freeze cannot find Modules/_hacl/libHacl_Hash_*.o

[chris21k]

Bug report

Bug description:

Dear maintainers,

unfortunately, the fix of issue #116043 in Python 3.14 does not fix the problem at all. Now freeze.py adds dependencies to some Modules/_hacl/libHacl_Hash*.o files,
although their functionality is already contained in cpython. The result again ist that linking is not possible:

/usr/bin/x86_64-linux-gnu-ld.bfd: cannot find Modules/_hacl/Hacl_Hash_Blake2s.o: No such file or directory
/usr/bin/x86_64-linux-gnu-ld.bfd: cannot find Modules/_hacl/Hacl_Hash_Blake2b.o: No such file or directory
/usr/bin/x86_64-linux-gnu-ld.bfd: cannot find Modules/_hacl/Lib_Memzero0.o: No such file or directory
/usr/bin/x86_64-linux-gnu-ld.bfd: cannot find Modules/_hacl/Hacl_Hash_Blake2s_Simd128.o: No such file or directory
/usr/bin/x86_64-linux-gnu-ld.bfd: cannot find Modules/_hacl/Hacl_Hash_Blake2b_Simd256.o: No such file or directory
/usr/bin/x86_64-linux-gnu-ld.bfd: cannot find Modules/_hacl/Hacl_Hash_MD5.o: No such file or directory
/usr/bin/x86_64-linux-gnu-ld.bfd: cannot find Modules/_hacl/Hacl_Hash_SHA1.o: No such file or directory
/usr/bin/x86_64-linux-gnu-ld.bfd: cannot find Modules/_hacl/Hacl_Hash_SHA2.o: No such file or directory
/usr/bin/x86_64-linux-gnu-ld.bfd: cannot find Modules/_hacl/Hacl_Hash_SHA3.o: No such file or directory
collect2: error: ld returned 1 exit status

See attached error-log.txtfor the full compilation log. Thus, only again the crude hack creating all these files as empty files helps.

Here the steps to reproduce:

$ lsb_release -a
No LSB modules are available.
Distributor ID: Ubuntu
Description:    Ubuntu Resolute Raccoon (development branch)
Release:        26.04
Codename:       resolute
$ python3 --version
Python 3.14.3
$ cat helloworld.py
#!/usr/bin/python3

import hashlib

print("Hello, World!")

m = hashlib.sha256()
#m = hashlib.sha3_224()
m.update(b"nobody inspects")
print(m.hexdigest())
$ /usr/share/doc/python3.14/examples/freeze/freeze.py -X setuptools -o frozen helloworld.py
$ cd frozen
$ make &> error_log.txt                              <-- see attachement to this issue
$ mkdir -p Modules/_hacl
$ touch Modules/_hacl/Hacl_Hash_Blake2s.o
$ touch Modules/_hacl/Hacl_Hash_Blake2b.o
$ touch Modules/_hacl/Lib_Memzero0.o
$ touch Modules/_hacl/Hacl_Hash_Blake2s_Simd128.o
$ touch Modules/_hacl/Hacl_Hash_Blake2b_Simd256.o
$ touch Modules/_hacl/Hacl_Hash_MD5.o
$ touch Modules/_hacl/Hacl_Hash_SHA1.o
$ touch Modules/_hacl/Hacl_Hash_SHA2.o
$ touch Modules/_hacl/Hacl_Hash_SHA3.o
$ make
$ ./helloworld
Hello, World!
5880e257e70b83842737244b412a2f282e0f230953027a22104efc615d90d3f9

Many thanks,
Chris

Attachements
error-log.txt

CPython versions tested on:

3.14

Operating systems tested on:

Linux

[ZeroIntensity]

cc @picnixz

[picnixz]

I think I was only aware of the issue for HMAC, which should be fixed but please tell me if using hmac also fails. IIRC, we are linking them statically so I don't know why we are trying to find a .o (but maybe hashlib isn't doing it actually). Can you also expand error_log.txt in a <details> section please? It's easier for people that just read the issue. TiA.

I unfortunately don't have time to address this issue for now because of my daily job and my travels. We have quite a lot of issues with the build system because HMAC requires some files to exist and there are also other subtleties due to the configure options and CPUID capabilities.