Osyotr

> The thing is that std::wstring does not work on Linux This is incorrect. It has different encoding (UTF-32 instead of UTF-16), but it absolutely does work.

The problem with `std::string` is that on Windows it's hard to tell what encoding the data is in. It may be UTF-8, but may very well be `CP_ACP` that depends...

Base64/SHA1 stuff is used in other places.

> Are there other better/faster alternatives? IMHO it's not worth adding an external dependency over these two algorithms. It's possible to rip out code from https://github.com/weidai11/cryptopp (BSL-1.0).

> Thanks! What's the best/simplest can be used? Depends on the usage. Non-cryptographic: xxhash64 Cryptographic: blake3 https://jolynch.github.io/posts/use_fast_data_algorithms/

> We don't need cryptographic strength here. Would it be hard to integrate https://xxhash.com/ ? Here's drop-in replacement: `3party/CMakeLists.txt`: ```cmake include(FetchContent) FetchContent_Declare( xxHash GIT_REPOSITORY https://github.com/Cyan4973/xxHash.git GIT_TAG bbb27a5efb85b92a0486cf361a8635715a53f6ba # v0.8.2 SOURCE_SUBDIR...

> all values are processed, doesn't matter it is enum or enum class Not correct. Compilers can't reliably prove that `m_Op` is not something like `OP_WRITE_EXISTING | OP_APPEND`, so the...

> 1. Why presets are needed? What pros and cons of config duplication there? Can we avoid them? I find it hard to explain it better than CMake's own [docs](https://cmake.org/cmake/help/latest/manual/cmake-presets.7.html)...

`--x-install-root`?

Duplicate #38047?