Hmm, I kind of don't want to take this because the windows module code is just so crappy? I think we should almost completely scrap the windows module and rebuild it from first principles to implement the now() and naive_to_local() functions with minimal use of the Windows API (and definitely avoiding all the Unixisms that are there now, like Timespace, Tm and a lot of strange indirection. Should probably also avoid the panic!() in call() in favor of yielding LocalResult::None.

Would you be interested in working on that? See also #989 which fixes other silliness in windows.

I don't know if it's exactly the way to go, but I did get a little OCD over the weekend after removing the panics and tried to clean up the module on this branch. Not sure if it's exactly the direction to go as it is mostly just cleaning up the current module and creating wrappers around windows_sys.

A bit of OCD is probably a good thing for this! Your branch seems to still maintain the Timespec and Tm which I think are complete distractions from the task at hand. Also your branch is targeting the main branch which uses windows-sys, but ideally we'd first land this on 0.4.x which still uses winapi (and then merge a similar change that uses windows-sys to main later). IMO the goal is to make the windows module as concise as possible, translating now() and naive_to_local() as directly as possible to Win32 APIs.

I think the changes here have been addressed by #992?

I think the changes here have been addressed by #992?

Looks like @nekevss covered this in #992 (including the test_from_naive_date_time_windows). Let's defer to that one. This PR should be Closed.