The Meson Build System
http://mesonbuild.com/
You can not select more than 25 topics
Topics must start with a letter or number, can include dashes ('-') and can be up to 35 characters long.
950 lines
33 KiB
950 lines
33 KiB
# Copyright 2012-2015 The Meson development team |
|
|
|
# Licensed under the Apache License, Version 2.0 (the "License"); |
|
# you may not use this file except in compliance with the License. |
|
# You may obtain a copy of the License at |
|
|
|
# http://www.apache.org/licenses/LICENSE-2.0 |
|
|
|
# Unless required by applicable law or agreed to in writing, software |
|
# distributed under the License is distributed on an "AS IS" BASIS, |
|
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. |
|
# See the License for the specific language governing permissions and |
|
# limitations under the License. |
|
|
|
"""A library of random helper functionality.""" |
|
|
|
import sys |
|
import stat |
|
import time |
|
import platform, subprocess, operator, os, shutil, re |
|
import collections |
|
from mesonbuild import mlog |
|
|
|
from glob import glob |
|
|
|
def detect_meson_py_location(): |
|
c = sys.argv[0] |
|
c_dir, c_fname = os.path.split(c) |
|
|
|
# get the absolute path to the <mesontool> folder |
|
m_dir = None |
|
if os.path.isabs(c): |
|
# $ /foo/<mesontool>.py <args> |
|
m_dir = c_dir |
|
elif c_dir == '': |
|
# $ <mesontool> <args> (gets run from /usr/bin/<mesontool>) |
|
in_path_exe = shutil.which(c_fname) |
|
if in_path_exe: |
|
m_dir, c_fname = os.path.split(in_path_exe) |
|
# Special case: when run like "./meson.py <opts>", |
|
# we need to expand it out, because, for example, |
|
# "ninja test" will be run from a different directory. |
|
if m_dir == '.': |
|
m_dir = os.getcwd() |
|
else: |
|
m_dir = os.path.abspath(c_dir) |
|
|
|
# find meson in m_dir |
|
if m_dir is not None: |
|
for fname in ['meson', 'meson.py']: |
|
m_path = os.path.join(m_dir, fname) |
|
if os.path.exists(m_path): |
|
return m_path |
|
|
|
# No meson found, which means that either: |
|
# a) meson is not installed |
|
# b) meson is installed to a non-standard location |
|
# c) the script that invoked mesonlib is not the one of meson tools (e.g. run_unittests.py) |
|
# The only thing remaining is to try to find the bundled executable and |
|
# pray distro packagers have not moved it. |
|
fname = os.path.normpath(os.path.join(os.path.dirname(__file__), '..', 'meson.py')) |
|
if not os.path.exists(fname): |
|
raise RuntimeError('Could not determine how to run Meson. Please file a bug with details.') |
|
return fname |
|
|
|
if os.path.basename(sys.executable) == 'meson.exe': |
|
# In Windows and using the MSI installed executable. |
|
meson_command = [sys.executable] |
|
python_command = [sys.executable, 'runpython'] |
|
else: |
|
python_command = [sys.executable] |
|
meson_command = python_command + [detect_meson_py_location()] |
|
|
|
def is_ascii_string(astring): |
|
try: |
|
if isinstance(astring, str): |
|
astring.encode('ascii') |
|
if isinstance(astring, bytes): |
|
astring.decode('ascii') |
|
except UnicodeDecodeError: |
|
return False |
|
return True |
|
|
|
def check_direntry_issues(direntry_array): |
|
import locale |
|
# Warn if the locale is not UTF-8. This can cause various unfixable issues |
|
# such as os.stat not being able to decode filenames with unicode in them. |
|
# There is no way to reset both the preferred encoding and the filesystem |
|
# encoding, so we can just warn about it. |
|
e = locale.getpreferredencoding() |
|
if e.upper() != 'UTF-8' and not is_windows(): |
|
if not isinstance(direntry_array, list): |
|
direntry_array = [direntry_array] |
|
for de in direntry_array: |
|
if is_ascii_string(de): |
|
continue |
|
mlog.warning('''You are using {!r} which is not a Unicode-compatible ' |
|
locale but you are trying to access a file system entry called {!r} which is |
|
not pure ASCII. This may cause problems. |
|
'''.format(e, de), file=sys.stderr) |
|
|
|
# Put this in objects that should not get dumped to pickle files |
|
# by accident. |
|
import threading |
|
an_unpicklable_object = threading.Lock() |
|
|
|
class MesonException(Exception): |
|
'''Exceptions thrown by Meson''' |
|
|
|
class EnvironmentException(MesonException): |
|
'''Exceptions thrown while processing and creating the build environment''' |
|
|
|
class FileMode: |
|
# The first triad is for owner permissions, the second for group permissions, |
|
# and the third for others (everyone else). |
|
# For the 1st character: |
|
# 'r' means can read |
|
# '-' means not allowed |
|
# For the 2nd character: |
|
# 'w' means can write |
|
# '-' means not allowed |
|
# For the 3rd character: |
|
# 'x' means can execute |
|
# 's' means can execute and setuid/setgid is set (owner/group triads only) |
|
# 'S' means cannot execute and setuid/setgid is set (owner/group triads only) |
|
# 't' means can execute and sticky bit is set ("others" triads only) |
|
# 'T' means cannot execute and sticky bit is set ("others" triads only) |
|
# '-' means none of these are allowed |
|
# |
|
# The meanings of 'rwx' perms is not obvious for directories; see: |
|
# https://www.hackinglinuxexposed.com/articles/20030424.html |
|
# |
|
# For information on this notation such as setuid/setgid/sticky bits, see: |
|
# https://en.wikipedia.org/wiki/File_system_permissions#Symbolic_notation |
|
symbolic_perms_regex = re.compile('[r-][w-][xsS-]' # Owner perms |
|
'[r-][w-][xsS-]' # Group perms |
|
'[r-][w-][xtT-]') # Others perms |
|
|
|
def __init__(self, perms=None, owner=None, group=None): |
|
self.perms_s = perms |
|
self.perms = self.perms_s_to_bits(perms) |
|
self.owner = owner |
|
self.group = group |
|
|
|
def __repr__(self): |
|
ret = '<FileMode: {!r} owner={} group={}' |
|
return ret.format(self.perms_s, self.owner, self.group) |
|
|
|
@classmethod |
|
def perms_s_to_bits(cls, perms_s): |
|
''' |
|
Does the opposite of stat.filemode(), converts strings of the form |
|
'rwxr-xr-x' to st_mode enums which can be passed to os.chmod() |
|
''' |
|
if perms_s is None: |
|
# No perms specified, we will not touch the permissions |
|
return -1 |
|
eg = 'rwxr-xr-x' |
|
if not isinstance(perms_s, str): |
|
msg = 'Install perms must be a string. For example, {!r}' |
|
raise MesonException(msg.format(eg)) |
|
if len(perms_s) != 9 or not cls.symbolic_perms_regex.match(perms_s): |
|
msg = 'File perms {!r} must be exactly 9 chars. For example, {!r}' |
|
raise MesonException(msg.format(perms_s, eg)) |
|
perms = 0 |
|
# Owner perms |
|
if perms_s[0] == 'r': |
|
perms |= stat.S_IRUSR |
|
if perms_s[1] == 'w': |
|
perms |= stat.S_IWUSR |
|
if perms_s[2] == 'x': |
|
perms |= stat.S_IXUSR |
|
elif perms_s[2] == 'S': |
|
perms |= stat.S_ISUID |
|
elif perms_s[2] == 's': |
|
perms |= stat.S_IXUSR |
|
perms |= stat.S_ISUID |
|
# Group perms |
|
if perms_s[3] == 'r': |
|
perms |= stat.S_IRGRP |
|
if perms_s[4] == 'w': |
|
perms |= stat.S_IWGRP |
|
if perms_s[5] == 'x': |
|
perms |= stat.S_IXGRP |
|
elif perms_s[5] == 'S': |
|
perms |= stat.S_ISGID |
|
elif perms_s[5] == 's': |
|
perms |= stat.S_IXGRP |
|
perms |= stat.S_ISGID |
|
# Others perms |
|
if perms_s[6] == 'r': |
|
perms |= stat.S_IROTH |
|
if perms_s[7] == 'w': |
|
perms |= stat.S_IWOTH |
|
if perms_s[8] == 'x': |
|
perms |= stat.S_IXOTH |
|
elif perms_s[8] == 'T': |
|
perms |= stat.S_ISVTX |
|
elif perms_s[8] == 't': |
|
perms |= stat.S_IXOTH |
|
perms |= stat.S_ISVTX |
|
return perms |
|
|
|
class File: |
|
def __init__(self, is_built, subdir, fname): |
|
self.is_built = is_built |
|
self.subdir = subdir |
|
self.fname = fname |
|
assert(isinstance(self.subdir, str)) |
|
assert(isinstance(self.fname, str)) |
|
|
|
def __str__(self): |
|
return self.relative_name() |
|
|
|
def __repr__(self): |
|
ret = '<File: {0}' |
|
if not self.is_built: |
|
ret += ' (not built)' |
|
ret += '>' |
|
return ret.format(self.relative_name()) |
|
|
|
@staticmethod |
|
def from_source_file(source_root, subdir, fname): |
|
if not os.path.isfile(os.path.join(source_root, subdir, fname)): |
|
raise MesonException('File %s does not exist.' % fname) |
|
return File(False, subdir, fname) |
|
|
|
@staticmethod |
|
def from_built_file(subdir, fname): |
|
return File(True, subdir, fname) |
|
|
|
@staticmethod |
|
def from_absolute_file(fname): |
|
return File(False, '', fname) |
|
|
|
def rel_to_builddir(self, build_to_src): |
|
if self.is_built: |
|
return self.relative_name() |
|
else: |
|
return os.path.join(build_to_src, self.subdir, self.fname) |
|
|
|
def absolute_path(self, srcdir, builddir): |
|
absdir = srcdir |
|
if self.is_built: |
|
absdir = builddir |
|
return os.path.join(absdir, self.relative_name()) |
|
|
|
def endswith(self, ending): |
|
return self.fname.endswith(ending) |
|
|
|
def split(self, s): |
|
return self.fname.split(s) |
|
|
|
def __eq__(self, other): |
|
return (self.fname, self.subdir, self.is_built) == (other.fname, other.subdir, other.is_built) |
|
|
|
def __hash__(self): |
|
return hash((self.fname, self.subdir, self.is_built)) |
|
|
|
def relative_name(self): |
|
return os.path.join(self.subdir, self.fname) |
|
|
|
def get_compiler_for_source(compilers, src): |
|
for comp in compilers: |
|
if comp.can_compile(src): |
|
return comp |
|
raise RuntimeError('No specified compiler can handle file {!s}'.format(src)) |
|
|
|
def classify_unity_sources(compilers, sources): |
|
compsrclist = {} |
|
for src in sources: |
|
comp = get_compiler_for_source(compilers, src) |
|
if comp not in compsrclist: |
|
compsrclist[comp] = [src] |
|
else: |
|
compsrclist[comp].append(src) |
|
return compsrclist |
|
|
|
def is_osx(): |
|
return platform.system().lower() == 'darwin' |
|
|
|
def is_linux(): |
|
return platform.system().lower() == 'linux' |
|
|
|
def is_haiku(): |
|
return platform.system().lower() == 'haiku' |
|
|
|
def is_windows(): |
|
platname = platform.system().lower() |
|
return platname == 'windows' or 'mingw' in platname |
|
|
|
def is_cygwin(): |
|
platname = platform.system().lower() |
|
return platname.startswith('cygwin') |
|
|
|
def is_debianlike(): |
|
return os.path.isfile('/etc/debian_version') |
|
|
|
def is_dragonflybsd(): |
|
return platform.system().lower() == 'dragonfly' |
|
|
|
def is_freebsd(): |
|
return platform.system().lower() == 'freebsd' |
|
|
|
def for_windows(is_cross, env): |
|
""" |
|
Host machine is windows? |
|
|
|
Note: 'host' is the machine on which compiled binaries will run |
|
""" |
|
if not is_cross: |
|
return is_windows() |
|
elif env.cross_info.has_host(): |
|
return env.cross_info.config['host_machine']['system'] == 'windows' |
|
return False |
|
|
|
def for_cygwin(is_cross, env): |
|
""" |
|
Host machine is cygwin? |
|
|
|
Note: 'host' is the machine on which compiled binaries will run |
|
""" |
|
if not is_cross: |
|
return is_cygwin() |
|
elif env.cross_info.has_host(): |
|
return env.cross_info.config['host_machine']['system'] == 'cygwin' |
|
return False |
|
|
|
def for_linux(is_cross, env): |
|
""" |
|
Host machine is linux? |
|
|
|
Note: 'host' is the machine on which compiled binaries will run |
|
""" |
|
if not is_cross: |
|
return is_linux() |
|
elif env.cross_info.has_host(): |
|
return env.cross_info.config['host_machine']['system'] == 'linux' |
|
return False |
|
|
|
def for_darwin(is_cross, env): |
|
""" |
|
Host machine is Darwin (iOS/OS X)? |
|
|
|
Note: 'host' is the machine on which compiled binaries will run |
|
""" |
|
if not is_cross: |
|
return is_osx() |
|
elif env.cross_info.has_host(): |
|
return env.cross_info.config['host_machine']['system'] == 'darwin' |
|
return False |
|
|
|
def for_haiku(is_cross, env): |
|
""" |
|
Host machine is Haiku? |
|
|
|
Note: 'host' is the machine on which compiled binaries will run |
|
""" |
|
if not is_cross: |
|
return is_haiku() |
|
elif env.cross_info.has_host(): |
|
return env.cross_info.config['host_machine']['system'] == 'haiku' |
|
return False |
|
|
|
def exe_exists(arglist): |
|
try: |
|
p = subprocess.Popen(arglist, stdout=subprocess.PIPE, stderr=subprocess.PIPE) |
|
p.communicate() |
|
if p.returncode == 0: |
|
return True |
|
except FileNotFoundError: |
|
pass |
|
return False |
|
|
|
def detect_vcs(source_dir): |
|
vcs_systems = [ |
|
dict(name = 'git', cmd = 'git', repo_dir = '.git', get_rev = 'git describe --dirty=+', rev_regex = '(.*)', dep = '.git/logs/HEAD'), |
|
dict(name = 'mercurial', cmd = 'hg', repo_dir = '.hg', get_rev = 'hg id -i', rev_regex = '(.*)', dep = '.hg/dirstate'), |
|
dict(name = 'subversion', cmd = 'svn', repo_dir = '.svn', get_rev = 'svn info', rev_regex = 'Revision: (.*)', dep = '.svn/wc.db'), |
|
dict(name = 'bazaar', cmd = 'bzr', repo_dir = '.bzr', get_rev = 'bzr revno', rev_regex = '(.*)', dep = '.bzr'), |
|
] |
|
|
|
segs = source_dir.replace('\\', '/').split('/') |
|
for i in range(len(segs), -1, -1): |
|
curdir = '/'.join(segs[:i]) |
|
for vcs in vcs_systems: |
|
if os.path.isdir(os.path.join(curdir, vcs['repo_dir'])) and shutil.which(vcs['cmd']): |
|
vcs['wc_dir'] = curdir |
|
return vcs |
|
return None |
|
|
|
def grab_leading_numbers(vstr, strict=False): |
|
result = [] |
|
for x in vstr.rstrip('.').split('.'): |
|
try: |
|
result.append(int(x)) |
|
except ValueError as e: |
|
if strict: |
|
msg = 'Invalid version to compare against: {!r}; only ' \ |
|
'numeric digits separated by "." are allowed: ' + str(e) |
|
raise MesonException(msg.format(vstr)) |
|
break |
|
return result |
|
|
|
numpart = re.compile('[0-9.]+') |
|
|
|
def version_compare(vstr1, vstr2, strict=False): |
|
match = numpart.match(vstr1.strip()) |
|
if match is None: |
|
msg = 'Uncomparable version string {!r}.' |
|
raise MesonException(msg.format(vstr1)) |
|
vstr1 = match.group(0) |
|
if vstr2.startswith('>='): |
|
cmpop = operator.ge |
|
vstr2 = vstr2[2:] |
|
elif vstr2.startswith('<='): |
|
cmpop = operator.le |
|
vstr2 = vstr2[2:] |
|
elif vstr2.startswith('!='): |
|
cmpop = operator.ne |
|
vstr2 = vstr2[2:] |
|
elif vstr2.startswith('=='): |
|
cmpop = operator.eq |
|
vstr2 = vstr2[2:] |
|
elif vstr2.startswith('='): |
|
cmpop = operator.eq |
|
vstr2 = vstr2[1:] |
|
elif vstr2.startswith('>'): |
|
cmpop = operator.gt |
|
vstr2 = vstr2[1:] |
|
elif vstr2.startswith('<'): |
|
cmpop = operator.lt |
|
vstr2 = vstr2[1:] |
|
else: |
|
cmpop = operator.eq |
|
varr1 = grab_leading_numbers(vstr1, strict) |
|
varr2 = grab_leading_numbers(vstr2, strict) |
|
return cmpop(varr1, varr2) |
|
|
|
def version_compare_many(vstr1, conditions): |
|
if not isinstance(conditions, (list, tuple, frozenset)): |
|
conditions = [conditions] |
|
found = [] |
|
not_found = [] |
|
for req in conditions: |
|
if not version_compare(vstr1, req, strict=True): |
|
not_found.append(req) |
|
else: |
|
found.append(req) |
|
return not_found == [], not_found, found |
|
|
|
def default_libdir(): |
|
if is_debianlike(): |
|
try: |
|
pc = subprocess.Popen(['dpkg-architecture', '-qDEB_HOST_MULTIARCH'], |
|
stdout=subprocess.PIPE, |
|
stderr=subprocess.DEVNULL) |
|
(stdo, _) = pc.communicate() |
|
if pc.returncode == 0: |
|
archpath = stdo.decode().strip() |
|
return 'lib/' + archpath |
|
except Exception: |
|
pass |
|
if os.path.isdir('/usr/lib64') and not os.path.islink('/usr/lib64'): |
|
return 'lib64' |
|
return 'lib' |
|
|
|
def default_libexecdir(): |
|
# There is no way to auto-detect this, so it must be set at build time |
|
return 'libexec' |
|
|
|
def default_prefix(): |
|
return 'c:/' if is_windows() else '/usr/local' |
|
|
|
def get_library_dirs(): |
|
if is_windows(): |
|
return ['C:/mingw/lib'] # Fixme |
|
if is_osx(): |
|
return ['/usr/lib'] # Fix me as well. |
|
# The following is probably Debian/Ubuntu specific. |
|
# /usr/local/lib is first because it contains stuff |
|
# installed by the sysadmin and is probably more up-to-date |
|
# than /usr/lib. If you feel that this search order is |
|
# problematic, please raise the issue on the mailing list. |
|
unixdirs = ['/usr/local/lib', '/usr/lib', '/lib'] |
|
plat = subprocess.check_output(['uname', '-m']).decode().strip() |
|
# This is a terrible hack. I admit it and I'm really sorry. |
|
# I just don't know what the correct solution is. |
|
if plat == 'i686': |
|
plat = 'i386' |
|
if plat.startswith('arm'): |
|
plat = 'arm' |
|
unixdirs += glob('/usr/lib/' + plat + '*') |
|
if os.path.exists('/usr/lib64'): |
|
unixdirs.append('/usr/lib64') |
|
unixdirs += glob('/lib/' + plat + '*') |
|
if os.path.exists('/lib64'): |
|
unixdirs.append('/lib64') |
|
unixdirs += glob('/lib/' + plat + '*') |
|
return unixdirs |
|
|
|
|
|
def do_replacement(regex, line, confdata): |
|
missing_variables = set() |
|
|
|
def variable_replace(match): |
|
# Pairs of escape characters before '@' or '\@' |
|
if match.group(0).endswith('\\'): |
|
num_escapes = match.end(0) - match.start(0) |
|
return '\\' * (num_escapes // 2) |
|
# Single escape character and '@' |
|
elif match.group(0) == '\\@': |
|
return '@' |
|
# Template variable to be replaced |
|
else: |
|
varname = match.group(1) |
|
if varname in confdata: |
|
(var, desc) = confdata.get(varname) |
|
if isinstance(var, str): |
|
pass |
|
elif isinstance(var, int): |
|
var = str(var) |
|
else: |
|
raise RuntimeError('Tried to replace a variable with something other than a string or int.') |
|
else: |
|
missing_variables.add(varname) |
|
var = '' |
|
return var |
|
return re.sub(regex, variable_replace, line), missing_variables |
|
|
|
def do_mesondefine(line, confdata): |
|
arr = line.split() |
|
if len(arr) != 2: |
|
raise MesonException('#mesondefine does not contain exactly two tokens: %s', line.strip()) |
|
varname = arr[1] |
|
try: |
|
(v, desc) = confdata.get(varname) |
|
except KeyError: |
|
return '/* #undef %s */\n' % varname |
|
if isinstance(v, bool): |
|
if v: |
|
return '#define %s\n' % varname |
|
else: |
|
return '#undef %s\n' % varname |
|
elif isinstance(v, int): |
|
return '#define %s %d\n' % (varname, v) |
|
elif isinstance(v, str): |
|
return '#define %s %s\n' % (varname, v) |
|
else: |
|
raise MesonException('#mesondefine argument "%s" is of unknown type.' % varname) |
|
|
|
|
|
def do_conf_file(src, dst, confdata): |
|
try: |
|
with open(src, encoding='utf-8') as f: |
|
data = f.readlines() |
|
except Exception as e: |
|
raise MesonException('Could not read input file %s: %s' % (src, str(e))) |
|
# Only allow (a-z, A-Z, 0-9, _, -) as valid characters for a define |
|
# Also allow escaping '@' with '\@' |
|
regex = re.compile(r'(?:\\\\)+(?=\\?@)|\\@|@([-a-zA-Z0-9_]+)@') |
|
result = [] |
|
missing_variables = set() |
|
for line in data: |
|
if line.startswith('#mesondefine'): |
|
line = do_mesondefine(line, confdata) |
|
else: |
|
line, missing = do_replacement(regex, line, confdata) |
|
missing_variables.update(missing) |
|
result.append(line) |
|
dst_tmp = dst + '~' |
|
with open(dst_tmp, 'w', encoding='utf-8') as f: |
|
f.writelines(result) |
|
shutil.copymode(src, dst_tmp) |
|
replace_if_different(dst, dst_tmp) |
|
return missing_variables |
|
|
|
def dump_conf_header(ofilename, cdata): |
|
with open(ofilename, 'w', encoding='utf-8') as ofile: |
|
ofile.write('''/* |
|
* Autogenerated by the Meson build system. |
|
* Do not edit, your changes will be lost. |
|
*/ |
|
|
|
#pragma once |
|
|
|
''') |
|
for k in sorted(cdata.keys()): |
|
(v, desc) = cdata.get(k) |
|
if desc: |
|
ofile.write('/* %s */\n' % desc) |
|
if isinstance(v, bool): |
|
if v: |
|
ofile.write('#define %s\n\n' % k) |
|
else: |
|
ofile.write('#undef %s\n\n' % k) |
|
elif isinstance(v, (int, str)): |
|
ofile.write('#define %s %s\n\n' % (k, v)) |
|
else: |
|
raise MesonException('Unknown data type in configuration file entry: ' + k) |
|
|
|
def replace_if_different(dst, dst_tmp): |
|
# If contents are identical, don't touch the file to prevent |
|
# unnecessary rebuilds. |
|
different = True |
|
try: |
|
with open(dst, 'rb') as f1, open(dst_tmp, 'rb') as f2: |
|
if f1.read() == f2.read(): |
|
different = False |
|
except FileNotFoundError: |
|
pass |
|
if different: |
|
os.replace(dst_tmp, dst) |
|
else: |
|
os.unlink(dst_tmp) |
|
|
|
def listify(item, flatten=True, unholder=False): |
|
''' |
|
Returns a list with all args embedded in a list if they are not a list. |
|
This function preserves order. |
|
@flatten: Convert lists of lists to a flat list |
|
@unholder: Replace each item with the object it holds, if required |
|
|
|
Note: unholding only works recursively when flattening |
|
''' |
|
if not isinstance(item, list): |
|
if unholder and hasattr(item, 'held_object'): |
|
item = item.held_object |
|
return [item] |
|
result = [] |
|
for i in item: |
|
if unholder and hasattr(i, 'held_object'): |
|
i = i.held_object |
|
if flatten and isinstance(i, list): |
|
result += listify(i, flatten=True, unholder=unholder) |
|
else: |
|
result.append(i) |
|
return result |
|
|
|
|
|
def extract_as_list(dict_object, *keys, pop=False, **kwargs): |
|
''' |
|
Extracts all values from given dict_object and listifies them. |
|
''' |
|
result = [] |
|
fetch = dict_object.get |
|
if pop: |
|
fetch = dict_object.pop |
|
# If there's only one key, we don't return a list with one element |
|
if len(keys) == 1: |
|
return listify(fetch(keys[0], []), **kwargs) |
|
# Return a list of values corresponding to *keys |
|
for key in keys: |
|
result.append(listify(fetch(key, []), **kwargs)) |
|
return result |
|
|
|
|
|
def typeslistify(item, types): |
|
''' |
|
Ensure that type(@item) is one of @types or a |
|
list of items all of which are of type @types |
|
''' |
|
if isinstance(item, types): |
|
item = [item] |
|
if not isinstance(item, list): |
|
raise MesonException('Item must be a list or one of {!r}'.format(types)) |
|
for i in item: |
|
if i is not None and not isinstance(i, types): |
|
raise MesonException('List item must be one of {!r}'.format(types)) |
|
return item |
|
|
|
def stringlistify(item): |
|
return typeslistify(item, str) |
|
|
|
def expand_arguments(args): |
|
expended_args = [] |
|
for arg in args: |
|
if not arg.startswith('@'): |
|
expended_args.append(arg) |
|
continue |
|
|
|
args_file = arg[1:] |
|
try: |
|
with open(args_file) as f: |
|
extended_args = f.read().split() |
|
expended_args += extended_args |
|
except Exception as e: |
|
print('Error expanding command line arguments, %s not found' % args_file) |
|
print(e) |
|
return None |
|
return expended_args |
|
|
|
def Popen_safe(args, write=None, stderr=subprocess.PIPE, **kwargs): |
|
if sys.version_info < (3, 6) or not sys.stdout.encoding: |
|
return Popen_safe_legacy(args, write=write, stderr=stderr, **kwargs) |
|
p = subprocess.Popen(args, universal_newlines=True, |
|
close_fds=False, |
|
stdout=subprocess.PIPE, |
|
stderr=stderr, **kwargs) |
|
o, e = p.communicate(write) |
|
return p, o, e |
|
|
|
def Popen_safe_legacy(args, write=None, stderr=subprocess.PIPE, **kwargs): |
|
p = subprocess.Popen(args, universal_newlines=False, |
|
stdout=subprocess.PIPE, |
|
stderr=stderr, **kwargs) |
|
if write is not None: |
|
write = write.encode('utf-8') |
|
o, e = p.communicate(write) |
|
if o is not None: |
|
if sys.stdout.encoding: |
|
o = o.decode(encoding=sys.stdout.encoding, errors='replace').replace('\r\n', '\n') |
|
else: |
|
o = o.decode(errors='replace').replace('\r\n', '\n') |
|
if e is not None: |
|
if sys.stderr.encoding: |
|
e = e.decode(encoding=sys.stderr.encoding, errors='replace').replace('\r\n', '\n') |
|
else: |
|
e = e.decode(errors='replace').replace('\r\n', '\n') |
|
return p, o, e |
|
|
|
def iter_regexin_iter(regexiter, initer): |
|
''' |
|
Takes each regular expression in @regexiter and tries to search for it in |
|
every item in @initer. If there is a match, returns that match. |
|
Else returns False. |
|
''' |
|
for regex in regexiter: |
|
for ii in initer: |
|
if not isinstance(ii, str): |
|
continue |
|
match = re.search(regex, ii) |
|
if match: |
|
return match.group() |
|
return False |
|
|
|
def _substitute_values_check_errors(command, values): |
|
# Error checking |
|
inregex = ('@INPUT([0-9]+)?@', '@PLAINNAME@', '@BASENAME@') |
|
outregex = ('@OUTPUT([0-9]+)?@', '@OUTDIR@') |
|
if '@INPUT@' not in values: |
|
# Error out if any input-derived templates are present in the command |
|
match = iter_regexin_iter(inregex, command) |
|
if match: |
|
m = 'Command cannot have {!r}, since no input files were specified' |
|
raise MesonException(m.format(match)) |
|
else: |
|
if len(values['@INPUT@']) > 1: |
|
# Error out if @PLAINNAME@ or @BASENAME@ is present in the command |
|
match = iter_regexin_iter(inregex[1:], command) |
|
if match: |
|
raise MesonException('Command cannot have {!r} when there is ' |
|
'more than one input file'.format(match)) |
|
# Error out if an invalid @INPUTnn@ template was specified |
|
for each in command: |
|
if not isinstance(each, str): |
|
continue |
|
match = re.search(inregex[0], each) |
|
if match and match.group() not in values: |
|
m = 'Command cannot have {!r} since there are only {!r} inputs' |
|
raise MesonException(m.format(match.group(), len(values['@INPUT@']))) |
|
if '@OUTPUT@' not in values: |
|
# Error out if any output-derived templates are present in the command |
|
match = iter_regexin_iter(outregex, command) |
|
if match: |
|
m = 'Command cannot have {!r} since there are no outputs' |
|
raise MesonException(m.format(match)) |
|
else: |
|
# Error out if an invalid @OUTPUTnn@ template was specified |
|
for each in command: |
|
if not isinstance(each, str): |
|
continue |
|
match = re.search(outregex[0], each) |
|
if match and match.group() not in values: |
|
m = 'Command cannot have {!r} since there are only {!r} outputs' |
|
raise MesonException(m.format(match.group(), len(values['@OUTPUT@']))) |
|
|
|
def substitute_values(command, values): |
|
''' |
|
Substitute the template strings in the @values dict into the list of |
|
strings @command and return a new list. For a full list of the templates, |
|
see get_filenames_templates_dict() |
|
|
|
If multiple inputs/outputs are given in the @values dictionary, we |
|
substitute @INPUT@ and @OUTPUT@ only if they are the entire string, not |
|
just a part of it, and in that case we substitute *all* of them. |
|
''' |
|
# Error checking |
|
_substitute_values_check_errors(command, values) |
|
# Substitution |
|
outcmd = [] |
|
rx_keys = [re.escape(key) for key in values if key not in ('@INPUT@', '@OUTPUT@')] |
|
value_rx = re.compile('|'.join(rx_keys)) if rx_keys else None |
|
for vv in command: |
|
if not isinstance(vv, str): |
|
outcmd.append(vv) |
|
elif '@INPUT@' in vv: |
|
inputs = values['@INPUT@'] |
|
if vv == '@INPUT@': |
|
outcmd += inputs |
|
elif len(inputs) == 1: |
|
outcmd.append(vv.replace('@INPUT@', inputs[0])) |
|
else: |
|
raise MesonException("Command has '@INPUT@' as part of a " |
|
"string and more than one input file") |
|
elif '@OUTPUT@' in vv: |
|
outputs = values['@OUTPUT@'] |
|
if vv == '@OUTPUT@': |
|
outcmd += outputs |
|
elif len(outputs) == 1: |
|
outcmd.append(vv.replace('@OUTPUT@', outputs[0])) |
|
else: |
|
raise MesonException("Command has '@OUTPUT@' as part of a " |
|
"string and more than one output file") |
|
# Append values that are exactly a template string. |
|
# This is faster than a string replace. |
|
elif vv in values: |
|
outcmd.append(values[vv]) |
|
# Substitute everything else with replacement |
|
elif value_rx: |
|
outcmd.append(value_rx.sub(lambda m: values[m.group(0)], vv)) |
|
else: |
|
outcmd.append(vv) |
|
return outcmd |
|
|
|
def get_filenames_templates_dict(inputs, outputs): |
|
''' |
|
Create a dictionary with template strings as keys and values as values for |
|
the following templates: |
|
|
|
@INPUT@ - the full path to one or more input files, from @inputs |
|
@OUTPUT@ - the full path to one or more output files, from @outputs |
|
@OUTDIR@ - the full path to the directory containing the output files |
|
|
|
If there is only one input file, the following keys are also created: |
|
|
|
@PLAINNAME@ - the filename of the input file |
|
@BASENAME@ - the filename of the input file with the extension removed |
|
|
|
If there is more than one input file, the following keys are also created: |
|
|
|
@INPUT0@, @INPUT1@, ... one for each input file |
|
|
|
If there is more than one output file, the following keys are also created: |
|
|
|
@OUTPUT0@, @OUTPUT1@, ... one for each output file |
|
''' |
|
values = {} |
|
# Gather values derived from the input |
|
if inputs: |
|
# We want to substitute all the inputs. |
|
values['@INPUT@'] = inputs |
|
for (ii, vv) in enumerate(inputs): |
|
# Write out @INPUT0@, @INPUT1@, ... |
|
values['@INPUT{}@'.format(ii)] = vv |
|
if len(inputs) == 1: |
|
# Just one value, substitute @PLAINNAME@ and @BASENAME@ |
|
values['@PLAINNAME@'] = plain = os.path.split(inputs[0])[1] |
|
values['@BASENAME@'] = os.path.splitext(plain)[0] |
|
if outputs: |
|
# Gather values derived from the outputs, similar to above. |
|
values['@OUTPUT@'] = outputs |
|
for (ii, vv) in enumerate(outputs): |
|
values['@OUTPUT{}@'.format(ii)] = vv |
|
# Outdir should be the same for all outputs |
|
values['@OUTDIR@'] = os.path.split(outputs[0])[0] |
|
# Many external programs fail on empty arguments. |
|
if values['@OUTDIR@'] == '': |
|
values['@OUTDIR@'] = '.' |
|
return values |
|
|
|
|
|
def windows_proof_rmtree(f): |
|
# On Windows if anyone is holding a file open you can't |
|
# delete it. As an example an anti virus scanner might |
|
# be scanning files you are trying to delete. The only |
|
# way to fix this is to try again and again. |
|
delays = [0.1, 0.1, 0.2, 0.2, 0.2, 0.5, 0.5, 1, 1, 1, 1, 2] |
|
for d in delays: |
|
try: |
|
shutil.rmtree(f) |
|
return |
|
except (OSError, PermissionError): |
|
time.sleep(d) |
|
# Try one last time and throw if it fails. |
|
shutil.rmtree(f) |
|
|
|
|
|
def detect_subprojects(spdir_name, current_dir='', result=None): |
|
if result is None: |
|
result = {} |
|
spdir = os.path.join(current_dir, spdir_name) |
|
if not os.path.exists(spdir): |
|
return result |
|
for trial in glob(os.path.join(spdir, '*')): |
|
basename = os.path.split(trial)[1] |
|
if trial == 'packagecache': |
|
continue |
|
append_this = True |
|
if os.path.isdir(trial): |
|
detect_subprojects(spdir_name, trial, result) |
|
elif trial.endswith('.wrap') and os.path.isfile(trial): |
|
basename = os.path.splitext(basename)[0] |
|
else: |
|
append_this = False |
|
if append_this: |
|
if basename in result: |
|
result[basename].append(trial) |
|
else: |
|
result[basename] = [trial] |
|
return result |
|
|
|
class OrderedSet(collections.MutableSet): |
|
"""A set that preserves the order in which items are added, by first |
|
insertion. |
|
""" |
|
def __init__(self, iterable=None): |
|
self.__container = collections.OrderedDict() |
|
if iterable: |
|
self.update(iterable) |
|
|
|
def __contains__(self, value): |
|
return value in self.__container |
|
|
|
def __iter__(self): |
|
return iter(self.__container.keys()) |
|
|
|
def __len__(self): |
|
return len(self.__container) |
|
|
|
def __repr__(self): |
|
# Don't print 'OrderedSet("")' for an empty set. |
|
if self.__container: |
|
return 'OrderedSet("{}")'.format( |
|
'", "'.join(repr(e) for e in self.__container.keys())) |
|
return 'OrderedSet()' |
|
|
|
def add(self, value): |
|
self.__container[value] = None |
|
|
|
def discard(self, value): |
|
if value in self.__container: |
|
del self.__container[value] |
|
|
|
def update(self, iterable): |
|
for item in iterable: |
|
self.__container[item] = None |
|
|
|
def difference(self, set_): |
|
return type(self)(e for e in self if e not in set_)
|
|
|