Route libgdx-port checks through the target GUI

[MostCromulent]

Bug report

Resolves #10613 — On the mobile client against a desktop host, search-your-library effects (Demonic Tutor, Solemn Simulacrum's basic-land trigger, fetch lands) present prompt but no cards are visible.

The general issue

Several UI decision points read GuiBase.getInterface().isLibgdxPort() to determine whether to route to desktop or mobile rendering paths.

This is a process-global static answering "is this JVM libgdx?" — when the right question is "is the player whose UI will render this libgdx?". The two answers agree in single-player and same-platform netplay, but diverge in cross-platform netplay: a desktop host serving a mobile client returns false and steers the choice down a desktop-renderer code path the mobile client cannot honor.

How it manifests in the tutor case

PlayerControllerHuman.chooseSingleEntityForEffect (line 585)
                        │
                        ▼
        ┌──────────────────────────────────────┐
        │ useSelectCardsInput(optionList, sa)  │
        │                                      │
        │  Reads GuiBase.getInterface()        │
        │       .isLibgdxPort()                │
        │  ── PROCESS-LOCAL on the HOST ──     │
        │                                      │
        │  Library zone qualifies only when    │
        │  !isLibgdxPort()  (PCH.java:459)     │
        └──────────────┬───────────────────────┘
                       │
        ┌──────────────┴───────────────┐
        │                              │
   returns TRUE                  returns FALSE
        │                              │
        ▼                              ▼
   PATH A                          PATH B
   InputSelectEntitiesFromList     getGui().chooseSingleEntityForEffect
        │                              │
        │ setSelectables ──┐           │ → GameEntityPicker / SGuiChoose
        │ tempShowZones ───┤           │   (a self-contained dialog —
        │ updateZones      │           │    does NOT use tempShowZones)
        │ hideZones ───────┘           │
        ▼                              ▼
   Renders cards INSIDE the         Renders cards INSIDE the picker
   matching floating zone           dialog. Works regardless of
   on the player's screen.          whether floating zones exist.
   ── REQUIRES tempShowZones
      to be implemented ──

Desktop host → check returns false → PATH A → host serializes tempShowZones over the wire to the mobile client. Mobile's MatchScreen.tempShowZones is an unimplemented stub (// pfps needs to actually do something), so cards never render. The same anti-pattern wrong-routes confirm dialogs, scry reorder, end-step discard, AI-skip reveals, and stack notifications.

The fix

Add boolean isLibgdxPort() to IGuiGame. Local GUIs return their process flavor; RemoteClientGuiGame returns the value the remote sent at lobby handshake (LoginEvent gains a libgdx field, populated by GameClientHandler and stored on RemoteClient at login). Every UI-flavor decision that calls getGui() afterwards is converted to the per-GUI check:

File	Sites	What it gates
PlayerControllerHuman.java	459	tutor / library-search picker (the reported bug)
PlayerControllerHuman.java	977	scry — also drops the !isNetPlay(getGui()) workaround for the same defect class
PlayerControllerHuman.java	814, 1444, 1871	trigger / replacement / payment confirm dialogs
PlayerControllerHuman.java	1564	end-step hand-size discard picker
PlayerControllerHuman.java	1712	spell-played value notification
PlayerControllerHuman.java	2221	AI-skipped-cards reveal
HumanCostDecision.java	1507	cost-action confirm
InputConfirm.java	58, 67, 80	generic confirm dialogs
InputPayMana.java	79	mobile multi-tap behaviour
FControlGameEventHandler.java	263, 285, 391, 449	stack notifications + zone-update gating

Why it's safe

Semantically a no-op except where it fixes a bug:

• Single-player and same-platform netplay: returns the same value as before.
• FControlGameEventHandler is only ever bound to a local GUI (HostedMatch.java:237 routes remotes to GameEventForwarder) — pure consistency cleanup.
• Cross-platform netplay: now routes per the receiving renderer.

LoginEvent gains one boolean field; the existing version-mismatch warning catches any wire-format incompatibility between divergent builds.

Testing

Original bug can be easily recreated on master by casting Solemn Simulacrum as mobile client with desktop host.

Fix verified on NetworkPlay/libgdx-handshake: Solemn Simulacrum tutor (desktop host + mobile client, the original repro) now shows the library cards in the picker instead of an empty selection.

Note: although the bug report describes a recent change in behaviour, none of the relevant code paths (PlayerControllerHuman.java:459, the MatchScreen.tempShowZones stub, useSelectCardsInput, InputSelectEntitiesFromList) appear to have been modified by any recent commit in a way that would introduce this issue. This appears to be a long-standing latent bug.

---

🤖 Generated with Claude Code

[tool4ever]

nice find + breakdown!