Windows Port

[gmittert]

This allows IndexStoreDB to build on Windows.

Once the other PRs go in, this enables compilation on Windows. I have some CMake locally which I've been using to build this, but it should work just as well with SwiftPM when that is ready on Windows.

[gmittert]

@swift-ci please test

[benlangmuir]

Instead of creating a new header/cpp for this, let's inline these implementations. If we someday need to use them in more places we can expose them, but under more idiomatic API names (e.g. getCanonicalPath, not realpath).

[gmittert]

I've removed the CrossPlatform.[h/cpp] and instead inlined the changes.

@swift-ci please test

[benlangmuir]

I left some comments on the canonical path implementation, but actually: can we just use llvm's sys::fs::real_path?

[gmittert]

I left some comments on the canonical path implementation, but actually: can we just use llvm's sys::fs::real_path?

Ugh, that was really leaky. I've changed it to use llvm real_path. Like the Unix realpath, if PathBuf is NULL, it'll allocate memory and the target of the stringref would have to be freed. Thoughts on just having it return a std::string? I believe it's only ever used on Path.cpp:86.

[benlangmuir]

Thoughts on just having it return a std::string? I believe it's only ever used on Path.cpp:86.

Ah, I meant why not replace this whole function with a call to llvm::sys::fs::real_path? Instead of passing in a PathBuf char array we can use a SmallString<PATH_MAX>. It will have an extra string copy compared to the current implementation on macOS/Linux, but won't have to heap allocate for any platform.

[gmittert]

Ah, I meant why not replace this whole function with a call to llvm::sys::fs::real_path? Instead of passing in a PathBuf char array we can use a SmallString<PATH_MAX>. It will have an extra string copy compared to the current implementation on macOS/Linux, but won't have to heap allocate for any platform.

Oh I see, yes, that makes much more sense in that context.

@swift-ci please test

[benlangmuir]

Not sure where the PR test status went, but it's still running apparently https://ci.swift.org/job/swift-indexstore-db-PR-macOS/24/ https://ci.swift.org/job/swift-indexstore-db-PR-Linux/25/