Tracking issue for OpenOptionsExt on Windows

[alexcrichton]

This is a tracking issue for the unstable open_options_ext feature in the standard library. Currently Unix has a stable extension trait for setting the mode of a file but Windows has not stabilized any of its functionality. The current set of methods on the Windows extension trait just provide the ability to override any of the arguments to CreateFile.

The open questions here are:

• Is the naming right?
• Are the types right? (e.g. should they be updated to use std::os::windows::raw?)
• Do more options want to be specified?
• Does Unix want any more methods on OpenOptionsExt?

APIs:

• OpenOptionsExt::access_mode
• OpenOptionsExt::share_mode
• OpenOptionsExt::attributes
• OpenOptionsExt::security_qos_flags

[retep998]

For CreateFile2 (which Rust would have to use if we wanted to support non-desktop apps since CreateFile is desktop only) it takes flags and attributes as two separate parameters, so it may be a good idea to separate those and then when we invoke CreateFile just combine them back together.

[retep998]

u32 is very much the right type to use, although whether it should be behind some sort of alias is an entirely different debate and would be a backwards compatible change to make so it isn't an issue regardless.

There are more options that we will want to eventually support.

[alexcrichton]

Interesting! Perhaps we should call CreateFile2 unless it's not available? (e.g. Windows 7 and below).

Also yeah u32 is fine, I figure we may want to take a DWORD alias though (although also perhaps not)

[kamalmarhubi]

Does Unix want any more methods on OpenOptionsExt?

I'd like it if Unix exposed the flags parameter to open.

[kamalmarhubi]

Additionally, it would be nice to have a way to toggle particular flags as opposed to just overriding all of them.

[alexcrichton]

🔔 This issue is now entering its cycle-long final comment period for stabilization 🔔

The libs team would like to perform one final pass to ensure that all of these arguments have very clearly documented interactions with the underlying arguments and other builder options, but otherwise these seem good for stabilizing!