Introduce and use SWIFT_HAVE_LIBEDIT

[hughbe]

This is an optional dependency (i.e. we can build the compiler without it).

I went for the CheckIncludeFiles CMake option, as this is simpler than having to adding a find_package implementation.

Extracted from #5904

[compnerd]

Does the repl work without it?

[compnerd]

@swift-ci please test

[hughbe]

The repl is automatically disabled for platforms without histedit.h

[swift-ci]

Build failed
Jenkins build - Swift Test OS X Platform
Git Commit - 877900bc042b3ac85abbed3b33df093fce62a5cd
Test requested by - @compnerd

[compnerd]

CC @jrose-apple (in case you have a strong opinion on this, I think this is reasonable).

[jrose-apple]

@mxswd just did something like this…or maybe it was limited to xpc.h. Max?

[mxswd]

Yeah I just added 81ae0fd
Which worked for me. Without this what error do you get on Windows?

[hughbe]

Oh, this may have been superseeded! I extracted this from the main PR - will let you know if this still doesn't work.. Will close if not

[swift-ci]

Build failed
Jenkins build - Swift Test Linux Platform
Git Commit - 877900bc042b3ac85abbed3b33df093fce62a5cd
Test requested by - @compnerd

[hughbe]

Actually!! I remember now - this is still necessary.

This is because we add target_link_libraries(edit). edit.lib is NOT available by default on Windows - this PR fixes that! This caused linker errors.

@compnerd could you retrigger the tests - I fixed a typo causing the builds to fail

[jrose-apple]

Still seems like we should be able to use the same variable in both cases.

[hughbe]

Still seems like we should be able to use the same variable in both cases.

True, good idea. I've moved the variable to CMakeLists.txt. I've found that the order of CMakeLists.txt resolving is unpredictable - its best to move this kind of check (for an optional package) to the root file, where it will always be run before target_link_libraries(edit)

[jrose-apple]

@swift-ci Please smoke test