fix: TryLoginAsync reported success inverted

[binn]

The bug

TryLoginAsync passes user == null as the success argument of VRChatLoginResult:

return new VRChatLoginResult(user == null, null);
//                           ^^^^^^^^^^^^ this is `bool success`

So the result is inverted on both paths that don't throw:

Outcome	user	Reported Success	Correct?
Login succeeded	non-null	false	❌
LoginAsync returned null	null	true	❌
LoginAsync threw	—	false	✅

A successful login reports failure with a null Exception. That's indistinguishable from a silent failure, so the natural conclusion is bad credentials — which is how I found it, after a while of being certain my credentials were wrong.

VRChatLoginResult's own documentation is unambiguous about the intent:

success: A value indicating whether the login operation was successful. Set to true if the login succeeded; otherwise, false.

The fix

user != null for success, plus a real exception on the failure path.

The second part is worth a note: LoginAsync returns null when the final GetCurrentUser response isn't 200 OK, and discards the status. Success = false, Exception = null left a caller with nothing to act on, log, or report to a user. It now carries an UnauthorizedAccessException so the failure is at least nameable.

Notes

• Only wrapper/VRChat.API/Client/VRChat.cs is touched — no generated code under src/.
• Behaviour change is intentional and is the point of the PR: anyone who worked around the inversion by testing !result.Success will need to drop the negation. Given the current behaviour makes TryLoginAsync unusable as documented, I'd expect most callers to be using LoginAsync directly instead.
• Builds clean against src/VRChat.API/VRChat.API.csproj in Release.

Not fixed here, but related

LoginAsync swallowing the status makes 401, 403, 429 and a Cloudflare WAF block indistinguishable to a caller — they all surface as null. That matters for anything that needs to respond differently to each (re-auth vs back off vs tell the operator). Happy to open a separate PR if you'd like that changed, since it's a larger behavioural question than this one.

🤖 Generated with Claude Code

https://claude.ai/code/session_015DPXtTDGbCQ6GebnuvxV2s