GeezerVison

@chris-eibl can I take some of your time to request a review please :)?

Oh no, it's back again - this was already a back and forth (#130040).

I think, your fix back then (exclude __clang__)

#if defined(MS_WINDOWS) && !defined(__clang__)
#define _GENERATE_DEBUG_SECTION_WINDOWS(name)                       \
   _Pragma(Py_STRINGIFY(section(Py_STRINGIFY(name), read, write))) \
   __declspec(allocate(Py_STRINGIFY(name)))

should stay: clangcl might not understand __declspec(allocate(...)).

But then, no GENERATE_DEBUG_SECTION is in effect, because

#if defined(__linux__) && (defined(__GNUC__) || defined(__clang__))
#define _GENERATE_DEBUG_SECTION_LINUX(name) \
   __attribute__((section("." Py_STRINGIFY(name))))               \
   __attribute__((used))

will never fire due to __linux__. Maybe rewrite that to

#if (defined(__linux__) && defined(__GNUC__)) || (defined(MS_WINDOWS) && defined(__clang__))
#define _GENERATE_DEBUG_SECTION_LINUX(name) \
   __attribute__((section("." Py_STRINGIFY(name))))               \
   __attribute__((used))

clang defines __GNUC__, so I dropped the first __clang__

Ah - while I was typing you already had the same idea :)

Well I have absolutely no clue what's going on with Windows here. Will sleep on it and let someone else take over. In the meantime, will just use this PR to upgrade the LLVM version.

@Fidget-Spinner: I've created #132112 to fix GENERATE_DEBUG_SECTION for clangcl on Windows. CI is green. Feel free to incorporate it in this PR if you want - I'll then just close it.

@chris-eibl can you please send me an email (it's listed on my GH profile), I want to ask some questions.

Email sent :)