Google Git
Sign in
chromium / external / github.com / python / cpython / refs/tags/v2.6.7 / . / README
blob: a315ad94d981b64d8c5501c7db84cca4529752df [file] [log] [blame]
Barry Warsaw16ec24a2011-06-04 00:02:47[diff] [blame]1This is Python version 2.6.7
2============================
Guido van Rossum91447632000-04-11 17:11:09[diff] [blame]3
Guido van Rossum197f7f62011-03-19 23:20:06[diff] [blame]4Copyright (c) 2001, 2002, 2003, 2004, 2005, 2006, 2007, 2008, 2009, 2010, 2011
Barry Warsawcf0d8ab2011-05-23 19:22:56[diff] [blame]5Python Software Foundation.
6All rights reserved.
Guido van Rossum67345eb2001-01-18 14:51:12[diff] [blame]7
Guido van Rossumc07d5fa2000-09-01 22:50:02[diff] [blame]8Copyright (c) 2000 BeOpen.com.
9All rights reserved.
10
Guido van Rossum23f7aed2001-04-12 20:53:31[diff] [blame]11Copyright (c) 1995-2001 Corporation for National Research Initiatives.
Guido van Rossumc07d5fa2000-09-01 22:50:02[diff] [blame]12All rights reserved.
13
Guido van Rossum6ebd2992000-09-03 04:47:47[diff] [blame]14Copyright (c) 1991-1995 Stichting Mathematisch Centrum.
Guido van Rossumc07d5fa2000-09-01 22:50:02[diff] [blame]15All rights reserved.
16
Guido van Rossum4f40c2d2000-09-05 16:46:22[diff] [blame]17
Guido van Rossumc07d5fa2000-09-01 22:50:02[diff] [blame]18License information
19-------------------
20
Guido van Rossum4f40c2d2000-09-05 16:46:22[diff] [blame]21See the file "LICENSE" for information on the history of this
22software, terms & conditions for usage, and a DISCLAIMER OF ALL
23WARRANTIES.
Guido van Rossumc07d5fa2000-09-01 22:50:02[diff] [blame]24
Guido van Rossum4f40c2d2000-09-05 16:46:22[diff] [blame]25This Python distribution contains no GNU General Public Licensed
26(GPLed) code so it may be used in proprietary projects just like prior
27Python distributions. There are interfaces to some GNU code but these
28are entirely optional.
29
30All trademarks referenced herein are property of their respective
31holders.
Guido van Rossum79808261997-12-11 18:01:47[diff] [blame]32
33
Guido van Rossum8d7d4ed1996-07-30 21:41:07[diff] [blame]34What's new in this release?
35---------------------------
Guido van Rossum3ff96dd1996-07-30 18:05:04[diff] [blame]36
Guido van Rossum67345eb2001-01-18 14:51:12[diff] [blame]37See the file "Misc/NEWS".
Guido van Rossum64773801997-11-26 17:07:02[diff] [blame]38
Michael W. Hudson71dcc3e2005-02-22 15:33:26[diff] [blame]39
Guido van Rossum64773801997-11-26 17:07:02[diff] [blame]40If you don't read instructions
41------------------------------
42
43Congratulations on getting this far. :-)
44
45To start building right away (on UNIX): type "./configure" in the
Guido van Rossum2dcec0c2002-02-04 01:59:23[diff] [blame]46current directory and when it finishes, type "make". This creates an
47executable "./python"; to install in /usr/local, first do "su root"
48and then "make install".
49
Michael W. Hudson71dcc3e2005-02-22 15:33:26[diff] [blame]50The section `Build instructions' below is still recommended reading.
Guido van Rossum3ff96dd1996-07-30 18:05:04[diff] [blame]51
52
Guido van Rossum91cb9d21995-04-10 11:47:38[diff] [blame]53What is Python anyway?
54----------------------
55
Michael W. Hudson71dcc3e2005-02-22 15:33:26[diff] [blame]56Python is an interpreted, interactive object-oriented programming
57language suitable (amongst other uses) for distributed application
58development, scripting, numeric computing and system testing. Python
59is often compared to Tcl, Perl, Java, JavaScript, Visual Basic or
60Scheme. To find out more about what Python can do for you, point your
61browser to http://www.python.org/.
Guido van Rossum91cb9d21995-04-10 11:47:38[diff] [blame]62
63
Guido van Rossumf501b4e1996-10-25 14:32:48[diff] [blame]64How do I learn Python?
65----------------------
66
Fred Drake0e6444c1999-05-17 19:35:01[diff] [blame]67The official tutorial is still a good place to start; see
Michael W. Hudson71dcc3e2005-02-22 15:33:26[diff] [blame]68http://docs.python.org/ for online and downloadable versions, as well
69as a list of other introductions, and reference documentation.
Guido van Rossumf501b4e1996-10-25 14:32:48[diff] [blame]70
Guido van Rossumd0a42e22000-03-31 20:16:45[diff] [blame]71There's a quickly growing set of books on Python. See
Michael W. Hudsonfad46402005-06-16 15:51:20[diff] [blame]72http://wiki.python.org/moin/PythonBooks for a list.
Guido van Rossum8d90f9d1997-05-22 20:13:25[diff] [blame]73
Guido van Rossumf501b4e1996-10-25 14:32:48[diff] [blame]74
Guido van Rossumc07d5fa2000-09-01 22:50:02[diff] [blame]75Documentation
76-------------
Guido van Rossum91cb9d21995-04-10 11:47:38[diff] [blame]77
Guido van Rossumc07d5fa2000-09-01 22:50:02[diff] [blame]78All documentation is provided online in a variety of formats. In
79order of importance for new users: Tutorial, Library Reference,
Guido van Rossum4f40c2d2000-09-05 16:46:22[diff] [blame]80Language Reference, Extending & Embedding, and the Python/C API. The
81Library Reference is especially of immense value since much of
82Python's power is described there, including the built-in data types
83and functions!
Guido van Rossum91cb9d21995-04-10 11:47:38[diff] [blame]84
Guido van Rossum4f40c2d2000-09-05 16:46:22[diff] [blame]85All documentation is also available online at the Python web site
Benjamin Peterson9d0b6042008-03-30 19:35:10[diff] [blame]86(http://docs.python.org/, see below). It is available online for occasional
87reference, or can be downloaded in many formats for faster access. The
Georg Brandl1c88e0f2008-03-30 19:41:39[diff] [blame]88documentation is downloadable in HTML, PostScript, PDF, LaTeX, and
Benjamin Peterson9d0b6042008-03-30 19:35:10[diff] [blame]89reStructuredText (2.6+) formats; the LaTeX and reStructuredText versions are
90primarily for documentation authors, translators, and people with special
91formatting requirements.
Guido van Rossumc07d5fa2000-09-01 22:50:02[diff] [blame]92
Guido van Rossumc07d5fa2000-09-01 22:50:02[diff] [blame]93
Guido van Rossum4f40c2d2000-09-05 16:46:22[diff] [blame]94Web sites
95---------
Guido van Rossumc07d5fa2000-09-01 22:50:02[diff] [blame]96
Guido van Rossum4f40c2d2000-09-05 16:46:22[diff] [blame]97New Python releases and related technologies are published at
Guido van Rossum67345eb2001-01-18 14:51:12[diff] [blame]98http://www.python.org/. Come visit us!
Guido van Rossumc07d5fa2000-09-01 22:50:02[diff] [blame]99
Guido van Rossum4f40c2d2000-09-05 16:46:22[diff] [blame]100
101Newsgroups and Mailing Lists
102----------------------------
Guido van Rossumc07d5fa2000-09-01 22:50:02[diff] [blame]103
104Read comp.lang.python, a high-volume discussion newsgroup about
105Python, or comp.lang.python.announce, a low-volume moderated newsgroup
106for Python-related announcements. These are also accessible as
Georg Brandlf004d9d2009-10-27 15:39:53[diff] [blame]107mailing lists: see http://www.python.org/community/lists/ for an
Michael W. Hudson71dcc3e2005-02-22 15:33:26[diff] [blame]108overview of these and many other Python-related mailing lists.
Guido van Rossumc07d5fa2000-09-01 22:50:02[diff] [blame]109
Anthony Baxter70e07962006-04-03 14:16:27[diff] [blame]110Archives are accessible via the Google Groups Usenet archive; see
Michael W. Hudsonbe6e0f42002-06-23 16:29:36[diff] [blame]111http://groups.google.com/. The mailing lists are also archived, see
Georg Brandlf004d9d2009-10-27 15:39:53[diff] [blame]112http://www.python.org/community/lists/ for details.
Guido van Rossumc07d5fa2000-09-01 22:50:02[diff] [blame]113
114
115Bug reports
116-----------
117
Guido van Rossum4f40c2d2000-09-05 16:46:22[diff] [blame]118To report or search for bugs, please use the Python Bug
Georg Brandlf004d9d2009-10-27 15:39:53[diff] [blame]119Tracker at http://bugs.python.org/.
Guido van Rossumc07d5fa2000-09-01 22:50:02[diff] [blame]120
121
122Patches and contributions
123-------------------------
124
Fred Drake6caae142000-10-25 17:51:02[diff] [blame]125To submit a patch or other contribution, please use the Python Patch
Georg Brandlf004d9d2009-10-27 15:39:53[diff] [blame]126Manager at http://bugs.python.org/. Guidelines
Andrew M. Kuchling951300e2008-01-04 14:47:17[diff] [blame]127for patch submission may be found at http://www.python.org/dev/patches/.
Guido van Rossumc07d5fa2000-09-01 22:50:02[diff] [blame]128
Benjamin Petersonb2355002008-04-02 21:20:35[diff] [blame]129If you have a proposal to change Python, you may want to send an email to the
130comp.lang.python or python-ideas mailing lists for inital feedback. A Python
131Enhancement Proposal (PEP) may be submitted if your idea gains ground. All
132current PEPs, as well as guidelines for submitting a new PEP, are listed at
133http://www.python.org/dev/peps/.
Guido van Rossumc07d5fa2000-09-01 22:50:02[diff] [blame]134
135
136Questions
137---------
138
139For help, if you can't find it in the manuals or on the web site, it's
140best to post to the comp.lang.python or the Python mailing list (see
141above). If you specifically don't want to involve the newsgroup or
Guido van Rossum67345eb2001-01-18 14:51:12[diff] [blame]142mailing list, send questions to help@python.org (a group of volunteers
143who answer questions as they can). The newsgroup is the most
144efficient way to ask public questions.
Guido van Rossum627b2d71993-12-24 10:39:16[diff] [blame]145
Guido van Rossum901454e2000-06-29 22:28:44[diff] [blame]146
Guido van Rossum433c8ad1994-08-01 12:07:07[diff] [blame]147Build instructions
Guido van Rossum91cb9d21995-04-10 11:47:38[diff] [blame]148==================
Guido van Rossum627b2d71993-12-24 10:39:16[diff] [blame]149
Guido van Rossum59b0b602002-02-08 22:22:35[diff] [blame]150Before you can build Python, you must first configure it.
151Fortunately, the configuration and build process has been automated
152for Unix and Linux installations, so all you usually have to do is
153type a few commands and sit back. There are some platforms where
154things are not quite as smooth; see the platform specific notes below.
155If you want to build for multiple platforms sharing the same source
156tree, see the section on VPATH below.
Guido van Rossum627b2d71993-12-24 10:39:16[diff] [blame]157
Guido van Rossum59b0b602002-02-08 22:22:35[diff] [blame]158Start by running the script "./configure", which determines your
159system configuration and creates the Makefile. (It takes a minute or
160two -- please be patient!) You may want to pass options to the
161configure script -- see the section below on configuration options and
162variables. When it's done, you are ready to run make.
Guido van Rossum6d9cc801996-08-01 17:31:22[diff] [blame]163
Guido van Rossum59b0b602002-02-08 22:22:35[diff] [blame]164To build Python, you normally type "make" in the toplevel directory.
165If you have changed the configuration, the Makefile may have to be
Georg Brandlb6ca2d62010-05-18 23:38:50[diff] [blame]166rebuilt. In this case, you may have to run make again to correctly
Guido van Rossum59b0b602002-02-08 22:22:35[diff] [blame]167build your desired target. The interpreter executable is built in the
168top level directory.
Guido van Rossum6d9cc801996-08-01 17:31:22[diff] [blame]169
Guido van Rossum4f40c2d2000-09-05 16:46:22[diff] [blame]170Once you have built a Python interpreter, see the subsections below on
Guido van Rossum59b0b602002-02-08 22:22:35[diff] [blame]171testing and installation. If you run into trouble, see the next
172section.
173
174Previous versions of Python used a manual configuration process that
175involved editing the file Modules/Setup. While this file still exists
176and manual configuration is still supported, it is rarely needed any
177more: almost all modules are automatically built as appropriate under
178guidance of the setup.py script, which is run by Make after the
179interpreter has been built.
Guido van Rossum627b2d71993-12-24 10:39:16[diff] [blame]180
Guido van Rossum0a516c91994-09-12 10:58:40[diff] [blame]181
Guido van Rossum91cb9d21995-04-10 11:47:38[diff] [blame]182Troubleshooting
183---------------
Guido van Rossum627b2d71993-12-24 10:39:16[diff] [blame]184
Guido van Rossum6d9cc801996-08-01 17:31:22[diff] [blame]185See also the platform specific notes in the next section.
186
Michael W. Hudson71dcc3e2005-02-22 15:33:26[diff] [blame]187If you run into other trouble, see the FAQ
Georg Brandlf004d9d2009-10-27 15:39:53[diff] [blame]188(http://www.python.org/doc/faq/) for hints on what can go wrong, and
Michael W. Hudson71dcc3e2005-02-22 15:33:26[diff] [blame]189how to fix it.
Guido van Rossum91cb9d21995-04-10 11:47:38[diff] [blame]190
Guido van Rossum6d9cc801996-08-01 17:31:22[diff] [blame]191If you rerun the configure script with different options, remove all
192object files by running "make clean" before rebuilding. Believe it or
193not, "make clean" sometimes helps to clean up other inexplicable
194problems as well. Try it before sending in a bug report!
195
Guido van Rossumd0fe8451996-08-26 03:02:37[diff] [blame]196If the configure script fails or doesn't seem to find things that
Michael W. Hudson71dcc3e2005-02-22 15:33:26[diff] [blame]197should be there, inspect the config.log file.
Guido van Rossumc0be2f51997-10-08 05:05:28[diff] [blame]198
199If you get a warning for every file about the -Olimit option being no
200longer supported, you can ignore it. There's no foolproof way to know
Guido van Rossum4f40c2d2000-09-05 16:46:22[diff] [blame]201whether this option is needed; all we can do is test whether it is
Guido van Rossumc0be2f51997-10-08 05:05:28[diff] [blame]202accepted without error. On some systems, e.g. older SGI compilers, it
203is essential for performance (specifically when compiling ceval.c,
204which has more basic blocks than the default limit of 1000). If the
205warning bothers you, edit the Makefile to remove "-Olimit 1500" from
206the OPT variable.
Guido van Rossumd0fe8451996-08-26 03:02:37[diff] [blame]207
Guido van Rossum7ac11532000-08-29 21:36:40[diff] [blame]208If you get failures in test_long, or sys.maxint gets set to -1, you
209are probably experiencing compiler bugs, usually related to
Martin v. Löwis9cea5f52002-11-15 07:28:05[diff] [blame]210optimization. This is a common problem with some versions of gcc, and
211some vendor-supplied compilers, which can sometimes be worked around
212by turning off optimization. Consider switching to stable versions
Anthony Baxtere9d719b2004-11-30 01:49:18[diff] [blame]213(gcc 2.95.2, gcc 3.x, or contact your vendor.)
Guido van Rossum7ac11532000-08-29 21:36:40[diff] [blame]214
215From Python 2.0 onward, all Python C code is ANSI C. Compiling using
216old K&R-C-only compilers is no longer possible. ANSI C compilers are
217available for all modern systems, either in the form of updated
Martin v. Löwis9cea5f52002-11-15 07:28:05[diff] [blame]218compilers from the vendor, or one of the free compilers (gcc).
Guido van Rossum91cb9d21995-04-10 11:47:38[diff] [blame]219
Georg Brandl250755b2006-11-15 17:42:03[diff] [blame]220If "make install" fails mysteriously during the "compiling the library"
221step, make sure that you don't have any of the PYTHONPATH or PYTHONHOME
222environment variables set, as they may interfere with the newly built
223executable which is compiling the library.
224
Martin v. Löwis779ffc02002-12-02 22:17:01[diff] [blame]225Unsupported systems
226-------------------
227
Georg Brandl4f6ebc22010-06-12 06:46:31[diff] [blame]228A number of systems is not supported in Python 2.6 anymore. Some
229support code is still present, but will be removed in later versions.
Martin v. Löwis779ffc02002-12-02 22:17:01[diff] [blame]230If you still need to use current Python versions on these systems,
231please send a message to python-dev@python.org indicating that you
Martin v. Löwis8cab8b02006-07-24 05:05:22[diff] [blame]232volunteer to support this system. For a more detailed discussion
233regarding no-longer-supported and resupporting platforms, as well
234as a list of platforms that became or will be unsupported, see PEP 11.
Martin v. Löwis779ffc02002-12-02 22:17:01[diff] [blame]235
236More specifically, the following systems are not supported any
237longer:
238- SunOS 4
239- DYNIX
240- dgux
241- Minix
Martin v. Löwis8cab8b02006-07-24 05:05:22[diff] [blame]242- NeXT
Martin v. Löwis779ffc02002-12-02 22:17:01[diff] [blame]243- Irix 4 and --with-sgi-dl
244- Linux 1
245- Systems defining __d6_pthread_create (configure.in)
246- Systems defining PY_PTHREAD_D4, PY_PTHREAD_D6,
247 or PY_PTHREAD_D7 in thread_pthread.h
248- Systems using --with-dl-dld
Martin v. Löwis7e4cfcb2002-12-19 16:21:49[diff] [blame]249- Systems using --without-universal-newlines
Martin v. Löwis8cab8b02006-07-24 05:05:22[diff] [blame]250- MacOS 9
251
252The following systems are still supported in Python 2.5, but
253support will be dropped in 2.6:
254- Systems using --with-wctype-functions
255- Win9x, WinME
256
257Warning on install in Windows 98 and Windows Me
258-----------------------------------------------
259
260Following Microsoft's closing of Extended Support for
261Windows 98/ME (July 11, 2006), Python 2.6 will stop
262supporting these platforms. Python development and
263maintainability becomes easier (and more reliable) when
264platform specific code targeting OSes with few users
265and no dedicated expert developers is taken out. The
266vendor also warns that the OS versions listed above
267"can expose customers to security risks" and recommends
268upgrade.
Martin v. Löwis779ffc02002-12-02 22:17:01[diff] [blame]269
Guido van Rossum91cb9d21995-04-10 11:47:38[diff] [blame]270Platform specific notes
271-----------------------
272
Guido van Rossum0447a321995-10-08 01:22:33[diff] [blame]273(Some of these may no longer apply. If you find you can build Python
Guido van Rossum67345eb2001-01-18 14:51:12[diff] [blame]274on these platforms without the special directions mentioned here,
275submit a documentation bug report to SourceForge (see Bug Reports
276above) so we can remove them!)
Guido van Rossum91cb9d21995-04-10 11:47:38[diff] [blame]277
Skip Montanaro6dc4a8e2003-05-02 15:27:59[diff] [blame]278Unix platforms: If your vendor still ships (and you still use) Berkeley DB
Anthony Baxter70e07962006-04-03 14:16:27[diff] [blame]279 1.85 you will need to edit Modules/Setup to build the bsddb185
280 module and add a line to sitecustomize.py which makes it the
281 default. In Modules/Setup a line like
Skip Montanaro6dc4a8e2003-05-02 15:27:59[diff] [blame]282
Anthony Baxter70e07962006-04-03 14:16:27[diff] [blame]283 bsddb185 bsddbmodule.c
Skip Montanaro6dc4a8e2003-05-02 15:27:59[diff] [blame]284
Anthony Baxter70e07962006-04-03 14:16:27[diff] [blame]285 should work. (You may need to add -I, -L or -l flags to direct the
286 compiler and linker to your include files and libraries.)
Skip Montanaro6dc4a8e2003-05-02 15:27:59[diff] [blame]287
Michael W. Hudson71dcc3e2005-02-22 15:33:26[diff] [blame]288XXX I think this next bit is out of date:
Skip Montanaro6dc4a8e2003-05-02 15:27:59[diff] [blame]289
Brett Cannondc48b742007-05-20 07:09:50[diff] [blame]29064-bit platforms: The modules audioop, and imageop don't work.
Anthony Baxter70e07962006-04-03 14:16:27[diff] [blame]291 The setup.py script disables them on 64-bit installations.
292 Don't try to enable them in the Modules/Setup file. They
293 contain code that is quite wordsize sensitive. (If you have a
294 fix, let us know!)
Guido van Rossumc0be2f51997-10-08 05:05:28[diff] [blame]295
Guido van Rossum4462e931997-01-22 21:00:32[diff] [blame]296Solaris: When using Sun's C compiler with threads, at least on Solaris
Anthony Baxter70e07962006-04-03 14:16:27[diff] [blame]297 2.5.1, you need to add the "-mt" compiler option (the simplest
298 way is probably to specify the compiler with this option as
299 the "CC" environment variable when running the configure
300 script).
Guido van Rossum4462e931997-01-22 21:00:32[diff] [blame]301
Guido van Rossum42992312002-09-23 14:02:18[diff] [blame]302 When using GCC on Solaris, beware of binutils 2.13 or GCC
303 versions built using it. This mistakenly enables the
304 -zcombreloc option which creates broken shared libraries on
305 Solaris. binutils 2.12 works, and the binutils maintainers
Skip Montanaro4de9cba2003-01-03 16:26:23[diff] [blame]306 are aware of the problem. Binutils 2.13.1 only partially
307 fixed things. It appears that 2.13.2 solves the problem
308 completely. This problem is known to occur with Solaris 2.7
309 and 2.8, but may also affect earlier and later versions of the
310 OS.
Guido van Rossum42992312002-09-23 14:02:18[diff] [blame]311
Anthony Baxter70e07962006-04-03 14:16:27[diff] [blame]312 When the dynamic loader complains about errors finding shared
313 libraries, such as
Martin v. Löwis9cea5f52002-11-15 07:28:05[diff] [blame]314
Anthony Baxter70e07962006-04-03 14:16:27[diff] [blame]315 ld.so.1: ./python: fatal: libstdc++.so.5: open failed:
316 No such file or directory
Martin v. Löwis9cea5f52002-11-15 07:28:05[diff] [blame]317
Anthony Baxter70e07962006-04-03 14:16:27[diff] [blame]318 you need to first make sure that the library is available on
319 your system. Then, you need to instruct the dynamic loader how
320 to find it. You can choose any of the following strategies:
Martin v. Löwis9cea5f52002-11-15 07:28:05[diff] [blame]321
Anthony Baxter70e07962006-04-03 14:16:27[diff] [blame]322 1. When compiling Python, set LD_RUN_PATH to the directories
323 containing missing libraries.
324 2. When running Python, set LD_LIBRARY_PATH to these directories.
325 3. Use crle(8) to extend the search path of the loader.
326 4. Modify the installed GCC specs file, adding -R options into the
327 *link: section.
Martin v. Löwis9cea5f52002-11-15 07:28:05[diff] [blame]328
Skip Montanaro9f6606a2005-11-17 18:19:39[diff] [blame]329 The complex object fails to compile on Solaris 10 with gcc 3.4 (at
330 least up to 3.4.3). To work around it, define Py_HUGE_VAL as
331 HUGE_VAL(), e.g.:
332
333 make CPPFLAGS='-D"Py_HUGE_VAL=HUGE_VAL()" -I. -I$(srcdir)/Include'
Anthony Baxter70e07962006-04-03 14:16:27[diff] [blame]334 ./python setup.py CPPFLAGS='-D"Py_HUGE_VAL=HUGE_VAL()"'
Skip Montanaro9f6606a2005-11-17 18:19:39[diff] [blame]335
Guido van Rossum7ac11532000-08-29 21:36:40[diff] [blame]336Linux: A problem with threads and fork() was tracked down to a bug in
Anthony Baxter70e07962006-04-03 14:16:27[diff] [blame]337 the pthreads code in glibc version 2.0.5; glibc version 2.0.7
338 solves the problem. This causes the popen2 test to fail;
339 problem and solution reported by Pablo Bleyer.
Guido van Rossum4462e931997-01-22 21:00:32[diff] [blame]340
Martin v. Löwis3efa04a2003-06-14 05:29:27[diff] [blame]341Red Hat Linux: Red Hat 9 built Python2.2 in UCS-4 mode and hacked
Anthony Baxter70e07962006-04-03 14:16:27[diff] [blame]342 Tcl to support it. To compile Python2.3 with Tkinter, you will
343 need to pass --enable-unicode=ucs4 flag to ./configure.
Martin v. Löwis3efa04a2003-06-14 05:29:27[diff] [blame]344
Anthony Baxter70e07962006-04-03 14:16:27[diff] [blame]345 There's an executable /usr/bin/python which is Python
346 1.5.2 on most older Red Hat installations; several key Red Hat tools
347 require this version. Python 2.1.x may be installed as
348 /usr/bin/python2. The Makefile installs Python as
349 /usr/local/bin/python, which may or may not take precedence
350 over /usr/bin/python, depending on how you have set up $PATH.
Guido van Rossum2dcec0c2002-02-04 01:59:23[diff] [blame]351
Guido van Rossum59230501999-06-21 20:51:46[diff] [blame]352FreeBSD 3.x and probably platforms with NCurses that use libmytinfo or
Anthony Baxter70e07962006-04-03 14:16:27[diff] [blame]353 similar: When using cursesmodule, the linking is not done in
354 the correct order with the defaults. Remove "-ltermcap" from
355 the readline entry in Setup, and use as curses entry: "curses
356 cursesmodule.c -lmytinfo -lncurses -ltermcap" - "mytinfo" (so
357 called on FreeBSD) should be the name of the auxiliary library
358 required on your platform. Normally, it would be linked
359 automatically, but not necessarily in the correct order.
Guido van Rossum59230501999-06-21 20:51:46[diff] [blame]360
Anthony Baxter70e07962006-04-03 14:16:27[diff] [blame]361BSDI: BSDI versions before 4.1 have known problems with threads,
362 which can cause strange errors in a number of modules (for
363 instance, the 'test_signal' test script will hang forever.)
364 Turning off threads (with --with-threads=no) or upgrading to
365 BSDI 4.1 solves this problem.
Guido van Rossum7ac11532000-08-29 21:36:40[diff] [blame]366
367DEC Unix: Run configure with --with-dec-threads, or with
Anthony Baxter70e07962006-04-03 14:16:27[diff] [blame]368 --with-threads=no if no threads are desired (threads are on by
369 default). When using GCC, it is possible to get an internal
370 compiler error if optimization is used. This was reported for
371 GCC 2.7.2.3 on selectmodule.c. Manually compile the affected
372 file without optimization to solve the problem.
Guido van Rossum8eca2c21996-02-14 18:37:46[diff] [blame]373
Guido van Rossume1cdb161999-04-10 16:04:07[diff] [blame]374DEC Ultrix: compile with GCC to avoid bugs in the native compiler,
Anthony Baxter70e07962006-04-03 14:16:27[diff] [blame]375 and pass SHELL=/bin/sh5 to Make when installing.
Guido van Rossume1cdb161999-04-10 16:04:07[diff] [blame]376
Anthony Baxter70e07962006-04-03 14:16:27[diff] [blame]377AIX: A complete overhaul of the shared library support is now in
378 place. See Misc/AIX-NOTES for some notes on how it's done.
379 (The optimizer bug reported at this place in previous releases
380 has been worked around by a minimal code change.) If you get
381 errors about pthread_* functions, during compile or during
382 testing, try setting CC to a thread-safe (reentrant) compiler,
383 like "cc_r". For full C++ module support, set CC="xlC_r" (or
384 CC="xlC" without thread support).
Guido van Rossum91cb9d21995-04-10 11:47:38[diff] [blame]385
Guido van Rossuma2534c82005-09-20 19:06:23[diff] [blame]386AIX 5.3: To build a 64-bit version with IBM's compiler, I used the
387 following:
388
389 export PATH=/usr/bin:/usr/vacpp/bin
Anthony Baxter70e07962006-04-03 14:16:27[diff] [blame]390 ./configure --with-gcc="xlc_r -q64" --with-cxx="xlC_r -q64" \
Guido van Rossuma2534c82005-09-20 19:06:23[diff] [blame]391 --disable-ipv6 AR="ar -X64"
Anthony Baxter70e07962006-04-03 14:16:27[diff] [blame]392 make
Guido van Rossuma2534c82005-09-20 19:06:23[diff] [blame]393
Anthony Baxter7c497472003-10-01 07:24:54[diff] [blame]394HP-UX: When using threading, you may have to add -D_REENTRANT to the
Anthony Baxter70e07962006-04-03 14:16:27[diff] [blame]395 OPT variable in the top-level Makefile; reported by Pat Knight,
396 this seems to make a difference (at least for HP-UX 10.20)
397 even though pyconfig.h defines it. This seems unnecessary when
398 using HP/UX 11 and later - threading seems to work "out of the
399 box".
Anthony Baxter7c497472003-10-01 07:24:54[diff] [blame]400
Michael W. Hudson71dcc3e2005-02-22 15:33:26[diff] [blame]401HP-UX ia64: When building on the ia64 (Itanium) platform using HP's
Anthony Baxter70e07962006-04-03 14:16:27[diff] [blame]402 compiler, some experience has shown that the compiler's
403 optimiser produces a completely broken version of python
Georg Brandlf004d9d2009-10-27 15:39:53[diff] [blame]404 (see http://bugs.python.org/814976). To work around this,
Anthony Baxter70e07962006-04-03 14:16:27[diff] [blame]405 edit the Makefile and remove -O from the OPT line.
Guido van Rossum539c6622005-09-14 17:49:54[diff] [blame]406
Anthony Baxter70e07962006-04-03 14:16:27[diff] [blame]407 To build a 64-bit executable on an Itanium 2 system using HP's
408 compiler, use these environment variables:
Guido van Rossum539c6622005-09-14 17:49:54[diff] [blame]409
Anthony Baxter70e07962006-04-03 14:16:27[diff] [blame]410 CC=cc
411 CXX=aCC
412 BASECFLAGS="+DD64"
413 LDFLAGS="+DD64 -lxnet"
Guido van Rossum539c6622005-09-14 17:49:54[diff] [blame]414
Anthony Baxter70e07962006-04-03 14:16:27[diff] [blame]415 and call configure as:
Guido van Rossum539c6622005-09-14 17:49:54[diff] [blame]416
Anthony Baxter70e07962006-04-03 14:16:27[diff] [blame]417 ./configure --without-gcc
Guido van Rossum539c6622005-09-14 17:49:54[diff] [blame]418
Anthony Baxter70e07962006-04-03 14:16:27[diff] [blame]419 then *unset* the environment variables again before running
420 make. (At least one of these flags causes the build to fail
421 if it remains set.) You still have to edit the Makefile and
422 remove -O from the OPT line.
Guido van Rossum2094e041998-05-14 15:47:35[diff] [blame]423
Georg Brandlf004d9d2009-10-27 15:39:53[diff] [blame]424HP PA-RISC 2.0: A recent bug report (http://bugs.python.org/546117)
Anthony Baxter70e07962006-04-03 14:16:27[diff] [blame]425 suggests that the C compiler in this 64-bit system has bugs
426 in the optimizer that break Python. Compiling without
427 optimization solves the problems.
Guido van Rossum4e1dd7d2002-04-23 13:06:07[diff] [blame]428
Anthony Baxter70e07962006-04-03 14:16:27[diff] [blame]429SCO: The following apply to SCO 3 only; Python builds out of the box
430 on SCO 5 (or so we've heard).
Guido van Rossuma16e2751998-04-13 20:14:05[diff] [blame]431
Anthony Baxter70e07962006-04-03 14:16:27[diff] [blame]432 1) Everything works much better if you add -U__STDC__ to the
433 defs. This is because all the SCO header files are broken.
434 Anything that isn't mentioned in the C standard is
435 conditionally excluded when __STDC__ is defined.
Guido van Rossum91cb9d21995-04-10 11:47:38[diff] [blame]436
Anthony Baxter70e07962006-04-03 14:16:27[diff] [blame]437 2) Due to the U.S. export restrictions, SCO broke the crypt
438 stuff out into a separate library, libcrypt_i.a so the LIBS
439 needed be set to:
Guido van Rossum91cb9d21995-04-10 11:47:38[diff] [blame]440
Anthony Baxter70e07962006-04-03 14:16:27[diff] [blame]441 LIBS=' -lsocket -lcrypt_i'
Guido van Rossum91cb9d21995-04-10 11:47:38[diff] [blame]442
Martin v. Löwis387c5472001-09-06 08:16:17[diff] [blame]443UnixWare: There are known bugs in the math library of the system, as well as
444 problems in the handling of threads (calling fork in one
445 thread may interrupt system calls in others). Therefore, test_math and
446 tests involving threads will fail until those problems are fixed.
447
Anthony Baxter70e07962006-04-03 14:16:27[diff] [blame]448QNX: Chris Herborth (chrish@qnx.com) writes:
449 configure works best if you use GNU bash; a port is available on
450 ftp.qnx.com in /usr/free. I used the following process to build,
451 test and install Python 1.5.x under QNX:
Guido van Rossumb7f454d1997-12-02 19:44:31[diff] [blame]452
Anthony Baxter70e07962006-04-03 14:16:27[diff] [blame]453 1) CONFIG_SHELL=/usr/local/bin/bash CC=cc RANLIB=: \
454 ./configure --verbose --without-gcc --with-libm=""
Guido van Rossumb7f454d1997-12-02 19:44:31[diff] [blame]455
Anthony Baxter70e07962006-04-03 14:16:27[diff] [blame]456 2) edit Modules/Setup to activate everything that makes sense for
457 your system... tested here at QNX with the following modules:
Guido van Rossumb7f454d1997-12-02 19:44:31[diff] [blame]458
Anthony Baxter70e07962006-04-03 14:16:27[diff] [blame]459 array, audioop, binascii, cPickle, cStringIO, cmath,
460 crypt, curses, errno, fcntl, gdbm, grp, imageop,
461 _locale, math, md5, new, operator, parser, pcre,
Neal Norwitz90340a12007-05-20 18:06:27[diff] [blame]462 posix, pwd, readline, regex, reop,
Anthony Baxter70e07962006-04-03 14:16:27[diff] [blame]463 select, signal, socket, soundex, strop, struct,
Brett Cannondc48b742007-05-20 07:09:50[diff] [blame]464 syslog, termios, time, timing, zlib, audioop, imageop
Guido van Rossumb7f454d1997-12-02 19:44:31[diff] [blame]465
Anthony Baxter70e07962006-04-03 14:16:27[diff] [blame]466 3) make SHELL=/usr/local/bin/bash
Guido van Rossum16e809e2001-04-16 02:07:08[diff] [blame]467
Anthony Baxter70e07962006-04-03 14:16:27[diff] [blame]468 or, if you feel the need for speed:
Guido van Rossum16e809e2001-04-16 02:07:08[diff] [blame]469
Anthony Baxter70e07962006-04-03 14:16:27[diff] [blame]470 make SHELL=/usr/local/bin/bash OPT="-5 -Oil+nrt"
Guido van Rossumb7f454d1997-12-02 19:44:31[diff] [blame]471
Anthony Baxter70e07962006-04-03 14:16:27[diff] [blame]472 4) make SHELL=/usr/local/bin/bash test
Guido van Rossumec95c7b1998-08-04 17:59:56[diff] [blame]473
Anthony Baxter70e07962006-04-03 14:16:27[diff] [blame]474 Using GNU readline 2.2 seems to behave strangely, but I
475 think that's a problem with my readline 2.2 port. :-\
Guido van Rossumec95c7b1998-08-04 17:59:56[diff] [blame]476
Anthony Baxter70e07962006-04-03 14:16:27[diff] [blame]477 5) make SHELL=/usr/local/bin/bash install
Guido van Rossumb7f454d1997-12-02 19:44:31[diff] [blame]478
Anthony Baxter70e07962006-04-03 14:16:27[diff] [blame]479 If you get SIGSEGVs while running Python (I haven't yet, but
480 I've only run small programs and the test cases), you're
481 probably running out of stack; the default 32k could be a
482 little tight. To increase the stack size, edit the Makefile
483 to read: LDFLAGS = -N 48k
Guido van Rossum40d63581997-08-14 19:45:30[diff] [blame]484
Anthony Baxter70e07962006-04-03 14:16:27[diff] [blame]485BeOS: See Misc/BeOS-NOTES for notes about compiling/installing
486 Python on BeOS R3 or later. Note that only the PowerPC
487 platform is supported for R3; both PowerPC and x86 are
488 supported for R4.
Guido van Rossumec95c7b1998-08-04 17:59:56[diff] [blame]489
Michael W. Hudson33876f52002-07-31 09:55:25[diff] [blame]490Cray T3E: Mark Hadfield (m.hadfield@niwa.co.nz) writes:
Anthony Baxter70e07962006-04-03 14:16:27[diff] [blame]491 Python can be built satisfactorily on a Cray T3E but based on
492 my experience with the NIWA T3E (2002-05-22, version 2.2.1)
493 there are a few bugs and gotchas. For more information see a
494 thread on comp.lang.python in May 2002 entitled "Building
495 Python on Cray T3E".
Michael W. Hudson33876f52002-07-31 09:55:25[diff] [blame]496
497 1) Use Cray's cc and not gcc. The latter was reported not to
498 work by Konrad Hinsen. It may work now, but it may not.
499
500 2) To set sys.platform to something sensible, pass the
501 following environment variable to the configure script:
502
503 MACHDEP=unicosmk
504
Anthony Baxter70e07962006-04-03 14:16:27[diff] [blame]505 2) Run configure with option "--enable-unicode=ucs4".
Michael W. Hudson33876f52002-07-31 09:55:25[diff] [blame]506
Anthony Baxter70e07962006-04-03 14:16:27[diff] [blame]507 3) The Cray T3E does not support dynamic linking, so extension
508 modules have to be built by adding (or uncommenting) lines
509 in Modules/Setup. The minimum set of modules is
Michael W. Hudson33876f52002-07-31 09:55:25[diff] [blame]510
Anthony Baxter70e07962006-04-03 14:16:27[diff] [blame]511 posix, new, _sre, unicodedata
Michael W. Hudson33876f52002-07-31 09:55:25[diff] [blame]512
Anthony Baxter70e07962006-04-03 14:16:27[diff] [blame]513 On NIWA's vanilla T3E system the following have also been
514 included successfully:
Michael W. Hudson33876f52002-07-31 09:55:25[diff] [blame]515
Anthony Baxter70e07962006-04-03 14:16:27[diff] [blame]516 _codecs, _locale, _socket, _symtable, _testcapi, _weakref
517 array, binascii, cmath, cPickle, crypt, cStringIO, dbm
518 errno, fcntl, grp, math, md5, operator, parser, pcre, pwd
519 regex, rotor, select, struct, strop, syslog, termios
520 time, timing, xreadlines
Michael W. Hudson33876f52002-07-31 09:55:25[diff] [blame]521
Anthony Baxter70e07962006-04-03 14:16:27[diff] [blame]522 4) Once the python executable and library have been built, make
523 will execute setup.py, which will attempt to build remaining
524 extensions and link them dynamically. Each of these attempts
525 will fail but should not halt the make process. This is
526 normal.
Michael W. Hudson33876f52002-07-31 09:55:25[diff] [blame]527
Anthony Baxter70e07962006-04-03 14:16:27[diff] [blame]528 5) Running "make test" uses a lot of resources and causes
529 problems on our system. You might want to try running tests
530 singly or in small groups.
Guido van Rossum1bf0bf41997-08-20 23:50:51[diff] [blame]531
Anthony Baxter70e07962006-04-03 14:16:27[diff] [blame]532SGI: SGI's standard "make" utility (/bin/make or /usr/bin/make)
533 does not check whether a command actually changed the file it
534 is supposed to build. This means that whenever you say "make"
535 it will redo the link step. The remedy is to use SGI's much
536 smarter "smake" utility (/usr/sbin/smake), or GNU make. If
537 you set the first line of the Makefile to #!/usr/sbin/smake
538 smake will be invoked by make (likewise for GNU make).
Guido van Rossum0078aaf1997-08-21 03:05:11[diff] [blame]539
Anthony Baxter70e07962006-04-03 14:16:27[diff] [blame]540 WARNING: There are bugs in the optimizer of some versions of
541 SGI's compilers that can cause bus errors or other strange
542 behavior, especially on numerical operations. To avoid this,
543 try building with "make OPT=".
Fred Drake0b5fb2b2000-09-29 17:45:05[diff] [blame]544
Guido van Rossum64773801997-11-26 17:07:02[diff] [blame]545OS/2: If you are running Warp3 or Warp4 and have IBM's VisualAge C/C++
546 compiler installed, just change into the pc\os2vacpp directory
547 and type NMAKE. Threading and sockets are supported by default
548 in the resulting binaries of PYTHON15.DLL and PYTHON.EXE.
549
Fred Drake36ed5602000-10-06 01:58:48[diff] [blame]550Monterey (64-bit AIX): The current Monterey C compiler (Visual Age)
551 uses the OBJECT_MODE={32|64} environment variable to set the
552 compilation mode to either 32-bit or 64-bit (32-bit mode is
553 the default). Presumably you want 64-bit compilation mode for
554 this 64-bit OS. As a result you must first set OBJECT_MODE=64
555 in your environment before configuring (./configure) or
556 building (make) Python on Monterey.
557
558Reliant UNIX: The thread support does not compile on Reliant UNIX, and
559 there is a (minor) problem in the configure script for that
560 platform as well. This should be resolved in time for a
561 future release.
Trent Mick635f6fb2000-08-23 21:33:05[diff] [blame]562
Barry Warsaw6a8557d2002-10-14 18:04:39[diff] [blame]563MacOSX: The tests will crash on both 10.1 and 10.2 with SEGV in
Barry Warsaw6e753642002-10-14 18:15:35[diff] [blame]564 test_re and test_sre due to the small default stack size. If
565 you set the stack size to 2048 before doing a "make test" the
Georg Brandl667eb7c2007-03-06 18:47:31[diff] [blame]566 failure can be avoided. If you're using the tcsh or csh shells,
567 use "limit stacksize 2048" and for the bash shell (the default
568 as of OSX 10.3), use "ulimit -s 2048".
Barry Warsaw06557452001-12-14 20:47:12[diff] [blame]569
Jack Jansen6a77cf12001-12-06 21:49:02[diff] [blame]570 On naked Darwin you may want to add the configure option
Jack Jansen4c398fd2001-10-08 13:21:15[diff] [blame]571 "--disable-toolbox-glue" to disable the glue code for the Carbon
Jack Jansen6a77cf12001-12-06 21:49:02[diff] [blame]572 interface modules. The modules themselves are currently only built
573 if you add the --enable-framework option, see below.
Barry Warsaw06557452001-12-14 20:47:12[diff] [blame]574
575 On a clean OSX /usr/local does not exist. Do a
Jack Jansen61fec302002-01-04 15:59:57[diff] [blame]576 "sudo mkdir -m 775 /usr/local"
Jack Jansen84a01642002-08-04 21:19:55[diff] [blame]577 before you do a make install. It is probably not a good idea to
578 do "sudo make install" which installs everything as superuser,
579 as this may later cause problems when installing distutils-based
580 additions.
Michael W. Hudson71dcc3e2005-02-22 15:33:26[diff] [blame]581
Jack Jansen84a01642002-08-04 21:19:55[diff] [blame]582 Some people have reported problems building Python after using "fink"
Anthony Baxter70e07962006-04-03 14:16:27[diff] [blame]583 to install additional unix software. Disabling fink (remove all
584 references to /sw from your .profile or .login) should solve this.
Barry Warsaw06557452001-12-14 20:47:12[diff] [blame]585
586 You may want to try the configure option "--enable-framework"
587 which installs Python as a framework. The location can be set
588 as argument to the --enable-framework option (default
Jack Jansen84a01642002-08-04 21:19:55[diff] [blame]589 /Library/Frameworks). A framework install is probably needed if you
590 want to use any Aqua-based GUI toolkit (whether Tkinter, wxPython,
591 Carbon, Cocoa or anything else).
Michael W. Hudson71dcc3e2005-02-22 15:33:26[diff] [blame]592
Georg Brandlb0885b72007-12-09 08:59:45[diff] [blame]593 You may also want to try the configure option "--enable-universalsdk"
594 which builds Python as a universal binary with support for the
595 i386 and PPC architetures. This requires Xcode 2.1 or later to build.
Ronald Oussoren988117f2006-04-29 11:31:35[diff] [blame]596
Georg Brandl3fc792b2007-12-09 08:54:02[diff] [blame]597 See Mac/README for more information on framework and
Georg Brandlb0885b72007-12-09 08:59:45[diff] [blame]598 universal builds.
Fred Drakedabed752001-02-01 19:41:13[diff] [blame]599
Michael W. Hudson1c4523f2001-12-19 19:49:58[diff] [blame]600Cygwin: With recent (relative to the time of writing, 2001-12-19)
601 Cygwin installations, there are problems with the interaction
602 of dynamic linking and fork(). This manifests itself in build
603 failures during the execution of setup.py.
Tim Peters42fc7cc2001-04-04 18:35:19[diff] [blame]604
Michael W. Hudson1c4523f2001-12-19 19:49:58[diff] [blame]605 There are two workarounds that both enable Python (albeit
606 without threading support) to build and pass all tests on
607 NT/2000 (and most likely XP as well, though reports of testing
608 on XP would be appreciated).
Guido van Rossum16e809e2001-04-16 02:07:08[diff] [blame]609
Michael W. Hudson1c4523f2001-12-19 19:49:58[diff] [blame]610 The workarounds:
611
612 (a) the band-aid fix is to link the _socket module statically
613 rather than dynamically (which is the default).
614
615 To do this, run "./configure --with-threads=no" including any
616 other options you need (--prefix, etc.). Then in Modules/Setup
Walter Dörwald7dc65d62003-02-02 23:37:05[diff] [blame]617 uncomment the lines:
Tim Peters42fc7cc2001-04-04 18:35:19[diff] [blame]618
Michael W. Hudsonbe5ce182001-12-12 11:46:52[diff] [blame]619 #SSL=/usr/local/ssl
620 #_socket socketmodule.c \
Anthony Baxter70e07962006-04-03 14:16:27[diff] [blame]621 # -DUSE_SSL -I$(SSL)/include -I$(SSL)/include/openssl \
622 # -L$(SSL)/lib -lssl -lcrypto
Guido van Rossum16e809e2001-04-16 02:07:08[diff] [blame]623
Michael W. Hudson1c4523f2001-12-19 19:49:58[diff] [blame]624 and remove "local/" from the SSL variable. Finally, just run
625 "make"!
Tim Peters42fc7cc2001-04-04 18:35:19[diff] [blame]626
Michael W. Hudsonb6554642001-12-19 22:09:09[diff] [blame]627 (b) The "proper" fix is to rebase the Cygwin DLLs to prevent
Michael W. Hudson1c4523f2001-12-19 19:49:58[diff] [blame]628 base address conflicts. Details on how to do this can be
629 found in the following mail:
630
631 http://sources.redhat.com/ml/cygwin/2001-12/msg00894.html
632
633 It is hoped that a version of this solution will be
Michael W. Hudsonb6554642001-12-19 22:09:09[diff] [blame]634 incorporated into the Cygwin distribution fairly soon.
Michael W. Hudson1c4523f2001-12-19 19:49:58[diff] [blame]635
636 Two additional problems:
637
638 (1) Threading support should still be disabled due to a known
639 bug in Cygwin pthreads that causes test_threadedtempfile to
640 hang.
641
Neal Norwitz4ebde092001-12-19 20:44:13[diff] [blame]642 (2) The _curses module does not build. This is a known
Michael W. Hudson1c4523f2001-12-19 19:49:58[diff] [blame]643 Cygwin ncurses problem that should be resolved the next time
644 that this package is released.
Guido van Rossum16e809e2001-04-16 02:07:08[diff] [blame]645
Michael W. Hudsonbe5ce182001-12-12 11:46:52[diff] [blame]646 On older versions of Cygwin, test_poll may hang and test_strftime
647 may fail.
Tim Peters42fc7cc2001-04-04 18:35:19[diff] [blame]648
Michael W. Hudson1c4523f2001-12-19 19:49:58[diff] [blame]649 The situation on 9X/Me is not accurately known at present.
650 Some time ago, there were reports that the following
651 regression tests failed:
Tim Peters42fc7cc2001-04-04 18:35:19[diff] [blame]652
653 test_pwd
654 test_select (hang)
655 test_socket
656
Michael W. Hudson1c4523f2001-12-19 19:49:58[diff] [blame]657 Due to the test_select hang on 9X/Me, one should run the
658 regression test using the following:
Tim Peters42fc7cc2001-04-04 18:35:19[diff] [blame]659
Michael W. Hudsonbe5ce182001-12-12 11:46:52[diff] [blame]660 make TESTOPTS='-l -x test_select' test
Tim Peters42fc7cc2001-04-04 18:35:19[diff] [blame]661
Michael W. Hudson1c4523f2001-12-19 19:49:58[diff] [blame]662 News regarding these platforms with more recent Cygwin
663 versions would be appreciated!
Guido van Rossum6d9cc801996-08-01 17:31:22[diff] [blame]664
Georg Brandl9b08e052009-04-05 21:21:05[diff] [blame]665Windows: When executing Python scripts on the command line using file type
666 associations (i.e. starting "script.py" instead of "python script.py"),
667 redirects may not work unless you set a specific registry key. See
668 the Knowledge Base article <http://support.microsoft.com/kb/321788>.
Skip Montanaro56d83012003-01-04 04:05:51[diff] [blame]669
670
Skip Montanaro57454e52002-06-14 20:30:31[diff] [blame]671Configuring the bsddb and dbm modules
672-------------------------------------
673
Skip Montanaro56d83012003-01-04 04:05:51[diff] [blame]674Beginning with Python version 2.3, the PyBsddb package
675<http://pybsddb.sf.net/> was adopted into Python as the bsddb package,
676exposing a set of package-level functions which provide
Gregory P. Smith3adc4aa2006-04-13 19:19:01[diff] [blame]677backwards-compatible behavior. Only versions 3.3 through 4.4 of
Skip Montanaro56d83012003-01-04 04:05:51[diff] [blame]678Sleepycat's libraries provide the necessary API, so older versions
679aren't supported through this interface. The old bsddb module has
680been retained as bsddb185, though it is not built by default. Users
681wishing to use it will have to tweak Modules/Setup to build it. The
682dbm module will still be built against the Sleepycat libraries if
Gregory P. Smith3adc4aa2006-04-13 19:19:01[diff] [blame]683other preferred alternatives (ndbm, gdbm) are not found.
Guido van Rossum62d45c02002-12-02 15:49:13[diff] [blame]684
Anthony Baxter70e07962006-04-03 14:16:27[diff] [blame]685Building the sqlite3 module
686---------------------------
687
688To build the sqlite3 module, you'll need the sqlite3 or libsqlite3
689packages installed, including the header files. Many modern operating
690systems distribute the headers in a separate package to the library -
691often it will be the same name as the main package, but with a -dev or
692-devel suffix.
693
694The version of pysqlite2 that's including in Python needs sqlite3 3.0.8
695or later. setup.py attempts to check that it can find a correct version.
Skip Montanaro57454e52002-06-14 20:30:31[diff] [blame]696
Guido van Rossumdc8a3cb1998-05-12 15:29:18[diff] [blame]697Configuring threads
698-------------------
699
Guido van Rossum7ac11532000-08-29 21:36:40[diff] [blame]700As of Python 2.0, threads are enabled by default. If you wish to
701compile without threads, or if your thread support is broken, pass the
702--with-threads=no switch to configure. Unfortunately, on some
703platforms, additional compiler and/or linker options are required for
704threads to work properly. Below is a table of those options,
Guido van Rossum4f40c2d2000-09-05 16:46:22[diff] [blame]705collected by Bill Janssen. We would love to automate this process
Guido van Rossum7ac11532000-08-29 21:36:40[diff] [blame]706more, but the information below is not enough to write a patch for the
707configure.in file, so manual intervention is required. If you patch
708the configure.in file and are confident that the patch works, please
Guido van Rossum4f40c2d2000-09-05 16:46:22[diff] [blame]709send in the patch. (Don't bother patching the configure script itself
Walter Dörwald7dc65d62003-02-02 23:37:05[diff] [blame]710-- it is regenerated each time the configure.in file changes.)
Guido van Rossumdc8a3cb1998-05-12 15:29:18[diff] [blame]711
712Compiler switches for threads
713.............................
714
Martin v. Löwis1c24f9c2001-03-06 12:07:27[diff] [blame]715The definition of _REENTRANT should be configured automatically, if
716that does not work on your system, or if _REENTRANT is defined
717incorrectly, please report that as a bug.
Guido van Rossumdc8a3cb1998-05-12 15:29:18[diff] [blame]718
Martin v. Löwis1c24f9c2001-03-06 12:07:27[diff] [blame]719 OS/Compiler/threads Switches for use with threads
720 (POSIX is draft 10, DCE is draft 4) compile & link
721
722 SunOS 5.{1-5}/{gcc,SunPro cc}/solaris -mt
723 SunOS 5.5/{gcc,SunPro cc}/POSIX (nothing)
Guido van Rossum16e809e2001-04-16 02:07:08[diff] [blame]724 DEC OSF/1 3.x/cc/DCE -threads
Anthony Baxter70e07962006-04-03 14:16:27[diff] [blame]725 (butenhof@zko.dec.com)
Guido van Rossum16e809e2001-04-16 02:07:08[diff] [blame]726 Digital UNIX 4.x/cc/DCE -threads
Anthony Baxter70e07962006-04-03 14:16:27[diff] [blame]727 (butenhof@zko.dec.com)
Guido van Rossum16e809e2001-04-16 02:07:08[diff] [blame]728 Digital UNIX 4.x/cc/POSIX -pthread
Anthony Baxter70e07962006-04-03 14:16:27[diff] [blame]729 (butenhof@zko.dec.com)
Guido van Rossum16e809e2001-04-16 02:07:08[diff] [blame]730 AIX 4.1.4/cc_r/d7 (nothing)
Anthony Baxter70e07962006-04-03 14:16:27[diff] [blame]731 (buhrt@iquest.net)
Guido van Rossum16e809e2001-04-16 02:07:08[diff] [blame]732 AIX 4.1.4/cc_r4/DCE (nothing)
Anthony Baxter70e07962006-04-03 14:16:27[diff] [blame]733 (buhrt@iquest.net)
Guido van Rossum16e809e2001-04-16 02:07:08[diff] [blame]734 IRIX 6.2/cc/POSIX (nothing)
Anthony Baxter70e07962006-04-03 14:16:27[diff] [blame]735 (robertl@cwi.nl)
Guido van Rossumdc8a3cb1998-05-12 15:29:18[diff] [blame]736
737
738Linker (ld) libraries and flags for threads
739...........................................
740
741 OS/threads Libraries/switches for use with threads
742
743 SunOS 5.{1-5}/solaris -lthread
744 SunOS 5.5/POSIX -lpthread
Guido van Rossum16e809e2001-04-16 02:07:08[diff] [blame]745 DEC OSF/1 3.x/DCE -lpthreads -lmach -lc_r -lc
Anthony Baxter70e07962006-04-03 14:16:27[diff] [blame]746 (butenhof@zko.dec.com)
Guido van Rossumdc8a3cb1998-05-12 15:29:18[diff] [blame]747 Digital UNIX 4.x/DCE -lpthreads -lpthread -lmach -lexc -lc
Anthony Baxter70e07962006-04-03 14:16:27[diff] [blame]748 (butenhof@zko.dec.com)
Guido van Rossum16e809e2001-04-16 02:07:08[diff] [blame]749 Digital UNIX 4.x/POSIX -lpthread -lmach -lexc -lc
Anthony Baxter70e07962006-04-03 14:16:27[diff] [blame]750 (butenhof@zko.dec.com)
Guido van Rossum16e809e2001-04-16 02:07:08[diff] [blame]751 AIX 4.1.4/{draft7,DCE} (nothing)
Anthony Baxter70e07962006-04-03 14:16:27[diff] [blame]752 (buhrt@iquest.net)
Guido van Rossum16e809e2001-04-16 02:07:08[diff] [blame]753 IRIX 6.2/POSIX -lpthread
Anthony Baxter70e07962006-04-03 14:16:27[diff] [blame]754 (jph@emilia.engr.sgi.com)
Guido van Rossumdc8a3cb1998-05-12 15:29:18[diff] [blame]755
756
Martin v. Löwis1142de32002-03-29 16:28:31[diff] [blame]757Building a shared libpython
758---------------------------
759
760Starting with Python 2.3, the majority of the interpreter can be built
Michael W. Hudson71dcc3e2005-02-22 15:33:26[diff] [blame]761into a shared library, which can then be used by the interpreter
Martin v. Löwis1142de32002-03-29 16:28:31[diff] [blame]762executable, and by applications embedding Python. To enable this feature,
Michael W. Hudson71dcc3e2005-02-22 15:33:26[diff] [blame]763configure with --enable-shared.
Martin v. Löwis65069672002-08-03 21:38:27[diff] [blame]764
Fred Drake55512dc2003-07-21 16:01:39[diff] [blame]765If you enable this feature, the same object files will be used to create
766a static library. In particular, the static library will contain object
767files using position-independent code (PIC) on platforms where PIC flags
768are needed for the shared library.
Martin v. Löwis1142de32002-03-29 16:28:31[diff] [blame]769
770
Guido van Rossum6d9cc801996-08-01 17:31:22[diff] [blame]771Configuring additional built-in modules
Guido van Rossum91cb9d21995-04-10 11:47:38[diff] [blame]772---------------------------------------
Guido van Rossum19e0c261995-01-17 16:36:34[diff] [blame]773
Andrew M. Kuchlingebe30452001-02-27 03:24:24[diff] [blame]774Starting with Python 2.1, the setup.py script at the top of the source
775distribution attempts to detect which modules can be built and
776automatically compiles them. Autodetection doesn't always work, so
Guido van Rossum59b0b602002-02-08 22:22:35[diff] [blame]777you can still customize the configuration by editing the Modules/Setup
778file; but this should be considered a last resort. The rest of this
779section only applies if you decide to edit the Modules/Setup file.
780You also need this to enable static linking of certain modules (which
781is needed to enable profiling on some systems).
782
Andrew M. Kuchlingebe30452001-02-27 03:24:24[diff] [blame]783This file is initially copied from Setup.dist by the configure script;
784if it does not exist yet, create it by copying Modules/Setup.dist
785yourself (configure will never overwrite it). Never edit Setup.dist
786-- always edit Setup or Setup.local (see below). Read the comments in
787the file for information on what kind of edits are allowed. When you
788have edited Setup in the Modules directory, the interpreter will
Guido van Rossumef67ded2001-10-05 17:04:42[diff] [blame]789automatically be rebuilt the next time you run make (in the toplevel
790directory).
Guido van Rossum627b2d71993-12-24 10:39:16[diff] [blame]791
Andrew M. Kuchlingebe30452001-02-27 03:24:24[diff] [blame]792Many useful modules can be built on any Unix system, but some optional
793modules can't be reliably autodetected. Often the quickest way to
794determine whether a particular module works or not is to see if it
795will build: enable it in Setup, then if you get compilation or link
Fred Drake6caae142000-10-25 17:51:02[diff] [blame]796errors, disable it -- you're either missing support or need to adjust
797the compilation and linking parameters for that module.
Guido van Rossum6d9cc801996-08-01 17:31:22[diff] [blame]798
799On SGI IRIX, there are modules that interface to many SGI specific
Andrew M. Kuchlingebe30452001-02-27 03:24:24[diff] [blame]800system libraries, e.g. the GL library and the audio hardware. These
801modules will not be built by the setup.py script.
Guido van Rossum91cb9d21995-04-10 11:47:38[diff] [blame]802
Guido van Rossum4462e931997-01-22 21:00:32[diff] [blame]803In addition to the file Setup, you can also edit the file Setup.local.
804(the makesetup script processes both). You may find it more
805convenient to edit Setup.local and leave Setup alone. Then, when
806installing a new Python version, you can copy your old Setup.local
807file.
808
Guido van Rossum91cb9d21995-04-10 11:47:38[diff] [blame]809
810Setting the optimization/debugging options
811------------------------------------------
812
Guido van Rossum6d9cc801996-08-01 17:31:22[diff] [blame]813If you want or need to change the optimization/debugging options for
814the C compiler, assign to the OPT variable on the toplevel make
815command; e.g. "make OPT=-g" will build a debugging version of Python
816on most platforms. The default is OPT=-O; a value for OPT in the
817environment when the configure script is run overrides this default
818(likewise for CC; and the initial value for LIBS is used as the base
819set of libraries to link with).
Guido van Rossum91cb9d21995-04-10 11:47:38[diff] [blame]820
Fred Drake6caae142000-10-25 17:51:02[diff] [blame]821When compiling with GCC, the default value of OPT will also include
822the -Wall and -Wstrict-prototypes options.
823
824Additional debugging code to help debug memory management problems can
825be enabled by using the --with-pydebug option to the configure script.
826
Brett Cannon08cd5982005-04-24 22:26:38[diff] [blame]827For flags that change binary compatibility, use the EXTRA_CFLAGS
828variable.
829
Guido van Rossum91cb9d21995-04-10 11:47:38[diff] [blame]830
Guido van Rossuma68ca352001-10-04 17:00:07[diff] [blame]831Profiling
832---------
833
834If you want C profiling turned on, the easiest way is to run configure
835with the CC environment variable to the necessary compiler
836invocation. For example, on Linux, this works for profiling using
837gprof(1):
838
839 CC="gcc -pg" ./configure
840
841Note that on Linux, gprof apparently does not work for shared
842libraries. The Makefile/Setup mechanism can be used to compile and
Walter Dörwald7dc65d62003-02-02 23:37:05[diff] [blame]843link most extension modules statically.
Guido van Rossuma68ca352001-10-04 17:00:07[diff] [blame]844
845
Georg Brandl405b5f32008-05-18 11:46:51[diff] [blame]846Coverage checking
847-----------------
848
849For C coverage checking using gcov, run "make coverage". This will
850build a Python binary with profiling activated, and a ".gcno" and
851".gcda" file for every source file compiled with that option. With
852the built binary, now run the code whose coverage you want to check.
853Then, you can see coverage statistics for each individual source file
854by running gcov, e.g.
855
856 gcov -o Modules zlibmodule
857
858This will create a "zlibmodule.c.gcov" file in the current directory
859containing coverage info for that source file.
860
861This works only for source files statically compiled into the
862executable; use the Makefile/Setup mechanism to compile and link
863extension modules you want to coverage-check statically.
864
865
Guido van Rossum91cb9d21995-04-10 11:47:38[diff] [blame]866Testing
867-------
Guido van Rossum433c8ad1994-08-01 12:07:07[diff] [blame]868
Guido van Rossum4f40c2d2000-09-05 16:46:22[diff] [blame]869To test the interpreter, type "make test" in the top-level directory.
Guido van Rossum6ae5d3d1997-05-14 21:39:05[diff] [blame]870This runs the test set twice (once with no compiled files, once with
871the compiled files left by the previous test run). The test set
872produces some output. You can generally ignore the messages about
Guido van Rossum59b0b602002-02-08 22:22:35[diff] [blame]873skipped tests due to optional features which can't be imported.
Guido van Rossum4f40c2d2000-09-05 16:46:22[diff] [blame]874If a message is printed about a failed test or a traceback or core
875dump is produced, something is wrong. On some Linux systems (those
Guido van Rossum24df6841997-12-30 04:32:30[diff] [blame]876that are not yet using glibc 6), test_strftime fails due to a
Guido van Rossum4f40c2d2000-09-05 16:46:22[diff] [blame]877non-standard implementation of strftime() in the C library. Please
878ignore this, or upgrade to glibc version 6.
Guido van Rossum433c8ad1994-08-01 12:07:07[diff] [blame]879
Georg Brandlb6ca2d62010-05-18 23:38:50[diff] [blame]880By default, tests are prevented from overusing resources like disk space and
881memory. To enable these tests, run "make testall".
882
Guido van Rossum91cb9d21995-04-10 11:47:38[diff] [blame]883IMPORTANT: If the tests fail and you decide to mail a bug report,
884*don't* include the output of "make test". It is useless. Run the
Guido van Rossum4f40c2d2000-09-05 16:46:22[diff] [blame]885failing test manually, as follows:
Guido van Rossum91cb9d21995-04-10 11:47:38[diff] [blame]886
Georg Brandlb6ca2d62010-05-18 23:38:50[diff] [blame]887 ./python Lib/test/regrtest.py -v test_whatever
Guido van Rossum91cb9d21995-04-10 11:47:38[diff] [blame]888
Andrew M. Kuchling83d042d2002-03-21 23:52:20[diff] [blame]889(substituting the top of the source tree for '.' if you built in a
Guido van Rossum6ae5d3d1997-05-14 21:39:05[diff] [blame]890different directory). This runs the test in verbose mode.
Guido van Rossum91cb9d21995-04-10 11:47:38[diff] [blame]891
892
893Installing
894----------
895
Guido van Rossum6d9cc801996-08-01 17:31:22[diff] [blame]896To install the Python binary, library modules, shared library modules
897(see below), include files, configuration files, and the manual page,
Guido van Rossum64773801997-11-26 17:07:02[diff] [blame]898just type
899
Anthony Baxter70e07962006-04-03 14:16:27[diff] [blame]900 make install
Guido van Rossum64773801997-11-26 17:07:02[diff] [blame]901
Guido van Rossum4f40c2d2000-09-05 16:46:22[diff] [blame]902This will install all platform-independent files in subdirectories of
903the directory given with the --prefix option to configure or to the
904`prefix' Make variable (default /usr/local). All binary and other
905platform-specific files will be installed in subdirectories if the
906directory given by --exec-prefix or the `exec_prefix' Make variable
907(defaults to the --prefix directory) is given.
Guido van Rossum64773801997-11-26 17:07:02[diff] [blame]908
Martin v. Löwis3b8ee082003-05-11 20:25:35[diff] [blame]909If DESTDIR is set, it will be taken as the root directory of the
910installation, and files will be installed into $(DESTDIR)$(prefix),
911$(DESTDIR)$(exec_prefix), etc.
912
Guido van Rossum64773801997-11-26 17:07:02[diff] [blame]913All subdirectories created will have Python's version number in their
914name, e.g. the library modules are installed in
Guido van Rossum67345eb2001-01-18 14:51:12[diff] [blame]915"/usr/local/lib/python<version>/" by default, where <version> is the
916<major>.<minor> release number (e.g. "2.1"). The Python binary is
917installed as "python<version>" and a hard link named "python" is
918created. The only file not installed with a version number in its
919name is the manual page, installed as "/usr/local/man/man1/python.1"
920by default.
Guido van Rossum433c8ad1994-08-01 12:07:07[diff] [blame]921
Skip Montanaro86d4e7a2008-03-05 16:41:09[diff] [blame]922If you want to install multiple versions of Python see the section below
923entitled "Installing multiple versions".
Guido van Rossum6d9cc801996-08-01 17:31:22[diff] [blame]924
925The only thing you may have to install manually is the Python mode for
Guido van Rossum4f40c2d2000-09-05 16:46:22[diff] [blame]926Emacs found in Misc/python-mode.el. (But then again, more recent
927versions of Emacs may already have it.) Follow the instructions that
928came with Emacs for installation of site-specific files.
Guido van Rossum433c8ad1994-08-01 12:07:07[diff] [blame]929
Jack Jansena39ef862001-08-19 21:17:03[diff] [blame]930On Mac OS X, if you have configured Python with --enable-framework, you
931should use "make frameworkinstall" to do the installation. Note that this
932installs the Python executable in a place that is not normally on your
933PATH, you may want to set up a symlink in /usr/local/bin.
934
Guido van Rossum91cb9d21995-04-10 11:47:38[diff] [blame]935
Skip Montanaro86d4e7a2008-03-05 16:41:09[diff] [blame]936Installing multiple versions
937----------------------------
938
939On Unix and Mac systems if you intend to install multiple versions of Python
940using the same installation prefix (--prefix argument to the configure
941script) you must take care that your primary python executable is not
Ezio Melottic0fd6ff2010-03-23 13:20:39[diff] [blame]942overwritten by the installation of a different version. All files and
Skip Montanaro86d4e7a2008-03-05 16:41:09[diff] [blame]943directories installed using "make altinstall" contain the major and minor
944version and can thus live side-by-side. "make install" also creates
945${prefix}/bin/python which refers to ${prefix}/bin/pythonX.Y. If you intend
946to install multiple versions using the same prefix you must decide which
947version (if any) is your "primary" version. Install that version using
948"make install". Install all other versions using "make altinstall".
949
950For example, if you want to install Python 2.5, 2.6 and 3.0 with 2.6 being
951the primary version, you would execute "make install" in your 2.6 build
952directory and "make altinstall" in the others.
953
954
Guido van Rossum91cb9d21995-04-10 11:47:38[diff] [blame]955Configuration options and variables
956-----------------------------------
Guido van Rossum433c8ad1994-08-01 12:07:07[diff] [blame]957
Guido van Rossum6d9cc801996-08-01 17:31:22[diff] [blame]958Some special cases are handled by passing options to the configure
959script.
Guido van Rossum433c8ad1994-08-01 12:07:07[diff] [blame]960
Guido van Rossum6d9cc801996-08-01 17:31:22[diff] [blame]961WARNING: if you rerun the configure script with different options, you
962must run "make clean" before rebuilding. Exceptions to this rule:
963after changing --prefix or --exec-prefix, all you need to do is remove
Guido van Rossumb06df271997-08-05 21:50:20[diff] [blame]964Modules/getpath.o.
Guido van Rossum433c8ad1994-08-01 12:07:07[diff] [blame]965
Guido van Rossum91cb9d21995-04-10 11:47:38[diff] [blame]966--with(out)-gcc: The configure script uses gcc (the GNU C compiler) if
Anthony Baxter70e07962006-04-03 14:16:27[diff] [blame]967 it finds it. If you don't want this, or if this compiler is
968 installed but broken on your platform, pass the option
969 --without-gcc. You can also pass "CC=cc" (or whatever the
970 name of the proper C compiler is) in the environment, but the
971 advantage of using --without-gcc is that this option is
972 remembered by the config.status script for its --recheck
973 option.
Guido van Rossum76be6ed1995-01-02 18:33:54[diff] [blame]974
Guido van Rossum91cb9d21995-04-10 11:47:38[diff] [blame]975--prefix, --exec-prefix: If you want to install the binaries and the
Anthony Baxter70e07962006-04-03 14:16:27[diff] [blame]976 Python library somewhere else than in /usr/local/{bin,lib},
977 you can pass the option --prefix=DIRECTORY; the interpreter
978 binary will be installed as DIRECTORY/bin/python and the
979 library files as DIRECTORY/lib/python/*. If you pass
980 --exec-prefix=DIRECTORY (as well) this overrides the
981 installation prefix for architecture-dependent files (like the
982 interpreter binary). Note that --prefix=DIRECTORY also
983 affects the default module search path (sys.path), when
984 Modules/config.c is compiled. Passing make the option
985 prefix=DIRECTORY (and/or exec_prefix=DIRECTORY) overrides the
986 prefix set at configuration time; this may be more convenient
987 than re-running the configure script if you change your mind
988 about the install prefix.
Guido van Rossum76be6ed1995-01-02 18:33:54[diff] [blame]989
Guido van Rossum59b0b602002-02-08 22:22:35[diff] [blame]990--with-readline: This option is no longer supported. GNU
Anthony Baxter70e07962006-04-03 14:16:27[diff] [blame]991 readline is automatically enabled by setup.py when present.
Guido van Rossum91cb9d21995-04-10 11:47:38[diff] [blame]992
Guido van Rossum7ac11532000-08-29 21:36:40[diff] [blame]993--with-threads: On most Unix systems, you can now use multiple
Anthony Baxter70e07962006-04-03 14:16:27[diff] [blame]994 threads, and support for this is enabled by default. To
995 disable this, pass --with-threads=no. If the library required
996 for threads lives in a peculiar place, you can use
997 --with-thread=DIRECTORY. IMPORTANT: run "make clean" after
998 changing (either enabling or disabling) this option, or you
999 will get link errors! Note: for DEC Unix use
1000 --with-dec-threads instead.
Guido van Rossum91cb9d21995-04-10 11:47:38[diff] [blame]1001
1002--with-sgi-dl: On SGI IRIX 4, dynamic loading of extension modules is
Anthony Baxter70e07962006-04-03 14:16:27[diff] [blame]1003 supported by the "dl" library by Jack Jansen, which is
1004 ftp'able from ftp://ftp.cwi.nl/pub/dynload/dl-1.6.tar.Z.
1005 This is enabled (after you've ftp'ed and compiled the dl
1006 library) by passing --with-sgi-dl=DIRECTORY where DIRECTORY
1007 is the absolute pathname of the dl library. (Don't bother on
1008 IRIX 5, it already has dynamic linking using SunOS style
1009 shared libraries.) THIS OPTION IS UNSUPPORTED.
Guido van Rossum91cb9d21995-04-10 11:47:38[diff] [blame]1010
Guido van Rossum4f40c2d2000-09-05 16:46:22[diff] [blame]1011--with-dl-dld: Dynamic loading of modules is rumored to be supported
Anthony Baxter70e07962006-04-03 14:16:27[diff] [blame]1012 on some other systems: VAX (Ultrix), Sun3 (SunOS 3.4), Sequent
1013 Symmetry (Dynix), and Atari ST. This is done using a
1014 combination of the GNU dynamic loading package
1015 (ftp://ftp.cwi.nl/pub/dynload/dl-dld-1.1.tar.Z) and an
1016 emulation of the SGI dl library mentioned above (the emulation
1017 can be found at
1018 ftp://ftp.cwi.nl/pub/dynload/dld-3.2.3.tar.Z). To
1019 enable this, ftp and compile both libraries, then call
1020 configure, passing it the option
1021 --with-dl-dld=DL_DIRECTORY,DLD_DIRECTORY where DL_DIRECTORY is
1022 the absolute pathname of the dl emulation library and
1023 DLD_DIRECTORY is the absolute pathname of the GNU dld library.
1024 (Don't bother on SunOS 4 or 5, they already have dynamic
1025 linking using shared libraries.) THIS OPTION IS UNSUPPORTED.
Guido van Rossum91cb9d21995-04-10 11:47:38[diff] [blame]1026
1027--with-libm, --with-libc: It is possible to specify alternative
Anthony Baxter70e07962006-04-03 14:16:27[diff] [blame]1028 versions for the Math library (default -lm) and the C library
1029 (default the empty string) using the options
1030 --with-libm=STRING and --with-libc=STRING, respectively. For
1031 example, if your system requires that you pass -lc_s to the C
1032 compiler to use the shared C library, you can pass
1033 --with-libc=-lc_s. These libraries are passed after all other
1034 libraries, the C library last.
Guido van Rossum16e809e2001-04-16 02:07:08[diff] [blame]1035
Guido van Rossum4f40c2d2000-09-05 16:46:22[diff] [blame]1036--with-libs='libs': Add 'libs' to the LIBS that the python interpreter
Anthony Baxter70e07962006-04-03 14:16:27[diff] [blame]1037 is linked against.
Guido van Rossumd02ba451996-07-31 17:36:01[diff] [blame]1038
Martin v. Löwis0f48d982006-04-14 14:34:26[diff] [blame]1039--with-cxx-main=<compiler>: If you plan to use C++ extension modules,
1040 then -- on some platforms -- you need to compile python's main()
1041 function with the C++ compiler. With this option, make will use
1042 <compiler> to compile main() *and* to link the python executable.
1043 It is likely that the resulting executable depends on the C++
1044 runtime library of <compiler>. (The default is --without-cxx-main.)
1045
1046 There are platforms that do not require you to build Python
1047 with a C++ compiler in order to use C++ extension modules.
1048 E.g., x86 Linux with ELF shared binaries and GCC 3.x, 4.x is such
1049 a platform. We recommend that you configure Python
1050 --without-cxx-main on those platforms because a mismatch
1051 between the C++ compiler version used to build Python and to
1052 build a C++ extension module is likely to cause a crash at
1053 runtime.
1054
1055 The Python installation also stores the variable CXX that
1056 determines, e.g., the C++ compiler distutils calls by default
1057 to build C++ extensions. If you set CXX on the configure command
1058 line to any string of non-zero length, then configure won't
1059 change CXX. If you do not preset CXX but pass
1060 --with-cxx-main=<compiler>, then configure sets CXX=<compiler>.
1061 In all other cases, configure looks for a C++ compiler by
1062 some common names (c++, g++, gcc, CC, cxx, cc++, cl) and sets
1063 CXX to the first compiler it finds. If it does not find any
1064 C++ compiler, then it sets CXX="".
1065
1066 Similarly, if you want to change the command used to link the
1067 python executable, then set LINKCC on the configure command line.
Martin v. Löwis2df66942000-12-13 14:14:32[diff] [blame]1068
1069
Fred Drake6caae142000-10-25 17:51:02[diff] [blame]1070--with-pydebug: Enable additional debugging code to help track down
Anthony Baxter70e07962006-04-03 14:16:27[diff] [blame]1071 memory management problems. This allows printing a list of all
1072 live objects when the interpreter terminates.
Michael W. Hudson71dcc3e2005-02-22 15:33:26[diff] [blame]1073
Jack Jansen7b8c7542002-04-14 20:12:41[diff] [blame]1074--with(out)-universal-newlines: enable reading of text files with
Anthony Baxter70e07962006-04-03 14:16:27[diff] [blame]1075 foreign newline convention (default: enabled). In other words,
1076 any of \r, \n or \r\n is acceptable as end-of-line character.
1077 If enabled import and execfile will automatically accept any newline
1078 in files. Python code can open a file with open(file, 'U') to
1079 read it in universal newline mode. THIS OPTION IS UNSUPPORTED.
Fred Drake6caae142000-10-25 17:51:02[diff] [blame]1080
Martin v. Löwisf30d60e2004-06-08 08:17:44[diff] [blame]1081--with-tsc: Profile using the Pentium timestamping counter (TSC).
Guido van Rossum76be6ed1995-01-02 18:33:54[diff] [blame]1082
Martin v. Löwis9176fc12006-04-11 11:12:43[diff] [blame]1083--with-system-ffi: Build the _ctypes extension module using an ffi
Georg Brandlb0885b72007-12-09 08:59:45[diff] [blame]1084 library installed on the system.
Martin v. Löwis9176fc12006-04-11 11:12:43[diff] [blame]1085
Michael W. Hudson71dcc3e2005-02-22 15:33:26[diff] [blame]1086
Guido van Rossum433c8ad1994-08-01 12:07:07[diff] [blame]1087Building for multiple architectures (using the VPATH feature)
1088-------------------------------------------------------------
1089
1090If your file system is shared between multiple architectures, it
1091usually is not necessary to make copies of the sources for each
1092architecture you want to support. If the make program supports the
1093VPATH feature, you can create an empty build directory for each
1094architecture, and in each directory run the configure script (on the
1095appropriate machine with the appropriate options). This creates the
1096necessary subdirectories and the Makefiles therein. The Makefiles
Guido van Rossum4f40c2d2000-09-05 16:46:22[diff] [blame]1097contain a line VPATH=... which points to a directory containing the
Guido van Rossumd0fe8451996-08-26 03:02:37[diff] [blame]1098actual sources. (On SGI systems, use "smake -J1" instead of "make" if
1099you use VPATH -- don't try gnumake.)
Guido van Rossum433c8ad1994-08-01 12:07:07[diff] [blame]1100
1101For example, the following is all you need to build a minimal Python
1102in /usr/tmp/python (assuming ~guido/src/python is the toplevel
1103directory and you want to build in /usr/tmp/python):
1104
Anthony Baxter70e07962006-04-03 14:16:27[diff] [blame]1105 $ mkdir /usr/tmp/python
1106 $ cd /usr/tmp/python
1107 $ ~guido/src/python/configure
1108 [...]
1109 $ make
1110 [...]
1111 $
Guido van Rossum433c8ad1994-08-01 12:07:07[diff] [blame]1112
Neil Schemenauer73f8ab22001-01-26 22:18:55[diff] [blame]1113Note that configure copies the original Setup file to the build
Guido van Rossum433c8ad1994-08-01 12:07:07[diff] [blame]1114directory if it finds no Setup file there. This means that you can
1115edit the Setup file for each architecture independently. For this
1116reason, subsequent changes to the original Setup file are not tracked
1117automatically, as they might overwrite local changes. To force a copy
1118of a changed original Setup file, delete the target Setup file. (The
1119makesetup script supports multiple input files, so if you want to be
1120fancy you can change the rules to create an empty Setup.local if it
1121doesn't exist and run it with arguments $(srcdir)/Setup Setup.local;
1122however this assumes that you only need to add modules.)
1123
Christian Heimes0a0e5832007-12-13 19:23:16[diff] [blame]1124Also note that you can't use a workspace for VPATH and non VPATH builds. The
1125object files left behind by one version confuses the other.
1126
Guido van Rossum433c8ad1994-08-01 12:07:07[diff] [blame]1127
Guido van Rossum433c8ad1994-08-01 12:07:07[diff] [blame]1128Building on non-UNIX systems
1129----------------------------
1130
Martin v. Löwis2eba0d62004-12-24 08:28:28[diff] [blame]1131For Windows (2000/NT/ME/98/95), assuming you have MS VC++ 7.1, the
Tim Peters03444242000-09-19 00:38:35[diff] [blame]1132project files are in PCbuild, the workspace is pcbuild.dsw. See
1133PCbuild\readme.txt for detailed instructions.
Guido van Rossum8d7d4ed1996-07-30 21:41:07[diff] [blame]1134
Martin v. Löwis2eba0d62004-12-24 08:28:28[diff] [blame]1135For other non-Unix Windows compilers, in particular MS VC++ 6.0 and
Guido van Rossum31ae2071999-04-12 14:47:30[diff] [blame]1136for OS/2, enter the directory "PC" and read the file "readme.txt".
Guido van Rossum6d9cc801996-08-01 17:31:22[diff] [blame]1137
1138For the Mac, a separate source distribution will be made available,
1139for use with the CodeWarrior compiler. If you are interested in Mac
1140development, join the PythonMac Special Interest Group
1141(http://www.python.org/sigs/pythonmac-sig/, or send email to
1142pythonmac-sig-request@python.org).
1143
1144Of course, there are also binary distributions available for these
Guido van Rossum67345eb2001-01-18 14:51:12[diff] [blame]1145platforms -- see http://www.python.org/.
Guido van Rossum6d9cc801996-08-01 17:31:22[diff] [blame]1146
1147To port Python to a new non-UNIX system, you will have to fake the
1148effect of running the configure script manually (for Mac and PC, this
1149has already been done for you). A good start is to copy the file
Michael W. Hudsonbe6e0f42002-06-23 16:29:36[diff] [blame]1150pyconfig.h.in to pyconfig.h and edit the latter to reflect the actual
Guido van Rossum433c8ad1994-08-01 12:07:07[diff] [blame]1151configuration of your system. Most symbols must simply be defined as
11521 only if the corresponding feature is present and can be left alone
Michael W. Hudsonbe6e0f42002-06-23 16:29:36[diff] [blame]1153otherwise; however the *_t type symbols must be defined as some
1154variant of int if they need to be defined at all.
Guido van Rossum433c8ad1994-08-01 12:07:07[diff] [blame]1155
Barry Warsaw06557452001-12-14 20:47:12[diff] [blame]1156For all platforms, it's important that the build arrange to define the
1157preprocessor symbol NDEBUG on the compiler command line in a release
1158build of Python (else assert() calls remain in the code, hurting
1159release-build performance). The Unix, Windows and Mac builds already
1160do this.
Guido van Rossum433c8ad1994-08-01 12:07:07[diff] [blame]1161
Guido van Rossum91cb9d21995-04-10 11:47:38[diff] [blame]1162
1163Miscellaneous issues
1164====================
1165
Guido van Rossum91cb9d21995-04-10 11:47:38[diff] [blame]1166Emacs mode
1167----------
1168
Guido van Rossumc0be2f51997-10-08 05:05:28[diff] [blame]1169There's an excellent Emacs editing mode for Python code; see the file
1170Misc/python-mode.el. Originally written by the famous Tim Peters, it
Guido van Rossum4f40c2d2000-09-05 16:46:22[diff] [blame]1171is now maintained by the equally famous Barry Warsaw (it's no
Guido van Rossum67345eb2001-01-18 14:51:12[diff] [blame]1172coincidence that they now both work on the same team). The latest
Guido van Rossumc07d5fa2000-09-01 22:50:02[diff] [blame]1173version, along with various other contributed Python-related Emacs
Georg Brandlf004d9d2009-10-27 15:39:53[diff] [blame]1174goodies, is online at http://www.python.org/emacs/python-mode/. And
Guido van Rossumc07d5fa2000-09-01 22:50:02[diff] [blame]1175if you are planning to edit the Python C code, please pick up the
Georg Brandlf004d9d2009-10-27 15:39:53[diff] [blame]1176latest version of CC Mode http://www.python.org/emacs/cc-mode/; it
Guido van Rossumc07d5fa2000-09-01 22:50:02[diff] [blame]1177contains a "python" style used throughout most of the Python C source
1178files. (Newer versions of Emacs or XEmacs may already come with the
1179latest version of python-mode.)
Guido van Rossum91cb9d21995-04-10 11:47:38[diff] [blame]1180
1181
Guido van Rossum59b0b602002-02-08 22:22:35[diff] [blame]1182Tkinter
1183-------
Guido van Rossum91cb9d21995-04-10 11:47:38[diff] [blame]1184
Guido van Rossum59b0b602002-02-08 22:22:35[diff] [blame]1185The setup.py script automatically configures this when it detects a
1186usable Tcl/Tk installation. This requires Tcl/Tk version 8.0 or
Fred Drake8179a9d2000-09-28 23:28:30[diff] [blame]1187higher.
Guido van Rossum6d9cc801996-08-01 17:31:22[diff] [blame]1188
Guido van Rossum973e4dc2000-07-01 00:34:39[diff] [blame]1189For more Tkinter information, see the Tkinter Resource page:
1190http://www.python.org/topics/tkinter/
Guido van Rossumc0be2f51997-10-08 05:05:28[diff] [blame]1191
Michael W. Hudson71dcc3e2005-02-22 15:33:26[diff] [blame]1192There are demos in the Demo/tkinter directory.
Guido van Rossum91cb9d21995-04-10 11:47:38[diff] [blame]1193
1194Note that there's a Python module called "Tkinter" (capital T) which
Guido van Rossumc07d5fa2000-09-01 22:50:02[diff] [blame]1195lives in Lib/lib-tk/Tkinter.py, and a C module called "_tkinter"
Guido van Rossum6d9cc801996-08-01 17:31:22[diff] [blame]1196(lower case t and leading underscore) which lives in
Guido van Rossum4f40c2d2000-09-05 16:46:22[diff] [blame]1197Modules/_tkinter.c. Demos and normal Tk applications import only the
Guido van Rossum59b0b602002-02-08 22:22:35[diff] [blame]1198Python Tkinter module -- only the latter imports the C _tkinter
1199module. In order to find the C _tkinter module, it must be compiled
1200and linked into the Python interpreter -- the setup.py script does
1201this. In order to find the Python Tkinter module, sys.path must be
1202set correctly -- normal installation takes care of this.
Guido van Rossum84c8c7f1995-08-28 02:44:24[diff] [blame]1203
Guido van Rossum91cb9d21995-04-10 11:47:38[diff] [blame]1204
Guido van Rossum433c8ad1994-08-01 12:07:07[diff] [blame]1205Distribution structure
1206----------------------
1207
Guido van Rossum4f40c2d2000-09-05 16:46:22[diff] [blame]1208Most subdirectories have their own README files. Most files have
Guido van Rossum433c8ad1994-08-01 12:07:07[diff] [blame]1209comments.
1210
Guido van Rossum6d9cc801996-08-01 17:31:22[diff] [blame]1211Demo/ Demonstration scripts, modules and programs
Georg Brandlc207c712008-02-09 21:38:54[diff] [blame]1212Doc/ Documentation sources (reStructuredText)
Guido van Rossum6d9cc801996-08-01 17:31:22[diff] [blame]1213Grammar/ Input for the parser generator
1214Include/ Public header files
Anthony Baxter70e07962006-04-03 14:16:27[diff] [blame]1215LICENSE Licensing information
Guido van Rossum6d9cc801996-08-01 17:31:22[diff] [blame]1216Lib/ Python library modules
Michael W. Hudsonbe6e0f42002-06-23 16:29:36[diff] [blame]1217Mac/ Macintosh specific resources
Neil Schemenauer3f52da52001-01-24 17:49:59[diff] [blame]1218Makefile.pre.in Source from which config.status creates the Makefile.pre
Guido van Rossumc0be2f51997-10-08 05:05:28[diff] [blame]1219Misc/ Miscellaneous useful files
Guido van Rossum6d9cc801996-08-01 17:31:22[diff] [blame]1220Modules/ Implementation of most built-in modules
1221Objects/ Implementation of most built-in object types
Guido van Rossumc07d5fa2000-09-01 22:50:02[diff] [blame]1222PC/ Files specific to PC ports (DOS, Windows, OS/2)
Anthony Baxter70e07962006-04-03 14:16:27[diff] [blame]1223PCbuild/ Build directory for Microsoft Visual C++
Guido van Rossum6d9cc801996-08-01 17:31:22[diff] [blame]1224Parser/ The parser and tokenizer and their input handling
Guido van Rossum4f40c2d2000-09-05 16:46:22[diff] [blame]1225Python/ The byte-compiler and interpreter
Guido van Rossum6d9cc801996-08-01 17:31:22[diff] [blame]1226README The file you're reading now
Neal Norwitzf8519c42007-05-20 18:43:00[diff] [blame]1227RISCOS/ Files specific to RISC OS port
Guido van Rossum6d9cc801996-08-01 17:31:22[diff] [blame]1228Tools/ Some useful programs written in Python
Michael W. Hudsonbe6e0f42002-06-23 16:29:36[diff] [blame]1229pyconfig.h.in Source from which pyconfig.h is created (GNU autoheader output)
Guido van Rossum6d9cc801996-08-01 17:31:22[diff] [blame]1230configure Configuration shell script (GNU autoconf output)
Guido van Rossumc07d5fa2000-09-01 22:50:02[diff] [blame]1231configure.in Configuration specification (input for GNU autoconf)
Guido van Rossum6d9cc801996-08-01 17:31:22[diff] [blame]1232install-sh Shell script used to install files
Anthony Baxter70e07962006-04-03 14:16:27[diff] [blame]1233setup.py Python script used to build extension modules
Guido van Rossum91cb9d21995-04-10 11:47:38[diff] [blame]1234
1235The following files will (may) be created in the toplevel directory by
1236the configuration and build processes:
1237
Guido van Rossum6d9cc801996-08-01 17:31:22[diff] [blame]1238Makefile Build rules
Neil Schemenauer3f52da52001-01-24 17:49:59[diff] [blame]1239Makefile.pre Build rules before running Modules/makesetup
Michael W. Hudsonbe6e0f42002-06-23 16:29:36[diff] [blame]1240buildno Keeps track of the build number
Guido van Rossumc07d5fa2000-09-01 22:50:02[diff] [blame]1241config.cache Cache of configuration variables
Michael W. Hudsonbe6e0f42002-06-23 16:29:36[diff] [blame]1242pyconfig.h Configuration header
Guido van Rossumc0be2f51997-10-08 05:05:28[diff] [blame]1243config.log Log from last configure run
Guido van Rossum4f40c2d2000-09-05 16:46:22[diff] [blame]1244config.status Status from last run of the configure script
Anthony Baxter70e07962006-04-03 14:16:27[diff] [blame]1245getbuildinfo.o Object file from Modules/getbuildinfo.c
1246libpython<version>.a The library archive
Guido van Rossum6d9cc801996-08-01 17:31:22[diff] [blame]1247python The executable interpreter
Neal Norwitzf8519c42007-05-20 18:43:00[diff] [blame]1248reflog.txt Output from running the regression suite with the -R flag
Guido van Rossum6d9cc801996-08-01 17:31:22[diff] [blame]1249tags, TAGS Tags files for vi and Emacs
Guido van Rossum433c8ad1994-08-01 12:07:07[diff] [blame]1250
1251
Guido van Rossumc07d5fa2000-09-01 22:50:02[diff] [blame]1252That's all, folks!
1253------------------
Guido van Rossum433c8ad1994-08-01 12:07:07[diff] [blame]1254
1255
Guido van Rossum67345eb2001-01-18 14:51:12[diff] [blame]1256--Guido van Rossum (home page: http://www.python.org/~guido/)