Strangely, two out of three patches there seem already committed; but
the essential one (get rid of the assert in object_filenames in
ccompiler.py) was not yet applied.
This makes the build procedure for Twisted work again.
This is *not* a backport candidate despite the fact that identical
code appears to exist in 2.2.2; Twisted builds fine there, so there
must have been a change elsewhere.
The default implementation calls _compile() to compile individual
files. This method must be implemented by the subclass. This change
factors out most of the remaining common code in all the compilers
except mwerks.
Two new tests are needed:
Don't skip building an extension if any of the depends files are newer
than the target.
Pass ext.depends to compiler.compile() so that it can track individual
files.
Always use _setup_compile() to do the grunt work of processing
arguments, figuring out which files to compile, and emitting debug
messages for files that are up-to-date.
Use _get_cc_args() when possible.
This change is not backwards compatible. If a compiler subclass
exists outside the distutils package, it may get called with the
unexpected keyword arg. It's easy to extend that compiler by having
it ignore the argument, and not much harder to do the right thing. If
this ends up being burdensome, we can change it before 2.3 final to
work harder at compatibility.
Also add _setup_compile() and _get_cc_args() helper functions that
factor out much of the boilerplate for each concrete compiler class.
Remove __init__ that just called base class __init__ with same args.
Fold long argument lists into fewer, shorter lines.
Remove parens in tuple unpacks.
Don't put multiple statements on one line with a semicolon.
In find_library_file() compute the library_filename() upfront.
This is a conservative version of SF patch 504889. It uses the log
module instead of calling print in various places, and it ignores the
verbose argument passed to many functions and set as an attribute on
some objects. Instead, it uses the verbosity set on the logger via
the command line.
The log module is now preferred over announce() and warn() methods
that exist only for backwards compatibility.
XXX This checkin changes a lot of modules that have no test suite and
aren't exercised by the Python build process. It will need
substantial testing.
# XXX this isn't used anywhere, and worse, it has the same name as a method
# in Command with subtly different semantics. (This one just has one
# source -> one dest; that one has many sources -> one dest.) Nuke it?
Yes. Nuke it.
modules, distutils does not understand that the build version of the
source tree is needed.
This patch fixes distutils.sysconfig to understand that the running
Python is part of the build tree and needs to use the appropriate
"shape" of the tree. This does not assume anything about the current
directory, so can be used to build 3rd-party modules using Python's
build tree as well.
This is useful since it allows us to use a non-installed debug-mode
Python with 3rd-party modules for testing. It as the side-effect that
set_python_build() is no longer needed (the hack which was added to
allow distutils to be used to build the "standard" extension modules).
This closes SF patch #547734.
--install-script ... command line option to bdist_wininst) at the end
of the installation and at the start of deinstallation. Output
(stdout, stderr) of the script (if any) is displayed in the last
screen at installation, or in a simple message box at deinstallation.
sys.argv[1] for the script will contain '-install' at installation
time or '-remove' at deinstallation time.
The installation script runs in an environment (embedded by the
bdist_wininst runtime) where an additional function is available as
builtin:
create_shortcut(path, description, filename,
[arguments[, workdir[, iconpath, iconindex]]])
Recreated this file after source changes.
installations are present, by always unlinking the destination file
before copying to it. Without the unlink(), the copied file remains
owned by its previous UID, causing the subsequent chmod() to fail.
Bugfix candidate, though it may cause changes on platforms where
file ownership behaves differently.
contain the type of the file (regular file, socket, link, &c.).
This means that install_scripts will now print
"changing mode of <file> to 775" instead of "... to 100775".
2.2 bugfix candidate, I suppose, though this isn't actually fixing a bug.
present - at least the swigged file should be named <name>_wrap.c as
this is also SWIG's default. (Even better would be to generate the
wrapped sources in a different location, but I'll leave this for
later).
Newer versions of SWIG don't accept the -dnone flag any more.
Since virtually nobody uses SWIG with distutils, this should do no
harm.
Suggested be Martin Bless on c.l.p.
crashes.
If no external zip-utility is found, the archive is created by the
zipfile module, which behaves different now than in 2.1: if the
zip-file is created in the root directory if the distribution, it will
contain an (empty) version of itself.
This triggered the above bug - so it's better to create the zip-file
far away in the TMP directory.
>
> When using 'distutils' (shipped with Python 2.1) I've found that my
> Python scripts installed with a first line of:
>
> #!/usr/bin/python2.1None
>
> This is caused by distutils trying to patch the first line of the python
> script to use the current interpreter.
distutils for the library modules built as shared objects. A better solution
appears possible, but with the threat that the distutils becomes more
magical ("complex").
This closes SF bug #458343.
bdist_wininst doesn't use the NT SCHEME any more, instead
a custom SCHEME is used, which is exchanged at installation
time, depending on the python version used.
Avoid a bogus warning frpom install_lib about installing
into a directory not on sys.path.
modules and extensions on Windows is now $PREFIX/Lib/site-packages.
Includes backwards compatibility code for pre-2.2 Pythons. Contributed
by Paul Moore.
libraries. This is done by adding a .get_source_files() method,
contributed by Rene Liebscher and slightly modified.
Remove an unused local variable spotted by PyChecker
though 'licence' is still supported for backward-compatibility
(Should I add a warning to get_licence(), or not bother?)
Also fixes an UnboundLocalError noticed by PyChecker
to the current Python interpreter (ie. the one used for
building/installation), even (especially!) if "/usr/bin/env" appears in
the #! line.
Rationale: installing scripts with "#!/usr/bin/env python" is asking for
trouble, because
1) it might pick the wrong interpreter (not the one used to
build/install the script)
2) it doesn't work on all platforms (try it on IRIX 5, or on Linux
with command-line options for python)
3) "env" might not be in /usr/bin
- compile() didn't return a (empty) list of objects. Fixed.
- the various _fix_xxx_args() methods weren't called (are they new or did I overlook them?). Fixed.
along with options to print them.
Add a finalize_options() method to Distribution to do final processing
on the platform and keyword attributes
Add DistributionMetadata.write_pkg_info() method to write a PKG-INFO file
into the release tree.
before this get forgotten again.
Should probably be set to 1.0.2 before final release of python 2.1
Does someone still release distutils separate from python?
has been changed to include an uninstaller.
I forgot to mention in the uninstaller checkin that the logfile
name (used for uninstalling) has been changed from
<module>.log to <module>-wininst.log. This should prevent
conflicts with a distutils logfile serving the same purpose.
The short form of the --bdist-dir (-d) option has been removed
because it caused conflicts with the short form of the --dist-dir
option.
the Cygwin-specific compiler class.
(According to Jason Tishler, cygwinccompiler needs some work to
handle the differences in Cygwin- and MSVC-Python. Makefile and
config files are currently ignored by cygwinccompiler, as it was
written to support cygwin for extensions which are intended to be
used with the standard MSVC built Python.)
--bitmap command line option allows to use a different bitmap file instead
of the build-in python powered logo.
--title lets you specify the text to display on the background.
The editbox in the first screen now longer is
selected (highlighted), it had the WS_TABSTOP flag.
This is the patch
http://sourceforge.net/patch/?func=detailpatch&patch_id=103687&group_id=5470
with two changes:
1. No messagebox displayed when the compilation to .pyc or .pyo files
failes, this will only confuse the user (and it will fail under certain
cases, where sys.path contains garbage).
2. A debugging print statement was removed from bdist_wininst.py.
and also takes the sys.platform name into account. This helps on
platforms where there are multiple possible compiler backends (the
one with which Python itself was compiled is preferred over others
in this case).
The patch uses this new technique to enable using cygwin compiler
per default for cygwin compiled Pythons.
Written by Marc-Andre Lemburg. Copyright assigned to Guido van Rossum.
prevent binding for str from masking use of builtin str in nested
function.
(This is the only case I found in the standard library where a local
shadows a global or builtin. There may be others, but the regression
test doesn't catch them.)
Lib/distutils/command/build_ext.py(build_ext.finalize_options): Add
Cygwin specific code to append Python's library directory to the
extension's list of library directories.
(build_ext.get_libraries): Add Cygwin specific code to append Python's
(import) library to the extension's list of libraries.
This patch adds support for Cygwin to util.get_platform(). A Cygwin
specific case is needed due to the format of Cygwin's uname command,
which contains '/' characters.
sys.prefix + 'config/Makefile'. When building Python for the first
time, these files aren't there, so the files from the build tree have
to be used instead; this file adds an entry point for specifying that
the build tree files should be used. (Perhaps 'set_python_build' should
should be preceded with an underscore?)
for done[n] can be integers as well as strings, but the code
concatenates them with strings (fixed by adding a str()) and calls
string.strip() on them (fixed by rearranging the logic)
(Presumably this wasn't noticed previously because parse_makefile()
was only called on Modules/Makefile, which contains no integer-valued
variables.)
produce a list of unique filenames:
"While attempting to build an RPM using distutils on Python 2.0,
rpm complained about duplicate files. The following patch fixed
that problem.
about how it would be nice to write absolute paths to the temporary
byte-compilation script, but this doesn't work because it screws up the
trailing-slash trickery done to 'prefix' in build_py's 'byte_compile()'
method.
Fixed to use 'execute()' instead of 'os.remove()' to remove the temporary
script: now it doesn't blow up in dry-run mode!
by default (since compiling at install time works just fine). Details:
- added 'compile' and 'optimize' options
- added 'byte_compile()' method
- changed 'get_outputs()' so it includes bytecode files
A lot of the code added is very similar to code in install_lib.py;
would be nice to factor it out further.
choice between (compile, no-compile) * (optimize=0, optimize=1,
optimize=2). Details:
- added --no-compile option to complement --compile, which has
been there for ages
- changed --optimize (which never worked) to a value option, which
expects 0, 1, or 2
- renamed 'bytecompile()' method to 'byte_compile()', and beefed
it up to handle both 'compile' and 'optimize' options
- fix '_bytecode_filenames()' to respect the new options
standard 'py_compile.compile()' function. Laundry list of features:
- handles standard Distutils 'force', 'verbose', 'dry_run' flags
- handles various levels of optimization: can compile directly in
this interpreter process, or write a temporary script that is
then executed by a new interpreter with the appropriate flags
- can rewrite the source filename by stripping an optional prefix
and preprending an optional base dir.
- added 'sub_commands' class attr
- added 'has_*()' predicates referenced by the sub-command list
- rewrote 'run()' so it's a trivial loop over relevant sub-commands
They are unneeded: All this stuff is already done by the
install command which is run by bdist_wininst.
One bug has been fixed:
The root of the fake install tree is install.install_purelib,
not install.install_lib!
They are different if the extra_path option is used in
the setup function.
Rebuild after the changes to wininst.exe.
Removed get_ext_libname() because it is unused.
Fixed get_libraries() to append an '_d' to the python debug
import library. If MSVC is used, do not add 'pythonxx.lib' to
the list of libraries, because this is handled better
by a pragma in config.h.
This should fix bug #115595, but it needs some more testing.
implementations. Details:
* replace 'link_shared_object()', 'link_shared_lib()', and
'link_executable()' with 'link()', which is (roughly)
the union of the three methods it replaces
* in all implementation classes (UnixCCompiler, MSVCCompiler, etc.),
ditch the old 'link_*()' methods and replace them with 'link()'
* in the abstract base class (CCompiler), add the old 'link_*()'
methods as wrappers around the new 'link()' (they also print
a warning of the deprecated interface)
Also increases consistency between MSVCCompiler and BCPPCompiler,
hopefully to make it easier to factor out the mythical WindowsCCompiler
class. Details:
* use 'self.linker' instead of 'self.link'
* add ability to compile resource files to BCPPCompiler
* added (redundant?) 'object_filename()' method to BCPPCompiler
* only generate a .def file if 'export_symbols' defined
* options can now be spelled "foo-bar" or "foo_bar" (handled in
'parse_config_files()', just after we parse a file)
* added a "[global]" section so there's a place to set global
options like verbose/quiet and dry-run
* respect the "negative alias" dictionary so (eg.) "quiet=1" is
the same as "verbose=0" (this had to be done twice: once in
'parse_config_file()' for global options, and once in
'_set_command_options()' for per-command options)
* the other half of handling boolean options correctly: allow
commands to list their boolean options in a 'boolean_options'
class attribute, and use it to translate strings (like "yes", "1",
"no", "0", etc) to true or false
'convert_paths()' method to convert them all to the local syntax (backslash
or colon or whatever) at the appropriate time.
Added SCHEME_KEYS to get rid of one hard-coded list of attributes (in
'select_scheme()').
Default 'install_path_file' to true, and never set it false (it's just
there in case some outsider somewhere wants to disable installation of the
.pth file for whatever reason).
Toned down the warning emitted when 'install_path_file' is false, since we
no longer know why it might be false.
Added 'warn_dir' flag to suppress warning when installing to a directory
not in sys.path (again, we never set this false -- it's there for outsiders
to use, specifically the "bdist_*" commands).
Pulled the loop of 'change_root()' calls out to new method 'change_roots()'.
Comment updates/deletions/additions.
resource files. The gist of the patch is to treat ".rc" and ".mc"
files as source files; ".mc" files are compiled to ".rc" and then
".res", and ".rc" files are compiled to ".res". Wish I knew what
all these things stood for...
in a string (gives you something to do with the dictionary returned
by 'parse_makefile()').
Pulled the regexes in 'parse_makefile()' out -- they're now globals,
as 'expand_makefile_vars()' needs (two of) them.
Cosmetic tweaks to 'parse_makefile()'.
easier for people porting Makefile.pre.in-based extensions to Distutils.
Also loosened argument-checking in Extension constructor to make life
easier for 'read_setup_file()'.
are completely skipped, rather than being treated as blank lines
(and then subject to the 'skip_blanks' flag). This allows us
to process old-style Setup files, which rely on
hello \\
# boo!
there
coming out as "hello there".
Dropped the 'collapse_ws' option and replaced it with 'collapse_join' --
it's *much* faster (no 're.sub()') and this is the reason I really added
'collapse_ws', ie. to remove leading whitespace from a line being joined
to the previous line.
(eg. "bdist_dumb", to generate both ZIP and tar archives in the same
run), tell all but the last run to keep temp files -- this just gets
rid of the need to pseudo-install the same files multiple times.
the command's sub-commands as well (off by default). This is essential if
we want to be be able to run (eg.) "install" twice in one run, as happens
when generating multiple built distributions in one run.
families" -- eg. install and its brood, build and its brood, and so forth.
Specifically: added the 'sub_commands' class attribute (empty list, sub-
classes must override it) and a comment describing it, and the
'get_sub_commands()' method.
of globals from sysconfig.
Added 'prefix' and 'exec_prefix' to the list of variables that can be
expanded in installation directories (preserving the stupid old names
of 'sys_prefix' and 'sys_exec_prefix, though).
all that work when someone asks for a "configuration variable" from the
Makefile. Details:
- added 'get_config_vars()': responsible for calling one of the
'_init_*()' functions to figure things out for this platform,
and to provide an interface to the resulting dictionary
- added 'get_config_var()' as a simple interface to the dictionary
loaded by 'get_config_vars()'
- changed the '_init_*()' functions so they load the global dictionary
'_config_vars', rather than spewing their findings all over
the module namespace
- don't delete the '_init_*()' functions when done importing
- adjusted 'customize_compiler()' to the new regime
used to create the distribution and the creation date.
Takes care of the extra_path argument to the setup function,
installs the modules into <prefix>/extra_path and creates
a -pth file (like install_lib does).
distutils/command/bdist_wininst.py:
- the windows installer is again able to compile after installing
the files. Note: The default has changed, the packager has to
give --no-target-compile/--no-target-optimize to NOT compile
on the target system. (Another note: install_lib's --compile
--optimize options have the same semantics to switch off
the compilation. Shouldn't the names change?)
- All references to specific python versions are gone.
- A small bug:
raise DistutilsPlatformError ("...")
instead of
raise DistutilsPlatformError, ("...")
- When bdist_wininst creates an installer for one specific python
version, this is reflected in the name:
Distutils-0.9.2.win32-py15.exe instead of
Distutils-0.9.2.win32.exe
- bdist_wininst, when run as script, reads the wininst.exe file
and rewrites itself. Previously this was done by hand.
misc/install.c
- All the changes needed for compilation
- Deleted a lot of debug/dead code
* ensure the "dist" directory exists
* raise exception if using for modules containing compiled extensions
on a non-win32 platform.
* don't create an .ini file anymore (it was just for debugging)
fairly tight control, and the '_setup_stop_after' and '_setup_distribution'
globals to provide the tight control.
This isn't entirely reliable yet: it dies horribly with a NameError on the
example PIL setup script in examples/pil_setup.py (at least with Python
1.5.2; untested with current Python). There's some strangeness going
on with execfile(), but I don't understand it and don't have time
to track it down right now.
according to the MS docs it enables exception-handling, and (according
to Alex Martelli <aleaxit@yahoo.com>) is needed to compile without
getting warnings from standard C++ library headers. Apparently
it doesn't cause any problems with C code, so I haven't bothered
conditionalizing the use of /GX.
Changed 'core.setup()' so it sets them to reasonable defaults.
Tweaked how the "usage" string is generated: 'core' now provides
'gen_usage()', which is used instead of 'USAGE'.
Modified "build_py" and "sdist" commands to refer to
'self.distribution.script_name' rather than 'sys.argv[0]'.
The known bug (bogus error message when an empty file is
extracted) is fixed.
Other changes:
- The target-compile and target-optimize flags of bdist_wininst
are gone. It is no longer possible to compile the python
files during installation.
- The zlib module is no longer required or used by bdist_wininst.
- I moved the decompression/extraction code into a separate
file (extract.c).
- The installer stub is now compressed by UPX (see
http://upx.tsx.org/). This reduces the size of the exe
(and thus the overhead of the final installer program)
from 40 kB to 16 kB.
- The installer displays a more uptodate user wizard-like
user interface, also containing a graphic: Just's Python Powered logo.
(I could not convince myself to use one of the BeOpen logos).
- The installation progress bar now moves correctly.
* use self.debug_print() for debug messages
* uses now copy.copy() to copy lists
* added 'shared_lib_extension=".dll"', ... , this is necessary if you
want use the compiler class outside of the standard distutils build
process.
* changed result type of check_config_h() from int to string
participates in the "--root" hack, ie. it also has a new root directory
hacked on at the very last minute (essential if the .pth file is to be
included in an RPM or other smart installer!).
- added big comment describing possible problems
- look for and react to versions of gcc, ld, and dlltool; mainly
this is done by the 'get_versions()' function and the CygwinCCompiler
and Mingw32CCompiler constructors
- move 'check_config_h()' to end of file and defer calling it until
we need to (ie. in the CygwinCCompiler constructor)
- lots of changes in 'link_shared_object()' -- mostly seems to be
library and DLL stuff, but I don't follow it entirely
it so BCPPCompiler actually works, so I'm provisionally accepting it
-- ugly and working is better than not working! Major changes:
- normalize paths (apparently BC++ doesn't like slashes)
- overhauled how we search for and specify libraries on the linker
command-line
- hacked up 'find_library_file()' so it knows about "debug" library
naming convention as well as "bcpp_xxx.lib" -- the question is,
is this a well-established and sensible convention?
Also:
- change to use 'util.write_file()' to write the .def file
- 'export_symbol_file' (and corresponding 'def_file' in the old
"build info" dict) are gone; warn if we see 'def_file' in the
dict
- the MSVC "pre-link hack" is gone -- all that stuff is now handled
elsewhere (eg. by using 'export_symbols', etc.)
- add 'get_export_symbols()' and 'get_libraries()' methods -- needed
because on Windows, both of those things are a tad more complicated
than fetching them from the Extension instance
Added 'remove_duplicates()'.
Simplified constructor: no longer take 'files' or 'allfiles' as args,
and no longer have 'dir' attribute at all.
Added 'set_allfiles()' and 'findall()' so the client does have a
way to set the list of all files.
Changed 'include_pattern()' to use the 'findall()' method instead of
the external function. (Of course, the method is just a trivial
wrapper around the function.)
Martin v. Löwis
Marc-André Lemburg
Skip Montanaro
Guido van Rossum
Jeremy Hylton
Andrew M. Kuchling
Neal Norwitz
Andrew MacIntyre
Thomas Heller
Michael W. Hudson
Jack Jansen
Fred Drake
Barry Warsaw
Tim Peters
Just van Rossum
Neil Schemenauer
Greg Ward
unknown
Ka-Ping Yee
Eric S. Raymond