Quick Start Guide

-----------------

1. Install Microsoft Visual Studio 2017 or later with Python workload and

Python native development component.

1a. Optionally install Python 3.10 or later. If not installed,

get_externals.bat (via build.bat) will download and use Python via

NuGet.

2. Run "build.bat" to build Python in 64-bit Release configuration.

3. (Optional, but recommended) Run the test suite with "rt.bat -q".

Building Python using Microsoft Visual C++

------------------------------------------

This directory is used to build CPython for Microsoft Windows on 32- and 64-

bit platforms. Using this directory requires an installation of

Microsoft Visual Studio (MSVC) with the *Python workload* and

its optional *Python native development* component selected.

Building from the command line is recommended in order to obtain any

external dependencies. To build, simply run the "build.bat" script without

any arguments. After this succeeds, you can open the "pcbuild.sln"

solution in Visual Studio to continue development.

To build an installer package, refer to PC/layout.

The solution currently supports two platforms. The Win32 platform is

used to build standard x86-compatible 32-bit binaries, output into the

win32 sub-directory. The x64 platform is used for building 64-bit AMD64

(aka x86_64 or EM64T) binaries, output into the amd64 sub-directory.

The Itanium (IA-64) platform is no longer supported.

Four configuration options are supported by the solution:

Debug

Used to build Python with extra debugging capabilities, equivalent

to using ./configure --with-pydebug on UNIX. All binaries built

using this configuration have "_d" added to their name:

python310_d.dll, python_d.exe, parser_d.pyd, and so on. Both the

build and rt (run test) batch files in this directory accept a -d

option for debug builds. If you are building Python to help with

development of CPython, you will most likely use this configuration.

PGInstrument, PGUpdate

Used to build Python in Release configuration using PGO, which

requires Premium Edition of Visual Studio. See the "Profile

Guided Optimization" section below for more information. Build

output from each of these configurations lands in its own

sub-directory of this directory. The official Python releases may

be built using these configurations.

Release

Used to build Python as it is meant to be used in production

settings, though without PGO.

Building Python using Clang/LLVM

--------------------------------

See https://learn.microsoft.com/cpp/build/clang-support-msbuild

for how to install and use clang-cl bundled with Microsoft Visual Studio.

You can use the IDE to switch to clang-cl for local development,

but because this alters the *.vcxproj files, the recommended way is

to use build.bat:

build.bat "/p:PlatformToolset=ClangCL"

All other build.bat options continue to work as with MSVC, so this

will create a 64bit release binary.

You can also use a specific version of clang-cl downloaded from

https://github.com/llvm/llvm-project/releases, e.g.

clang+llvm-18.1.8-x86_64-pc-windows-msvc.tar.xz.

Given you have extracted that to <my-clang-dir>, you can use it like so

build.bat --pgo "/p:PlatformToolset=ClangCL" "/p:LLVMInstallDir=<my-clang-dir>" "/p:LLVMToolsVersion=18"

Setting LLVMToolsVersion to the major version is enough, although you

can be specific and use 18.1.8 in the above example, too.

Use the --pgo option to build with PGO (Profile Guided Optimization).

However, if you want to run the PGO task

on a different host than the build host, you must pass

"/p:CLANG_PROFILE_PATH=<relative-path-to-instrumented-dir-on-remote-host>"

in the PGInstrument step to make sure the profile data is generated

into the instrumented directory when running the PGO task.

E.g., if you place the instrumented binaries into the folder

"workdir/instrumented" and then run the PGO task using "workdir"

as the current working directory, the usage is

"/p:CLANG_PROFILE_PATH=instrumented"

Like in the MSVC case, after fetching (or manually copying) the instrumented

folder back into your build tree, you can continue with the PGUpdate

step with no further parameters.

Building Python using the build.bat script

----------------------------------------------

In this directory you can find build.bat, a script designed to make

building Python on Windows simpler. This script will use the env.bat

script to detect either Visual Studio 2017 or later, either of

which may be used to build Python. Currently Visual Studio 2017 is

officially supported.

By default, build.bat will build Python in Release configuration for

the 32-bit Win32 platform. It accepts several arguments to change

this behavior, try `build.bat -h` to learn more.

C Runtime

---------

Visual Studio 2017 uses version 14.0 of the C runtime (vcruntime140).

The executables no longer use the "Side by Side" assemblies used in

previous versions of the compiler. This simplifies distribution of

applications.

The run time libraries are available under the redist folder of your

Visual Studio distribution. For more info, see the Readme in the

redist folder.

Sub-Projects

------------

The CPython project is split up into several smaller sub-projects which

are managed by the pcbuild.sln solution file. Each sub-project is

represented by a .vcxproj and a .vcxproj.filters file starting with the

name of the sub-project. These sub-projects fall into a few general

categories:

The following sub-projects represent the bare minimum required to build

a functioning CPython interpreter. If nothing else builds but these,

you'll have a very limited but usable python.exe:

pythoncore

.dll and .lib

python

.exe

These sub-projects provide extra executables that are useful for running

CPython in different ways:

pythonw

pythonw.exe, a variant of python.exe that doesn't open a Command

Prompt window

_testembed

_testembed.exe, a small program that embeds Python for testing

purposes, used by test_capi.py

These are miscellaneous sub-projects that don't really fit the other

categories:

_freeze_module

_freeze_module.exe, used to regenerate frozen modules in Python

after changes have been made to the corresponding source files

(e.g. Lib\importlib\_bootstrap.py).

python3dll

python3.dll, the PEP 384 Stable ABI dll

(not installed on free-threaded builds)

python3tdll

python3t.dll, the PEP 803 free-threading Stable ABI dll

(built from the same source as python3.dll)

xxlimited

builds an example module that makes use of the PEP 384 Stable ABI,

see Modules\xxlimited.c

xxlimited_35

xxlimited_3_13

ditto for testing older Limited API, see

Modules\xxlimited_*.c

The following sub-projects are for individual modules of the standard

library which are implemented in C; each one builds a DLL (renamed to

.pyd) of the same name as the project:

* _asyncio

* _ctypes

* _ctypes_test

* _decimal

* _elementtree

* _hashlib

* _multiprocessing

* _overlapped

* _queue

* _remote_debugging

* _socket

* _testbuffer

* _testcapi

* _testclinic

* _testclinic_limited

* _testconsole

* _testimportmultiple

* _testinternalcapi

* _testlimitedcapi

* _testmultiphase

* _testsinglephase

* _uuid

* _wmi

* _zoneinfo

* pyexpat

* select

* unicodedata

* winsound

The following Python-controlled sub-projects wrap external projects.

Note that these external libraries are not necessary for a working

interpreter, but they do implement several major features. See the

"Getting External Sources" section below for additional information

about getting the source for building these libraries. The sub-projects

are:

_bz2

Python wrapper for version 1.0.8 of the libbzip2 compression library

Homepage:

http://www.bzip.org/

_lzma

Python wrapper for version 5.2.2 of the liblzma compression library,

which is itself built by liblzma.vcxproj.

Homepage:

https://tukaani.org/xz/

_ssl

Python wrapper for version 3.5 of the OpenSSL secure sockets

library, which is itself downloaded from our binaries repository at

https://github.com/python/cpython-bin-deps and built by openssl.vcxproj.

Homepage:

https://www.openssl.org/

Building OpenSSL requires Perl on your path, and can be performed by

running PCbuild\prepare_ssl.bat. This will retrieve the version of

the sources matched to the current commit from the OpenSSL branch

in our source repository at

https://github.com/python/cpython-source-deps.

To use an alternative build of OpenSSL completely, you should replace

the files in the externals/openssl-bin-<version> folder with your own.

As long as this folder exists, its contents will not be downloaded

again when building.

_sqlite3

Wraps SQLite 3.53.2, which is itself built by sqlite3.vcxproj

Homepage:

https://www.sqlite.org/

_tkinter

Wraps version 9.0.4 of the Tk windowing system, which is downloaded

from our binaries repository at

https://github.com/python/cpython-bin-deps.

Homepage:

https://www.tcl.tk/

Building Tcl and Tk can be performed by running

PCbuild\prepare_tcltk.bat. This will retrieve the version of the

sources matched to the current commit from the Tcl and Tk branches

in our source repository at

https://github.com/python/cpython-source-deps and build them via the

tcl.vcxproj and tk.vcxproj sub-projects.

The two projects install their respective components in a

directory alongside the source directories called "tcltk" on

Win32 and "tcltk64" on x64. They also copy the Tcl and Tk DLLs

into the current output directory, which should ensure that Tkinter

is able to load Tcl/Tk without having to change your PATH.

_zstd

Python wrapper for version 1.5.7 of the zstd compression library

Homepage:

https://facebook.github.io/zstd/

zlib-ng

Compiles zlib-ng as a static library, which is later included by

pythoncore.vcxproj. This was generated using CMake against zlib-ng

version 2.2.4, and should be minimally updated as needed to adapt

to changes in their source layout. The zbuild.h, zconf.h and

zconf-ng.h file in the PC directory were likewise generated and

vendored.

Sources for zlib-ng are imported unmodified into our source

repository at https://github.com/python/cpython-source-deps.

_zstd

Python wrapper for version 1.5.7 of the Zstandard compression library

Homepage:

https://facebook.github.io/zstd/

Getting External Sources

------------------------

The last category of sub-projects listed above wrap external projects

Python doesn't control, and as such a little more work is required in

order to download the relevant source files for each project before they

can be built. However, a simple script is provided to make this as

painless as possible, called "get_externals.bat" and located in this

directory. This script extracts all the external sub-projects from

https://github.com/python/cpython-source-deps

and

https://github.com/python/cpython-bin-deps

via a Python script called "get_external.py", located in this directory.

Everything downloaded by these scripts is stored in ..\externals

(relative to this directory), or the path specified by the EXTERNALS_DIR

environment variable.

The path or command to use for Python can be provided with the

PYTHON_FOR_BUILD environment variable. If this is not set, an active

virtual environment will be used. If none is active, and HOST_PYTHON is

set to a recent enough version or "py.exe" is able to find a recent

enough version, those will be used. If all else fails, a copy of Python

will be downloaded from NuGet and extracted to the externals directory.

This will then be used for later builds (see PCbuild/find_python.bat

for the full logic).

It is also possible to download sources from each project's homepage,

though you may have to change folder names or pass the names to MSBuild

as the values of certain properties in order for the build solution to

find them. This is an advanced topic and not necessarily fully

supported.

The get_externals.bat script is called automatically by build.bat

unless you pass the '-E' option.

Profile Guided Optimization

---------------------------

The solution has two configurations for PGO. The PGInstrument

configuration must be built first. The PGInstrument binaries are linked

against a profiling library and contain extra debug information. The

PGUpdate configuration takes the profiling data and generates optimized

binaries.

The build.bat script has an argument `--pgo` that automate the creation

of optimized binaries.

It creates the PGI files, runs the unit test suite with the PGI python,

and finally creates the optimized files.

You can customize the job for profiling with `--pgo-job <job>` option.

See

https://docs.microsoft.com/en-us/cpp/build/profile-guided-optimizations

for more on this topic.

Optimization flags

------------------

You can set optimization flags either via

* environment variables, for example:

set WITH_COMPUTED_GOTOS=true

* or pass them as parameters to `build.bat`, for example:

build.bat "/p:WITH_COMPUTED_GOTOS=true"

* or put them in `msbuild.rsp` in the `PCbuild` directory, one flag per line.

Supported flags are:

* WITH_COMPUTED_GOTOS: build the interpreter using "computed gotos".

Currently only supported by clang-cl.

* UsePymallocHugepages: enable huge page support for pymalloc arenas.

When enabled, the arena size on 64-bit platforms is increased to 2 MiB

and arena allocation uses MEM_LARGE_PAGES with automatic fallback to

regular pages. Can also be enabled via `--pymalloc-hugepages` flag.

Static library

--------------

The solution has no configuration for static libraries. However it is

easy to build a static library instead of a DLL. You simply have to set

the "Configuration Type" to "Static Library (.lib)" and alter the

preprocessor macro "Py_ENABLE_SHARED" to "Py_NO_ENABLE_SHARED". You may

also have to change the "Runtime Library" from "Multi-threaded DLL

(/MD)" to "Multi-threaded (/MT)".

Visual Studio properties

------------------------

The PCbuild solution makes use of Visual Studio property files (*.props)

to simplify each project. The properties can be viewed in the Property

Manager (View -> Other Windows -> Property Manager) but should be

carefully modified by hand.

The property files used are:

* python (versions, directories and build names)

* pyproject (base settings for all projects)

* openssl (used by projects dependent upon OpenSSL)

* tcltk (used by _tkinter, tcl, and tk projects)

The pyproject property file defines all of the build settings for each

project, with some projects overriding certain specific values. The GUI

doesn't always reflect the correct settings and may confuse the user

with false information, especially for settings that automatically adapt

for different configurations.

Add a new project

-----------------

For example, add a new _testclinic_limited project to build a new

_testclinic_limited extension, the file Modules/_testclinic_limited.c:

* In PCbuild/, copy _testclinic.vcxproj to _testclinic_limited.vcxproj,

replace RootNamespace value with `_testclinic_limited`, replace

`_asyncio.c` with `_testclinic_limited.c`.

* In PCbuild/, copy _testclinic.vcxproj.filters to

_testclinic_limited.vcxproj.filters, edit the list of files in the new file.

* Open Visual Studio, open PCbuild\pcbuild.sln solution, add the

PCbuild\_testclinic_limited.vcxproj project to the solution ("add existing

project).

* Add a dependency on the python project to the new _testclinic_limited

project.

* Save and exit Visual Studio.

* Add `;_testclinic_limited` to `<TestModules Include="...">` in

PCbuild\pcbuild.proj.

* PC\layout\main.py needs updating if you add a test-only extension whose name

doesn't start with "_test".

* Add the extension to PCbuild\readme.txt (this file).

* Build Python from scratch (clean the solution) to check that the new project

is built successfully.

* Ensure the new .vcxproj and .vcxproj.filters files are added to your commit,

as well as the changes to pcbuild.sln, pcbuild.proj and any other modified

files.