The typical initialization of PyMutex in its comment document cannot be compiled on Windows

[aisk]

cpython/Include/internal/pycore_lock.h

Lines 28 to 29 in 2939ad0

	// Typical initialization:
	// PyMutex m = (PyMutex){0};

But it cannot be compiled outside a function on Windows: https://github.com/python/cpython/actions/runs/7862752441/job/21452578591 with this error:

`3>D:\a\cpython\cpython\Modules\rotatingtree.c(19,16): error C2099: initializer is not a constant [D:\a\cpython\cpython\PCbuild\pythoncore.vcxproj]` 

Also check here be0427f#diff-266bed9d854389a88ddd0696e9a225a2570559901edf0e909de43dd1f49ce80aR19 to see the usage and the compile error.

Change it to PyMutex m = {0}; can resolve this issue.

Linked PRs

• gh-115304: Add doc for initialize PyMutex as global variables #115305

[terryjreedy]

I don't understand the links above. pycore_lock.h, 28-29, are as listed, but have nothing obvious to do with commit 2939ad0 for a pydoc fix. Neither do those lines have any obvious connections with the compile error in random_value *= 1082527; in rotatingtree.c. The build report is not obviously linked back to a particular PR.

[aisk]

Hi @terryjreedy the compile error in the link should be:

3>D:\a\cpython\cpython\Modules\rotatingtree.c(19,16): error C2099: initializer is not a constant [D:\a\cpython\cpython\PCbuild\pythoncore.vcxproj] (I'll update this to the first comment, sorry I forgot it yesterday)

And change the usage to PyMutex m = {0}; got compile passed on Windows, see the commits in #115301

commit 2939ad0 have no connections to this issue, I guess you mixed this with other issues?