python / python/cpython

Build Windows debug extensions without debug Python

Open
#142,760 8 comments 1 reaction 0 assignees View on GitHub

Nobody has claimed this yet.

build OS-windows type-feature
Dominant language
Python
Stars
77.2k
Forks
35.9k
PR merge metrics
PR metrics pending

Description

Feature or enhancement

Proposal:

Summary of a problem

On Windows when users are building a debug version of anything, including Python extensions, they typically provide /MDd/MTd flag to link against the debug version of the C runtime library.

Which automatically sets _DEBUG, which, historically, sets Py_DEBUG macro.

Py_DEBUG macro makes it impossible to link against release version of Python library:

  • because of implicit linking against _d.lib by # pragma comment(lib,"python315_d.lib").
    This restriction was overcomed in Python 3.14 by introduction of Py_NO_LINK_LIB macro.

  • Py_DEBUG also automatically sets Py_REF_DEBUG, which in turn enables additional symbols (namely Py_NegativeRefcount, Py_INCREF_IncRefTotal, Py_DECREF_DecRefTotal), which of course are not present in Release library, so linking Debug build against it is still impossible.
    But if we disable Py_REF_DEBUG when Py_DEBUG is set, then linking succeeds, so there are no more obstacles
    to build debug extensions against release Python.

Typically, users don't want to use debug Python for their extensions, they just want debug version of their own extension, as debug Python binaries are not widely available and even if they are, after build user now would also need debug versions of the entire extensions ecosystem they were using.

So it would be great to provide an option to finally figure this out after all those years.


Changes

There are two approaches to consider (PR currently implements both of them just to showcase the options, but we need to pick one):

  • Py_NO_PY_REF_DEBUG - macro to avoid setting Py_REF_DEBUG based on Py_DEBUG.
    It's a minimal change change to overcome the current problem.
    It seems currently there are no implications when set Py_DEBUG would conflict with unset Py_REF_DEBUG, but Python developers will need to keep it in mind in future that Py_REF_DEBUG might be unset even in debug builds.

  • Py_WIN_NO_PY_DEBUG - that would make _DEBUG not imply Py_DEBUG.
    Honestly, I would prefer this option, as it seems to be a fix for the root cause of the problem and makes flags more consistent between Windows and other platforms - making Py_DEBUG deliberate choice on both Windows and Unix systems.
    _DEBUG implying Py_DEBUG seems to be mostly historical thing introduced awhile ago (https://github.com/python/cpython/commit/9f9fa6d0c937a2cbc51feb3c92212688b09235f0) and they're not closely related.
    Adding Py_WIN_NO_PY_DEBUG also opens an opportunity in the future to completely deprecate _DEBUG setting Py_DEBUG.

Has this already been discussed elsewhere?

I have already discussed this feature proposal on Discourse

Links to previous discussion of this feature:

https://discuss.python.org/t/building-debug-version-of-extensions-on-windows-without-debug-binaries

Linked PRs
  • gh-142761

Contributor guide

Open the contributing guide

First steps

  1. Read the whole issue, then the project's contributing guide.
  2. Comment on the issue to say you are picking it up — it saves two people doing the same work.
  3. Fork the repository and make your change on a branch.
  4. Open a pull request that references the issue number.

Research direction

Start with linked PR gh-142761 and the Discourse discussion, then trace the Py_DEBUG, Py_REF_DEBUG, Py_NO_LINK_LIB, Py_NO_PY_REF_DEBUG, and Py_WIN_NO_PY_DEBUG entry points described here. Done means selecting one proposed macro approach and validating that Windows debug extensions can link against release Python without requiring debug-only symbols.

Written by the indexing model from the issue text.

Assessment

Tech stack
c, python
Domain
build-system
Issue type
Feature
Difficulty
5/5
Estimated time
Over a week
Activity status
Stale
Clarity
Mostly clear
Newbie friendliness
25/100

Get new issues in your inbox

A short digest of beginner-friendly GitHub issues.