3 Miscellaneous utility functions -- anything that doesn't fit into
4 one of the other *util.py modules.
9 import sys
, os
, string
, re
11 from distutils
.errors
import DistutilsPlatformError
12 from distutils
.dep_util
import newer
13 from distutils
.spawn
import spawn
, find_executable
14 from distutils
import log
15 from distutils
.version
import LooseVersion
16 from distutils
.errors
import DistutilsByteCompileError
19 """Return a string that identifies the current platform.
21 This is used mainly to distinguish platform-specific build directories and
22 platform-specific built distributions. Typically includes the OS name
23 and version and the architecture (as supplied by 'os.uname()'),
24 although the exact information included depends on the OS; eg. for IRIX
25 the architecture isn't particularly important (IRIX only runs on SGI
26 hardware), but for Linux the kernel version isn't particularly
29 Examples of returned values:
36 Windows will return one of:
37 win-amd64 (64bit Windows on AMD64 (aka x86_64, Intel64, EM64T, etc)
38 win-ia64 (64bit Windows on Itanium)
39 win32 (all others - specifically, sys.platform is returned)
41 For other non-POSIX platforms, currently just returns 'sys.platform'.
44 # sniff sys.version for architecture.
46 i
= sys
.version
.find(prefix
)
49 j
= sys
.version
.find(")", i
)
50 look
= sys
.version
[i
+len(prefix
):j
].lower()
57 if os
.name
!= "posix" or not hasattr(os
, 'uname'):
58 # XXX what about the architecture? NT is Intel or Alpha,
59 # Mac OS is M68k or PPC, etc.
62 # Try to distinguish various flavours of Unix
64 (osname
, host
, release
, version
, machine
) = os
.uname()
66 # Convert the OS name to lowercase, remove '/' characters
67 # (to accommodate BSD/OS), and translate spaces (for "Power Macintosh")
68 osname
= osname
.lower().replace('/', '')
69 machine
= machine
.replace(' ', '_')
70 machine
= machine
.replace('/', '-')
72 if osname
[:5] == "linux":
73 # At least on Linux/Intel, 'machine' is the processor --
75 # XXX what about Alpha, SPARC, etc?
76 return "%s-%s" % (osname
, machine
)
77 elif osname
[:5] == "sunos":
78 if release
[0] >= "5": # SunOS 5 == Solaris 2
80 release
= "%d.%s" % (int(release
[0]) - 3, release
[2:])
81 # fall through to standard osname-release-machine representation
82 elif osname
[:4] == "irix": # could be "irix64"!
83 return "%s-%s" % (osname
, release
)
84 elif osname
[:3] == "aix":
85 return "%s-%s.%s" % (osname
, version
, release
)
86 elif osname
[:6] == "cygwin":
88 rel_re
= re
.compile (r
'[\d.]+')
89 m
= rel_re
.match(release
)
92 elif osname
[:6] == "darwin":
94 # For our purposes, we'll assume that the system version from
95 # distutils' perspective is what MACOSX_DEPLOYMENT_TARGET is set
96 # to. This makes the compatibility story a bit more sane because the
97 # machine is going to compile and link as if it were
98 # MACOSX_DEPLOYMENT_TARGET.
99 from distutils
.sysconfig
import get_config_vars
100 cfgvars
= get_config_vars()
102 macver
= os
.environ
.get('MACOSX_DEPLOYMENT_TARGET')
104 macver
= cfgvars
.get('MACOSX_DEPLOYMENT_TARGET')
107 # Always calculate the release of the running machine,
108 # needed to determine if we can build fat binaries or not.
111 # Get the system version. Reading this plist is a documented
112 # way to get the system version (see the documentation for
113 # the Gestalt Manager)
115 f
= open('/System/Library/CoreServices/SystemVersion.plist')
117 # We're on a plain darwin box, fall back to the default
122 r
'<key>ProductUserVisibleVersion</key>\s*' +
123 r
'<string>(.*?)</string>', f
.read())
126 macrelease
= '.'.join(m
.group(1).split('.')[:2])
127 # else: fall back to the default behaviour
133 from distutils
.sysconfig
import get_config_vars
137 if (macrelease
+ '.') >= '10.4.' and \
138 '-arch' in get_config_vars().get('CFLAGS', '').strip():
139 # The universal build will build fat binaries, but not on
140 # systems before 10.4
142 # Try to detect 4-way universal builds, those have machine-type
143 # 'universal' instead of 'fat'.
146 cflags
= get_config_vars().get('CFLAGS')
148 archs
= re
.findall('-arch\s+(\S+)', cflags
)
154 elif archs
== ('i386', 'ppc'):
156 elif archs
== ('i386', 'x86_64'):
158 elif archs
== ('i386', 'ppc', 'x86_64'):
160 elif archs
== ('ppc64', 'x86_64'):
162 elif archs
== ('i386', 'ppc', 'ppc64', 'x86_64'):
163 machine
= 'universal'
166 "Don't know machine value for archs=%r"%(archs
,))
169 elif machine
in ('PowerPC', 'Power_Macintosh'):
170 # Pick a sane name for the PPC architecture.
173 return "%s-%s-%s" % (osname
, release
, machine
)
176 def convert_path(pathname
):
177 """Return 'pathname' as a name that will work on the native filesystem.
179 i.e. split it on '/' and put it back together again using the current
180 directory separator. Needed because filenames in the setup script are
181 always supplied in Unix style, and have to be converted to the local
182 convention before we can actually use them in the filesystem. Raises
183 ValueError on non-Unix-ish systems if 'pathname' either starts or
190 if pathname
[0] == '/':
191 raise ValueError("path '%s' cannot be absolute" % pathname
)
192 if pathname
[-1] == '/':
193 raise ValueError("path '%s' cannot end with '/'" % pathname
)
195 paths
= pathname
.split('/')
200 return os
.path
.join(*paths
)
203 def change_root(new_root
, pathname
):
204 """Return 'pathname' with 'new_root' prepended.
206 If 'pathname' is relative, this is equivalent to
207 "os.path.join(new_root,pathname)".
208 Otherwise, it requires making 'pathname' relative and then joining the
209 two, which is tricky on DOS/Windows and Mac OS.
211 if os
.name
== 'posix':
212 if not os
.path
.isabs(pathname
):
213 return os
.path
.join(new_root
, pathname
)
215 return os
.path
.join(new_root
, pathname
[1:])
217 elif os
.name
== 'nt':
218 (drive
, path
) = os
.path
.splitdrive(pathname
)
221 return os
.path
.join(new_root
, path
)
223 elif os
.name
== 'os2':
224 (drive
, path
) = os
.path
.splitdrive(pathname
)
225 if path
[0] == os
.sep
:
227 return os
.path
.join(new_root
, path
)
229 elif os
.name
== 'mac':
230 if not os
.path
.isabs(pathname
):
231 return os
.path
.join(new_root
, pathname
)
233 # Chop off volume name from start of path
234 elements
= pathname
.split(":", 1)
235 pathname
= ":" + elements
[1]
236 return os
.path
.join(new_root
, pathname
)
239 raise DistutilsPlatformError("nothing known about "
240 "platform '%s'" % os
.name
)
245 """Ensure that 'os.environ' has all the environment variables needed.
247 We guarantee that users can use in config files, command-line options,
248 etc. Currently this includes:
249 HOME - user's home directory (Unix only)
250 PLAT - description of the current platform, including hardware
251 and OS (see 'get_platform()')
253 global _environ_checked
257 if os
.name
== 'posix' and 'HOME' not in os
.environ
:
259 os
.environ
['HOME'] = pwd
.getpwuid(os
.getuid())[5]
261 if 'PLAT' not in os
.environ
:
262 os
.environ
['PLAT'] = get_platform()
266 def subst_vars(s
, local_vars
):
267 """Perform shell/Perl-style variable substitution on 'string'.
269 Every occurrence of '$' followed by a name is considered a variable, and
270 variable is substituted by the value found in the 'local_vars'
271 dictionary, or in 'os.environ' if it's not in 'local_vars'.
272 'os.environ' is first checked/augmented to guarantee that it contains
273 certain values: see 'check_environ()'. Raise ValueError for any
274 variables not found in either 'local_vars' or 'os.environ'.
277 def _subst (match
, local_vars
=local_vars
):
278 var_name
= match
.group(1)
279 if var_name
in local_vars
:
280 return str(local_vars
[var_name
])
282 return os
.environ
[var_name
]
285 return re
.sub(r
'\$([a-zA-Z_][a-zA-Z_0-9]*)', _subst
, s
)
286 except KeyError, var
:
287 raise ValueError("invalid variable '$%s'" % var
)
289 def grok_environment_error(exc
, prefix
="error: "):
290 """Generate a useful error message from an EnvironmentError.
292 This will generate an IOError or an OSError exception object.
293 Handles Python 1.5.1 and 1.5.2 styles, and
294 does what it can to deal with exception objects that don't have a
295 filename (which happens when the error is due to a two-file operation,
296 such as 'rename()' or 'link()'. Returns the error message as a string
297 prefixed with 'prefix'.
299 # check for Python 1.5.2-style {IO,OS}Error exception objects
300 if hasattr(exc
, 'filename') and hasattr(exc
, 'strerror'):
302 error
= prefix
+ "%s: %s" % (exc
.filename
, exc
.strerror
)
304 # two-argument functions in posix module don't
305 # include the filename in the exception object!
306 error
= prefix
+ "%s" % exc
.strerror
308 error
= prefix
+ str(exc
[-1])
312 # Needed by 'split_quoted()'
313 _wordchars_re
= _squote_re
= _dquote_re
= None
316 global _wordchars_re
, _squote_re
, _dquote_re
317 _wordchars_re
= re
.compile(r
'[^\\\'\"%s ]*' % string.whitespace)
318 _squote_re = re.compile(r"'(?
:[^
'\\]|\\.)*'")
319 _dquote_re = re.compile(r'"(?
:[^
"\\]|\\.)*"')
322 """Split a string up according to Unix shell-like rules for quotes and
325 In short: words are delimited by spaces, as long as those
326 spaces are not escaped by a backslash, or inside a quoted string.
327 Single and double quotes are equivalent, and the quote characters can
328 be backslash-escaped. The backslash is stripped from any two-character
329 escape sequence, leaving only the escaped character. The quote
330 characters are stripped from any quoted string. Returns a list of
333 # This is a nice algorithm for splitting up a single string, since it
334 # doesn't require character
-by
-character examination
. It was a little
335 # bit of a brain-bender to get it working right, though...
336 if _wordchars_re
is None: _init_regex()
343 m
= _wordchars_re
.match(s
, pos
)
346 words
.append(s
[:end
])
349 if s
[end
] in string
.whitespace
: # unescaped, unquoted whitespace: now
350 words
.append(s
[:end
]) # we definitely have a word delimiter
354 elif s
[end
] == '\\': # preserve whatever is being escaped;
355 # will become part of the current word
356 s
= s
[:end
] + s
[end
+1:]
360 if s
[end
] == "'": # slurp singly-quoted string
361 m
= _squote_re
.match(s
, end
)
362 elif s
[end
] == '"': # slurp doubly-quoted string
363 m
= _dquote_re
.match(s
, end
)
365 raise RuntimeError("this can't happen "
366 "(bad char '%c')" % s
[end
])
369 raise ValueError("bad string (mismatched %s quotes?)" % s
[end
])
371 (beg
, end
) = m
.span()
372 s
= s
[:beg
] + s
[beg
+1:end
-1] + s
[end
:]
382 def execute(func
, args
, msg
=None, verbose
=0, dry_run
=0):
383 """Perform some action that affects the outside world.
385 eg. by writing to the filesystem). Such actions are special because
386 they are disabled by the 'dry_run' flag. This method takes care of all
387 that bureaucracy for you; all you have to do is supply the
388 function to call and an argument tuple for it (to embody the
389 "external action" being performed), and an optional message to
393 msg
= "%s%r" % (func
.__name
__, args
)
394 if msg
[-2:] == ',)': # correct for singleton tuple
395 msg
= msg
[0:-2] + ')'
403 """Convert a string representation of truth to true (1) or false (0).
405 True values are 'y', 'yes', 't', 'true', 'on', and '1'; false values
406 are 'n', 'no', 'f', 'false', 'off', and '0'. Raises ValueError if
407 'val' is anything else.
410 if val
in ('y', 'yes', 't', 'true', 'on', '1'):
412 elif val
in ('n', 'no', 'f', 'false', 'off', '0'):
415 raise ValueError, "invalid truth value %r" % (val
,)
418 def byte_compile(py_files
, optimize
=0, force
=0, prefix
=None, base_dir
=None,
419 verbose
=1, dry_run
=0, direct
=None):
420 """Byte-compile a collection of Python source files to either .pyc
421 or .pyo files in the same directory.
423 'py_files' is a list of files to compile; any files that don't end in
424 ".py" are silently skipped. 'optimize' must be one of the following:
425 0 - don't optimize (generate .pyc)
426 1 - normal optimization (like "python -O")
427 2 - extra optimization (like "python -OO")
428 If 'force' is true, all files are recompiled regardless of
431 The source filename encoded in each bytecode file defaults to the
432 filenames listed in 'py_files'; you can modify these with 'prefix' and
433 'basedir'. 'prefix' is a string that will be stripped off of each
434 source filename, and 'base_dir' is a directory name that will be
435 prepended (after 'prefix' is stripped). You can supply either or both
436 (or neither) of 'prefix' and 'base_dir', as you wish.
438 If 'dry_run' is true, doesn't actually do anything that would
439 affect the filesystem.
441 Byte-compilation is either done directly in this interpreter process
442 with the standard py_compile module, or indirectly by writing a
443 temporary script and executing it. Normally, you should let
444 'byte_compile()' figure out to use direct compilation or not (see
445 the source for details). The 'direct' flag is used by the script
446 generated in indirect mode; unless you know what you're doing, leave
449 # nothing is done if sys.dont_write_bytecode is True
450 if sys
.dont_write_bytecode
:
451 raise DistutilsByteCompileError('byte-compiling is disabled.')
453 # First, if the caller didn't force us into direct or indirect mode,
454 # figure out which mode we should be in. We take a conservative
455 # approach: choose direct mode *only* if the current interpreter is
456 # in debug mode and optimize is 0. If we're not in debug mode (-O
457 # or -OO), we don't know which level of optimization this
458 # interpreter is running with, so we can't do direct
459 # byte-compilation and be certain that it's the right thing. Thus,
460 # always compile indirectly if the current interpreter is in either
461 # optimize mode, or if either optimization level was requested by
464 direct
= (__debug__
and optimize
== 0)
466 # "Indirect" byte-compilation: write a temporary script and then
467 # run it with the appropriate flags.
470 from tempfile
import mkstemp
471 (script_fd
, script_name
) = mkstemp(".py")
473 from tempfile
import mktemp
474 (script_fd
, script_name
) = None, mktemp(".py")
475 log
.info("writing byte-compilation script '%s'", script_name
)
477 if script_fd
is not None:
478 script
= os
.fdopen(script_fd
, "w")
480 script
= open(script_name
, "w")
483 from distutils.util import byte_compile
487 # XXX would be nice to write absolute filenames, just for
488 # safety's sake (script should be more robust in the face of
489 # chdir'ing before running it). But this requires abspath'ing
490 # 'prefix' as well, and that breaks the hack in build_lib's
491 # 'byte_compile()' method that carefully tacks on a trailing
492 # slash (os.sep really) to make sure the prefix here is "just
493 # right". This whole prefix business is rather delicate -- the
494 # problem is that it's really a directory, but I'm treating it
495 # as a dumb string, so trailing slashes and so forth matter.
497 #py_files = map(os.path.abspath, py_files)
499 # prefix = os.path.abspath(prefix)
501 script
.write(",\n".join(map(repr, py_files
)) + "]\n")
503 byte_compile(files, optimize=%r, force=%r,
504 prefix=%r, base_dir=%r,
505 verbose=%r, dry_run=0,
507 """ % (optimize
, force
, prefix
, base_dir
, verbose
))
511 cmd
= [sys
.executable
, script_name
]
516 spawn(cmd
, dry_run
=dry_run
)
517 execute(os
.remove
, (script_name
,), "removing %s" % script_name
,
520 # "Direct" byte-compilation: use the py_compile module to compile
521 # right here, right now. Note that the script generated in indirect
522 # mode simply calls 'byte_compile()' in direct mode, a weird sort of
523 # cross-process recursion. Hey, it works!
525 from py_compile
import compile
527 for file in py_files
:
528 if file[-3:] != ".py":
529 # This lets us be lazy and not filter filenames in
530 # the "install_lib" command.
533 # Terminology from the py_compile module:
534 # cfile - byte-compiled file
535 # dfile - purported source filename (same as 'file' by default)
536 cfile
= file + (__debug__
and "c" or "o")
539 if file[:len(prefix
)] != prefix
:
540 raise ValueError("invalid prefix: filename %r doesn't "
541 "start with %r" % (file, prefix
))
542 dfile
= dfile
[len(prefix
):]
544 dfile
= os
.path
.join(base_dir
, dfile
)
546 cfile_base
= os
.path
.basename(cfile
)
548 if force
or newer(file, cfile
):
549 log
.info("byte-compiling %s to %s", file, cfile_base
)
551 compile(file, cfile
, dfile
)
553 log
.debug("skipping byte-compilation of %s to %s",
557 def rfc822_escape(header
):
558 """Return a version of the string escaped for inclusion in an
559 RFC-822 header, by ensuring there are 8 spaces space after each newline.
561 lines
= [x
.strip() for x
in header
.split('\n')]
563 return sep
.join(lines
)
565 _RE_VERSION
= re
.compile('(\d+\.\d+(\.\d+)*)')
566 _MAC_OS_X_LD_VERSION
= re
.compile('^@\(#\)PROGRAM:ld PROJECT:ld64-((\d+)(\.\d+)*)')
568 def _find_ld_version():
569 """Finds the ld version. The version scheme differs under Mac OSX."""
570 if sys
.platform
== 'darwin':
571 return _find_exe_version('ld -v', _MAC_OS_X_LD_VERSION
)
573 return _find_exe_version('ld -v')
575 def _find_exe_version(cmd
, pattern
=_RE_VERSION
):
576 """Find the version of an executable by running `cmd` in the shell.
578 `pattern` is a compiled regular expression. If not provided, default
579 to _RE_VERSION. If the command is not found, or the output does not
580 match the mattern, returns None.
582 from subprocess
import Popen
, PIPE
583 executable
= cmd
.split()[0]
584 if find_executable(executable
) is None:
586 pipe
= Popen(cmd
, shell
=True, stdout
=PIPE
, stderr
=PIPE
)
588 stdout
, stderr
= pipe
.stdout
.read(), pipe
.stderr
.read()
592 # some commands like ld under MacOS X, will give the
593 # output in the stderr, rather than stdout.
599 result
= pattern
.search(out_string
)
602 return LooseVersion(result
.group(1))
604 def get_compiler_versions():
605 """Returns a tuple providing the versions of gcc, ld and dllwrap
607 For each command, if a command is not found, None is returned.
608 Otherwise a LooseVersion instance is returned.
610 gcc
= _find_exe_version('gcc -dumpversion')
611 ld
= _find_ld_version()
612 dllwrap
= _find_exe_version('dllwrap --version')
613 return gcc
, ld
, dllwrap