Python 3.14.0rc1 #805

xhochy · 2025-07-11T19:47:33Z

Depends on conda-forge/python_abi-feedstock#30

…nda-forge-pinning 2025.07.08.08.15.46

…nda-forge-pinning 2025.07.11.09.44.45

xhochy · 2025-07-11T19:48:29Z

Let's get this building, we shouldn't merge it before we have updated it to a release candidate.

conda-forge-admin · 2025-07-11T19:49:11Z

Hi! This is the friendly automated conda-forge-linting service.

I just wanted to let you know that I linted all conda-recipes in your PR (recipe/meta.yaml) and found it was in an excellent condition.

I do have some suggestions for making it better though...

For recipe/meta.yaml:

ℹ️ The recipe is not parsable by parser conda-souschef (grayskull). This parser is not currently used by conda-forge, but may be in the future. We are collecting information to see which recipes are compatible with grayskull.

_{This message was generated by GitHub Actions workflow run https://github.com/conda-forge/conda-forge-webservices/actions/runs/16779745473. Examine the logs at this URL for more detail.}

This reverts commit b88c21a.

xhochy · 2025-07-21T15:07:42Z

The following error is related to the compiler update and also shows up in Python 3.13:

In function 'memset',
    inlined from 'format_float_short' at /usr/local/src/conda/python-3.14.0b3/Python/pystrtod.c:1167:0:
/home/conda/feedstock_root/build_artifacts/python-split_1753098885527/_build_env/x86_64-conda-linux-gnu/sysroot/usr/include/bits/string3.h:81: warning: call to '__warn_memset_zero_len' declared with attribute warning: memset used with constant zero length parameter; this could be due to transposed parameters [-Wattribute-warning]
/home/conda/feedstock_root/build_artifacts/python-split_1753098885527/_test_env_placehold_placehold_placehold_placehold_placehold_placehold_placehold_placehold_placehold_placehold_placehold_placehold_placehold_placehold_placehold_placehold_placehold_pla/bin/../lib/gcc/x86_64-conda-linux-gnu/14.3.0/../../../../x86_64-conda-linux-gnu/bin/ld: /tmp/ccrEo6GC.lto.o: in function `format_float_short':
/home/conda/feedstock_root/build_artifacts/python-split_1753098885527/_build_env/x86_64-conda-linux-gnu/sysroot/usr/include/bits/string3.h:81:(.text.hot.format_float_short+0x1c8): undefined reference to `__warn_memset_zero_len'
collect2: error: ld returned 1 exit status

xhochy · 2025-07-21T15:29:44Z

Pinning the compiler to GCC 13 leads to the same error, so it could also be related to the sysroot.

…nda-forge-pinning 2025.07.22.16.20.36

xhochy · 2025-07-22T18:57:47Z

@conda-forge/python This is ready for review. We should not merge before the RC is built but I would like to get reviews in as there will only be minimal changes (or none) needed for that.

xhochy · 2025-07-22T19:54:55Z

Actually: Here is the RC ;)

h-vetinari · 2025-07-23T05:44:54Z

recipe/meta.yaml

+        - clang-19         # [not win]
+        - llvm-tools-19    # [not win]
+        - clang 19.*       # [win]
+        - llvm-tools 19.*  # [win]


Python 3.14 provides a new build option --with-tail-call-interp, which provides ~5% or more speedup:

This new interpreter type is an internal implementation detail of the CPython interpreter. It doesn’t change the visible behavior of Python programs at all. It can improve their performance, but doesn’t change anything else.

We could perhaps consider enabling it by default? 🤔

As noted in that section of the release notes, there's a performance-affecting clang bug (llvm/llvm-project#106846), which is still present in clang 19 and will affect the regular interpreter too AFAIU. So in either case we should probably use v20 here.

This interpreter currently only works with Clang 19 and newer on x86-64 and AArch64 architectures. However, we expect that a future release of GCC will support this as well.

So for now, we would enable this only on macOS as we don't use clang to build Python on Linux, right?

More precisely: Only on osx-64 as we don't do PGO on osx-arm64?

AFAIU, PGO is "only" recommended for that option, but not strictly speaking necessary.

So for now, we would enable this only on macOS as we don't use clang to build Python on Linux, right?

Presumably, unless people are comfortable with switching compilers. In principle we could make it another (non-default?) variant, but we already have quite a few on this feedstock...

…nda-forge-pinning 2025.07.23.06.31.04

…nda-forge-pinning 2025.07.23.06.32.39

djhoese · 2025-07-24T21:05:01Z

As an impatiently waiting user, is this only waiting on reviews before merge or is this waiting on the final 3.14 release?

Thanks for everyone's work on this! Very excited!

h-vetinari

LGTM (as core, and active contributor to this feedstock)

h-vetinari · 2025-07-24T21:50:23Z

recipe/conda_build_config.yaml

+# Needs pthread_jit_write_protect_np
+c_stdlib_version:              # [osx and x86_64]
+  - 11.0                       # [osx and x86_64]


Python requiring this is a good reason to revisit conda-forge/conda-forge.github.io#2467

djhoese · 2025-08-04T14:50:29Z

I'm preparing a presentation about free-threading and decided to build this locally while waiting for a merge on my linux_64 system, but when using pip it tells me that SSL isn't available in the build of Python. Is there an easy way to:

Make sure this PR's builds have it.
Make sure it gets added to my build when I do it locally.

Edit: conda install openssl got things working locally.

isuruf · 2025-08-04T17:12:31Z

recipe/patches/0025-Use-zlib-ng.h-on-windows.patch

+From 5248a46154b3ca6bc9b7f3546dfa40c319b16486 Mon Sep 17 00:00:00 2001
+From: "Uwe L. Korn" <[email protected]>
+Date: Mon, 21 Jul 2025 20:03:19 +0200
+Subject: [PATCH 25/25] Use zlib-ng.h on windows


Are we using zlib-ng on windows? I don't see a change in recipe.yaml

isuruf · 2025-08-04T17:14:04Z

recipe/patches/0024-Do-not-define-pid_t-as-it-might-conflict-with-the-ac.patch

-diff --git a/PC/pyconfig.h.in b/PC/pyconfig.h.in
-index 424421f6ff1..2c1b5a5b834 100644
--- a/PC/pyconfig.h.in
-+++ b/PC/pyconfig.h.in
-@@ -230,8 +230,6 @@ typedef _W64 int Py_ssize_t;
- #endif
- #endif /* MS_WIN32 && !MS_WIN64 */

-typedef int pid_t;
-
- /* define some ANSI types that are not defined in earlier Win headers */
- #if _MSC_VER >= 1200
- /* This file only exists in VC 6.0 or higher */


If this is not there anymore, this whole patch can be dropped right?

isuruf · 2025-08-04T17:20:56Z

recipe/conda_build_config.yaml

@@ -1,9 +1,17 @@
 MACOSX_SDK_VERSION:            # [osx and x86_64]
  - 11.0                       # [osx and x86_64]
+# Needs pthread_jit_write_protect_np


pthread_jit_write_protect_np is not in the codebase right?

xhochy added 6 commits July 8, 2025 15:51

python 3.14.0b3

d81e283

MNT: Re-rendered with conda-build 25.5.0, conda-smithy 3.51.0, and co…

b20a5ea

…nda-forge-pinning 2025.07.08.08.15.46

python 3.14.0b3

7848655

MNT: Re-rendered with conda-build 25.5.0, conda-smithy 3.51.0, and co…

3a29243

…nda-forge-pinning 2025.07.11.09.44.45

Update clang and osx

6655464

MNT: Re-rendered with conda-build 25.5.0, conda-smithy 3.51.0, and co…

13a3d9e

…nda-forge-pinning 2025.07.11.09.44.45

xhochy requested review from chrisburr, isuruf, jakirkham, katietz, mbargull, msarahan, ocefpaf, pelson and scopatz as code owners July 11, 2025 19:47

xhochy and others added 5 commits July 21, 2025 08:25

Skip named aliases for now

b88c21a

Cannot mix freethreading and JIT

d4ff52d

Depend on zlib-ng on Windows

97cb05a

Revert "Skip named aliases for now"

03f415a

This reverts commit b88c21a.

use the same version in requirements as for outputs

113008e

xhochy added 6 commits July 21, 2025 15:52

Depend on stdlib for static tests

d5d1906

Require stdlib in all tests where a compiler is present

33221df

Don't depend on zlib on Windows

481da7f

win: freethreading and jit don't work together

8805435

python 3.14.0b4

7adadc0

Use zlib-ng.h on Windows

ecf4e65

xhochy added 7 commits July 22, 2025 19:21

MNT: Re-rendered with conda-build 25.5.0, conda-smithy 3.51.1, and co…

ff91f09

…nda-forge-pinning 2025.07.22.16.20.36

pyconfig.h is no longer generated

1b21a86

Limit to win

6a3f5bb

MNT: Re-rendered with conda-build 25.5.0, conda-smithy 3.51.1, and co…

6ecabc9

…nda-forge-pinning 2025.07.22.16.20.36

Fix TZPATH patch

2a4674f

Back to full matrix

1b78ec8

MNT: Re-rendered with conda-build 25.5.0, conda-smithy 3.51.1, and co…

93321db

…nda-forge-pinning 2025.07.22.16.20.36

python 3.14.0rc1

10d22d0

h-vetinari changed the title ~~Python 3.14.0b4~~ Python 3.14.0rc1 Jul 23, 2025

h-vetinari reviewed Jul 23, 2025

View reviewed changes

xhochy mentioned this pull request Jul 23, 2025

Add Python 3.14 migration conda-forge/conda-forge-pinning-feedstock#7598

Open

xhochy added 4 commits July 23, 2025 08:34

Enable tail-call-interp on osx

be701a0

MNT: Re-rendered with conda-build 25.5.0, conda-smithy 3.51.1, and co…

9f39e7a

…nda-forge-pinning 2025.07.23.06.31.04

Less strict clang pin

160fed9

MNT: Re-rendered with conda-build 25.5.0, conda-smithy 3.51.1, and co…

23cd7d6

…nda-forge-pinning 2025.07.23.06.32.39

h-vetinari approved these changes Jul 24, 2025

View reviewed changes

isuruf reviewed Aug 4, 2025

View reviewed changes

isuruf added 8 commits August 6, 2025 07:18

remove stdlib version change

52d195b

fix patch

0d6d0fb

no signature

9788287

update pid_t patch and remove zlib-ng patch

db281a1

Update 0011-Make-dyld-search-work-with-SYSTEM_VERSION_COMPAT-1.patch

fcdbb49

backport gh137211

9073cc6

update pid_t patch

3791cdb

fix patch again

a87dcd9

Uh oh!

Python 3.14.0rc1 #805

Are you sure you want to change the base?

Python 3.14.0rc1 #805

Uh oh!

Conversation

xhochy commented Jul 11, 2025

Uh oh!

xhochy commented Jul 11, 2025

Uh oh!

conda-forge-admin commented Jul 11, 2025 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

xhochy commented Jul 21, 2025

Uh oh!

xhochy commented Jul 21, 2025

Uh oh!

xhochy commented Jul 22, 2025

Uh oh!

xhochy commented Jul 22, 2025

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

djhoese commented Jul 24, 2025

Uh oh!

h-vetinari left a comment

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

djhoese commented Aug 4, 2025 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Uh oh!

conda-forge-admin commented Jul 11, 2025 •

edited

Loading

djhoese commented Aug 4, 2025 •

edited

Loading