GlobalUnlock return value seems weird

[knopp]

Summary

It is translated to a result, but in both case GetLastError returns NO_ERROR so it seems a bit weird.

https://learn.microsoft.com/en-us/windows/win32/api/winbase/nf-winbase-globalunlock

Crate manifest

No response

Crate code

No response

[kennykerr]

There are cases where the function can fail, in which case GetLastError reports that information.

[knopp]

The problem is that false in case of GlobalUnlock does not necessarily mean that error has happened, but windows-rs will return Error unconditionally, even if GetLastError is 0. So something like

GlobalUnlock(hglobal)?

will abort calling function with an error, which is in fact a success.

[tim-weis]

GlobalUnlock() doesn't follow the conventional error reporting pattern, and the Result<> transformation carries the surprises over into Rust.

The API should probably get the [CanReturnMultipleSuccessValues] or [CanReturnErrorsAsSuccess] attribute to suppress this transformation here. I don't recall the rationale behind choosing one over the other.

[kennykerr]

Starting to think the default for the return value transformation is all wrong. The transformation should be opt-in rather than opt-out for Win32.

[kennykerr]

Anyway, I'll transfer to the Win32 metadata repo for their consideration.

Rust only considers the CanReturnMultipleSuccessValues attribute.