GeezerVison

@chris21k can you comment as to whether this solves your issue?

@msprotz Thanks for looking into it and fixing it.
Unfortunately it is somehow hard to test it for me, or could you give me some hints? Currently I am using Ubuntu 24.04 standard python 3.12 installation...

However, I can give you detailed but easy instructions to test it on your fixed python installation (you just have to apdapt your path to the python installation containing the PR):

First we create a simple python program. Using hashlib or not I think plays no role. However, we take the following one:

import hashlib

print("Hello, World!")

m = hashlib.sha256()
m.update(b"nobody inspects")
print(m.hexdigest())

Then we do the follwing steps:

$ ls
helloworld.py
$ /usr/bin/python3 helloworld.py
Hello, World!
5880e257e70b83842737244b412a2f282e0f230953027a22104efc615d90d3f9
$ /usr/share/doc/python3.12/examples/freeze/freeze.py -o frozen helloworld.py
$ cd frozen
$ #mkdir -p Modules/_hacl; touch Modules/_hacl/libHacl_Hash_SHA2.a # should not be necessary any more
$ make
$ ./helloworld
Hello, World!
5880e257e70b83842737244b412a2f282e0f230953027a22104efc615d90d3f9

If make runs through without errors and creates the helloworld binary then everthing should be all right.

Sorry for the late reply.

Given that this pull request has been open for several months, I kindly request the maintainers consideration in reviewing and merging it. Implementing these changes would be beneficial for developers who rely on a simplified and robust build process.

Thank you for your attention.

My issue here is that I am not familiar with freeze.py, and since the time I submitted this PR, the build system of Python gained another .a file (for Hacl's Blake2), so it sounds like we should fix freeze.py rather than undo the changes to the build system and no longer rely on .a files.

Maybe @picnixz or @chris-eibl know about freeze.py? Also pinging @gpshead FYI

Sorry, I am not familiar with freeze.py

And me neither :( maybe @vstinner?

I'm closing this one in favor of #132438 (since this one also has conflicts). Thanks for the initial work though!