Test Python 3.14

[DimitriPapadopoulos]

No description provided.

[DimitriPapadopoulos]

It looks like hatch itself is failing with Python 3.14:

│ ❱ 199 │   │   for context in app.runner_context([selected_envs[0]]):         │

╰──────────────────────────────────────────────────────────────────────────────╯
IndexError: list index out of range

I cannot find a relevant hatch issue. I'll rebease to make sure we use the newly released Python 3.14.0 instead of 3.14.0-rc.3.

[ofek]

I think this needs to be set since there has not yet been a release that modifies the default https://hatch.pypa.io/latest/config/internal/testing/#matrix

[ofek]

Well, at least we got past that error!

[DimitriPapadopoulos]

Strange that this error only occurs with Python 3.14:

      In file included from src/curve.c:1:
      src/curve.h:4:10: fatal error: 'gmp.h' file not found
          4 | #include "gmp.h"
            |          ^~~~~~~
      1 error generated.

• Cannot install in macOS BigSur (M1 chip)
• Cannot link gmp: library not found for -lgmp

export LIBRARY_PATH=$LIBRARY_PATH:/opt/homebrew/lib
export INCLUDE_PATH=$INCLUDE_PATH:/opt/homebrew/include

Looks like a build issue of fastecdsa itself,

[acul71]

In file included from src/curve.c:1:
src/curve.h:4:10: fatal error: 'gmp.h' file not found
4 | #include "gmp.h"
| ^~~~~~~
1 error generated.

See if this helps

https://py-libp2p.readthedocs.io/en/latest/contributing.html#macos-setup

brew install cmake pkgconfig gmp

# On macOS, help the build command find and link against the gmp library:

CFLAGS="`pkg-config --cflags gmp`" LDFLAGS="`pkg-config --libs gmp`" python3 -m pip install -e ".[dev]"

#############

Hello I'm Luca from py-libp2p waiting for the `coincurve` new release for python v3.14 (needed for windows)

[DimitriPapadopoulos]

This had already been pointed out in #203 (comment).

The whole remaining question is where exactly to define these environment variables.

[DimitriPapadopoulos]

Not sure this is the best location for defining these variables, but I cannot find anything better. Suggestions?

[ofek]

This doesn't appear to be valid syntax, how is this working? 😅

[DimitriPapadopoulos]

I guess the parser is permissive. Fixed.

[ofek]

Try removing this because of the new release https://github.com/pypa/hatch/releases/tag/hatch-v1.15.0

[ofek]

This doesn't appear to be valid syntax, how is this working? 😅

[ofek]

Fixing the syntax error appears to have broken the build somehow, wow! Can you add that back and print out those environment variables? As you mentioned, the parser must be lenient because I just tried https://raw.githubusercontent.com/ofek/coincurve/6bb0ae22705d0cef2e2774ddfc66f6922b729b16/.github/workflows/build.yml:

>>> import pyperclip,yaml
>>> s=yaml.safe_load(pyperclip.paste())
Traceback (most recent call last):
  File "<stdin>", line 1, in <module>
...
  File "C:\Users\ofek\AppData\Local\Programs\Python\Python312\Lib\site-packages\yaml\scanner.py", line 291, in stale_possible_simple_keys
    raise ScannerError("while scanning a simple key", key.mark,
yaml.scanner.ScannerError: while scanning a simple key
  in "<unicode string>", line 41, column 3:
      LIBRARY_PATH /opt/homebrew/lib
      ^
could not find expected ':'
  in "<unicode string>", line 42, column 3:
      INCLUDE_PATH /opt/homebrew/include
      ^

[DimitriPapadopoulos]

Fixed by using CFLAGS and LDFLAGS instead of INCLUDE_PATH and LIBRARY_PATH.

[ofek]

Nice, thanks a lot! Do you think we could set those conditionally just on the jobs that were failing?

[DimitriPapadopoulos]

This really is a workaround that I hope to get rid of, see AntonKueltz/fastecdsa#114. And I still fail to understand why the build fails with Python 3.14 only. I suspect Python 3.14 might pull a different version of a dependency. So I'm not certain we should set the variables conditionally.

[DimitriPapadopoulos]

I've moved the definition of the environment variables to the specific step that needs them:

coincurve/.github/workflows/build.yml

Lines 65 to 66 in dbac6b6

	- name: Run static analysis
	run: hatch fmt --check

However, I haven't found a way to define the variables conditionally.

[DimitriPapadopoulos]

Is there a way to retrieve the details of what hatch fmt --check does under the hood? The difference between Python 3.13 and 3.14 originates in there, but the output is not verbose enough to observe what could be the cause.

[ofek]

hatch -v ... should do it

[DimitriPapadopoulos]

Unfortunately hatch -v ... seems to fix the issue for some reason - unless updates deep inside GitHub Actions or runners fixed the issue.

I'll try again without -v.

[DimitriPapadopoulos]

@ofek Somehow, macOS issues fixed themselves. I guess a bug has been fixed somewhere in CI software/runners.

[ofek]

Thanks for persevering through the weirdness!

[ofek]

Interestingly, master broke https://github.com/ofek/coincurve/actions/runs/18672653134/job/53236594428

[ofek]

I found the cause of the strange behavior:

• https://github.com/ofek/coincurve/actions/runs/18672653134/job/53236594428#step:11:40
• https://github.com/ofek/coincurve/actions/runs/18654992495/job/53181796080#step:11:45

[DimitriPapadopoulos]

I'm not sure how to interpret this:

• The exact reasons behind cache hits/misses are irrelevant — we should be able to handle both any way.
• I'm not sure how cache hits/misses have an impact on finding the GMP library, or not. What's in a cache? Does the cache include environment variables such as CFLAGS/LDFLAGS?
• Why the failure with Python 3.14 only? Because it's the only case missing a GitHub Action cache? But then, how did it work in the first place with other versions of Python?

[DimitriPapadopoulos]

Any way, I reapplied 117ffab in #214.