Internal linkage for _obf_date_hash() to prevent date/time macro issues with string obfuscation

[Dooomsickle]

Currently, _obf_date_hash() is only declared as constexpr which makes it implicitly inline and therefore only one instance is created per-compilation. This causes significant issues with string obfuscation.

When compiling in parallel, or when compiling single objects instead of the entire project, the __DATE__ and __TIME__ expansions in the single instance of _obf_date_hash() may not match the one used to create the key for a newly-compiled obfuscated string, breaking the decryption.

Example:

// a.cpp
#include "AES8.hpp"
const wchar_t* get_a() {
    static auto s = OBFUSCATE_WSTRING(L"hello from A");
    return s.c_str(); // may decrypt correctly or garbage depending on link order
}

// b.cpp
#include "AES8.hpp"
const wchar_t* get_b() {
    static auto s = OBFUSCATE_WSTRING(L"hello from B");
    return s.c_str(); // same — one of these will be wrong
}

Depending on which file gets compiled first, only one those strings may decrypt correctly because it's using its own hash function, while the other fails.

Fix

Add static to the declaration of constexpr uint32_t _obf_date_hash() -> static constexpr uint32_t _obf_date_hash()

This way, each compilation unit will get its own, correct definition of the function with the same TIME and DATE used to create the encryption key.