

    def __init__(self, root: str, *file_patterns: str, bfs_depth: int = 2):
        """Create a small summary of the given file. Does not error
        when file does not exist.

        Args:
            root (str): directory in which to recursively search
            file_patterns (str): glob file patterns understood by fnmatch
            bfs_depth (int): until this depth breadth-first traversal is used,
                when no match is found, the mode is switched to depth-first search.
        """
        self.root = root
        self.bfs_depth = bfs_depth
        self.match: Callable

        # normcase is trivial on posix
        regex = re.compile("|".join(fnmatch.translate(os.path.normcase(p)) for p in file_patterns))

        # On case sensitive filesystems match against normcase'd paths.
        if os.path is posixpath:
            self.match = regex.match
        else:
            self.match = lambda p: regex.match(os.path.normcase(p))

    def find(self) -> Optional[str]:
        """Run the file search

        Returns:
            str or None: path of the matching file
        """
        self.file = None

        # First do iterative deepening (i.e. bfs through limited depth dfs)
        for i in range(self.bfs_depth + 1):
            if self._find_at_depth(self.root, i):
                return self.file

        # Then fall back to depth-first search
        return self._find_dfs()

    def _find_at_depth(self, path, max_depth, depth=0) -> bool:
        """Returns True when done. Notice search can be done
        either because a file was found, or because it recursed
        through all directories."""
        try:
            entries = os.scandir(path)
        except OSError:
            return True

        done = True

        with entries:
            # At max depth we look for matching files.
            if depth == max_depth:
                for f in entries:
                    # Exit on match
                    if self.match(f.name):
                        self.file = os.path.join(path, f.name)
                        return True

                    # is_dir should not require a stat call, so it's a good optimization.
                    if self._is_dir(f):
                        done = False
                return done

            # At lower depth only recurse into subdirs
            for f in entries:
                if not self._is_dir(f):
                    continue

                # If any subdir is not fully traversed, we're not done yet.
                if not self._find_at_depth(os.path.join(path, f.name), max_depth, depth + 1):
                    done = False

                # Early exit when we've found something.
                if self.file:
                    return True

            return done

    def _is_dir(self, f: os.DirEntry) -> bool:
        """Returns True when f is dir we can enter (and not a symlink)."""
        try:
            return f.is_dir(follow_symlinks=False)
        except OSError:
            return False

    def _find_dfs(self) -> Optional[str]:
        """Returns match or None"""
        for dirpath, _, filenames in os.walk(self.root):
            for file in filenames:
                if self.match(file):
                    return os.path.join(dirpath, file)
        return None