is_windows_10 documentation somewhat incorrect

[DHowett]

Hi! Windows Console Subsystem maintainer here :)

is_windows_10 is documented implying that you cannot use ENABLE_VIRTUAL_TERMINAL_PROCESSING without first checking that you're on Windows 10, and goes to great lengths to explain how versioning is handled.

None of that is strictly necessary! If you call SetConsoleMode(..., ENABLE_VIRTUAL_TERMINAL_PROCESSING) and the console host does not support it, regardless of the version of Windows, it will give you a status code and ignore you.

I know we haven't made this very easily discoverable. Sorry :/.

That means you can collapse all your version checking logic into "does the console host support VT processing?" In so doing, you get is_microsoft_terminal for free (since it does support VT processing) and the handful of custom console hosts that exist for Windows 8 which also support VT processing. You can collapse every micosoft_terminal and Windows case to "get mode, check VT_PROCESSING, try to set it, bail out¹ if it's rejected."

To put a finer point on it... here's a console on Windows 8.1 displaying Sixels, a nominally "Windows 11+" feature which requires VIRTUAL_TERMINAL_PROCESSING (side by side with one that doesn't support that mode!)

Footnotes

1. bail out and do it the legacy way, FWIW. ↩

[passcod]

Oh, cool! If you want to PR that, would be appreciated, otherwise I'll get to it eventually. Thanks very much for explaining :)

[DHowett]

Thanks!

I started work on it over here, and left myself a pretty big TODO list: da08c5f (in the commit body)

I think the main "pivot" I'll need to understand is whether exposing WindowsVtClear for your public consumers is meaningful; in every way on my branch it is the same as XtermClear.

I'm not the best at idiomatic Rust, so I wouldn't expect an actual PR for a while, but I'm happy to keep working on it :)

[passcod]

I think having XtermClear and XtermReset do something else than documented on Windows if VT isn't supported would be surprising. I think probably the way to go is to have clear() call win::vt() like you've done, and remove both WindowsVtClear and WindowsVt entirely.

[DHowett]

ah, I may have been overly focused on keeping the public API unchanged :) thanks!