Nathan Goldbaum

FWIW, the problem is even worse for object string arrays: ``` >>> hashlib.sha256(np.asarray(['foo'], dtype=object)).hexdigest() '1eba5cdda7e69e88b33286bc8b70ea7a8a80de0ae49eab51b54a3aa0e365f2fe' >>> hashlib.sha256(np.asarray(['foo'], dtype=object)).hexdigest() 'cd2f7e615ead16439899efa31ec330d63599a54067e1822deafa7ab5e2e1801b' >>> hashlib.sha256(np.asarray(['foo'], dtype=object)).hexdigest() '1eba5cdda7e69e88b33286bc8b70ea7a8a80de0ae49eab51b54a3aa0e365f2fe' >>> hashlib.sha256(np.asarray(['foo'], dtype=object)).hexdigest() 'cd2f7e615ead16439899efa31ec330d63599a54067e1822deafa7ab5e2e1801b' ``` Also this...

We talked about this at the community meeting, which led me to creating https://github.com/numpy/numpy/issues/29229 right afterwards after we collectively decided that hashing an array using `hashlib` has some corner cases...

For free-threading, I'd like to see my PRs (#4523, #4528, #4544, and #4546) as well as a few others (#4439, #4566, and #4512) merged. I think most of those are...

@davidhewitt this can be closed, right?

I can't find any discussion about bytearray and free-threading in the CPython issue tracker, you may want to file an issue, especially if you can make a pure-python reproducer using...

Hopefully a future Python release will fix the thread safety issues you identified and we can at least make the free-threaded build have similar guarantees compared with the GIL-enabled build.

bytearray is now thread-safe, but the fixes are going to have to wait until 3.14 I think. There are a number of thread-safety fixes that only live in the `main`...

I don't understand this error. Out of curiosity, can you compile CPython itself using this version of the intel compilers? The code we're using was adapted from the CPython `pyatomic.h`...

``` In file included from ../numpy/_core/src/common/common.hpp(10), from ../numpy/_core/src/npymath/halffloat.cpp(12): ../numpy/_core/src/common/half.hpp(45): error: expected a ")" ); ``` Looking at `half.cpp`, there's some sort of complicated preprocessor magic happening right above line 45,...

For the `half.hpp` error, I don't see how you can be hitting an error about an unclosed paren. No matter what the preprocessor macros are set to, I don't think...