feat(ww3d2): add IRenderBackend Interface#2613
Conversation
5e014bf to
e648a3e
Compare
|
| Filename | Overview |
|---|---|
| Core/Libraries/Source/WWVegas/WW3D2/dx8wrapper.cpp | Wires Init/Shutdown into the DX8 device lifecycle; shutdown path calls g_renderBackend->Shutdown() without a null guard before the guarded Shutdown_Render_Backend(), creating a crash if reached before Init runs. |
| Core/Libraries/Source/WWVegas/WW3D2/IRenderBackend.h | New abstract interface covering the W3D-facing DX8Wrapper API; TransformKind enum pins values to D3DTS_* constants, leaking a D3D8 ABI detail into the backend-agnostic interface. |
| Core/Libraries/Source/WWVegas/WW3D2/Backend/DX8Backend.h | Concrete adapter header; all overrides declared with virtual but without the override specifier, missing compile-time mismatch detection. |
| Core/Libraries/Source/WWVegas/WW3D2/Backend/DX8Backend.cpp | Pure one-line trampolines to DX8Wrapper static methods; Set_Viewport and Set_Transform include correct struct/enum conversions. No issues found. |
| Core/Libraries/Source/WWVegas/WW3D2/Backend/RenderBackend.cpp | Owns the g_renderBackend global; Init/Shutdown functions have correct null guards internally. |
| Core/Libraries/Source/WWVegas/WW3D2/Backend/RenderBackend.h | Declares the g_renderBackend extern and the two lifecycle functions; clean header with #pragma once. |
| Core/Libraries/Source/WWVegas/WW3D2/CMakeLists.txt | Adds the four new Backend/ sources and IRenderBackend.h to the WW3D2 source list; straightforward and correct. |
Class Diagram
%%{init: {'theme': 'neutral'}}%%
classDiagram
class IRenderBackend {
<<abstract>>
+Initialize(window, width, height)
+Shutdown()
+Is_Device_Lost() bool
+Begin_Scene()
+End_Scene(flip_frame)
+Set_Shader(shader)
+Set_Texture(stage, texture)
+Set_Transform(kind, matrix)
+Draw_Triangles(...)
}
class DX8Backend {
+Is_Device_Lost() bool
+Begin_Scene()
+End_Scene(flip_frame)
+Set_Shader(shader)
+Set_Texture(stage, texture)
+Set_Transform(kind, matrix)
+Draw_Triangles(...)
}
class DX8Wrapper {
<<static facade>>
+Is_Device_Lost()$
+Begin_Scene()$
+Set_Transform(type, matrix)$
+Draw_Triangles(...)$
+Do_Onetime_Device_Dependent_Inits()$
+Do_Onetime_Device_Dependent_Shutdowns()$
}
class RenderBackend {
<<global>>
+g_renderBackend IRenderBackend*
+Init_Render_Backend()
+Shutdown_Render_Backend()
}
IRenderBackend <|-- DX8Backend : implements
DX8Backend --> DX8Wrapper : forwards every call to
DX8Wrapper --> RenderBackend : calls Init/Shutdown
RenderBackend o-- IRenderBackend : owns g_renderBackend
Prompt To Fix All With AI
Fix the following 3 code review issues. Work through them one at a time, proposing concise fixes.
---
### Issue 1 of 3
Core/Libraries/Source/WWVegas/WW3D2/dx8wrapper.cpp:460-461
`g_renderBackend->Shutdown()` is called unconditionally before `Shutdown_Render_Backend()`, which contains the null guard. If `Do_Onetime_Device_Dependent_Shutdowns()` is ever reached without a paired `Do_Onetime_Device_Dependent_Inits()` — for example through an error-recovery path that resets the device partway through initialization — this is a null-pointer dereference crash. The guard should wrap both calls.
```suggestion
if (g_renderBackend != nullptr)
{
g_renderBackend->Shutdown();
}
Shutdown_Render_Backend();
```
### Issue 2 of 3
Core/Libraries/Source/WWVegas/WW3D2/IRenderBackend.h:46-51
`TransformKind` values are pinned to the D3DTS_* integer constants — including the non-sequential `RB_TRANSFORM_WORLD = 256` — so `DX8Backend` can cast them directly. This leaks a D3D8 ABI detail into the abstract interface: any future non-D3D backend will receive magic numbers (2, 3, 256) it must silently understand to mean View/Projection/World. Consider sequential values (0, 1, 2) in the interface and letting `DX8Backend::Set_Transform` map them to the D3D8 constants locally.
### Issue 3 of 3
Core/Libraries/Source/WWVegas/WW3D2/Backend/DX8Backend.h:35-103
All overriding virtual methods use only `virtual` without `override`. The `override` specifier would catch any signature mismatch between `DX8Backend` and `IRenderBackend` at compile time. Prefer `virtual ... override` (or just `override` without `virtual`) for all overriding declarations.
Reviews (4): Last reviewed commit: "feat(ww3d2): add IRenderBackend Initiali..." | Re-trigger Greptile
xezon
left a comment
There was a problem hiding this comment.
Logical approach in an effort to start introducing new render backends. Some open questions.
| // Create the render backend. Called by DX8Wrapper::Do_Onetime_Device_Dependent_Inits | ||
| // after the D3D device has been successfully created. | ||
| // | ||
| // Phase 1 always creates a DX8Backend. Phase 2 will add a compile-time |
There was a problem hiding this comment.
I suggest only describe the status quo, not Phase 2 futures.
| // during device teardown. Never null between those two calls. | ||
| extern IRenderBackend * g_renderBackend; | ||
|
|
||
| // Create the render backend. Called by DX8Wrapper::Do_Onetime_Device_Dependent_Inits |
There was a problem hiding this comment.
I am not fond of the function declarations explaining which third parties calls it, because the callers can change and then these comments are outdated. Several times.
| // that migrating callers is a mechanical `DX8Wrapper::X(...)` → | ||
| // `g_renderBackend->X(...)` rewrite with minimal diff noise. | ||
| // | ||
| // **This header is included from VC6-compiled translation units** (the |
There was a problem hiding this comment.
This section goes into way too many details. This applies to the whole code base.
| return; | ||
| } | ||
|
|
||
| // Phase 1: the DX8 backend is the only option. Phase 2 will introduce |
There was a problem hiding this comment.
Please make a pass on comments and clean them up.
| // TheSuperHackers @refactor bobtista 10/04/2026 Construct the global | ||
| // IRenderBackend instance now that the D3D device is ready. See | ||
| // Core/Libraries/Source/WWVegas/WW3D2/RENDER_BACKEND.md. | ||
| Init_Render_Backend(); |
There was a problem hiding this comment.
This looks suspicious. Shouldn't Init_Render_Backend call DX8Wrapper::Do_Onetime_Device_Dependent_Inits instead?
There was a problem hiding this comment.
I don't think so. The contract is: DX8Wrapper owns when the renderer comes up (it owns the D3D device), and the backend owns how it brings up its own device. To make that explicit rather than implied, I'll add a lifecycle pair to the interface eg Initialize(hwnd, w, h) / Shutdown() with empty default bodies, called after Init_Render_Backend() and before Shutdown_Render_Backend(). DX8Backend treats them as no-ops since DX8Wrapper still owns the real device, and non-DX8 backends use Initialize() to create their own device/swapchain against the game window. Device-lost/reset stays DX8-internal, it's a D3D8 artifact, and modern backends recover via swapchain reset without surfacing it.
|
|
||
| // TheSuperHackers @refactor bobtista 10/04/2026 Construct the global | ||
| // IRenderBackend instance now that the D3D device is ready. See | ||
| // Core/Libraries/Source/WWVegas/WW3D2/RENDER_BACKEND.md. |
There was a problem hiding this comment.
This comment refers to a file that is not part of the change.
| // after the D3D device has been successfully created. | ||
| // | ||
| // Phase 1 always creates a DX8Backend. Phase 2 will add a compile-time | ||
| // option to select between DX8Backend, BgfxBackend, and DiligentBackend. |
There was a problem hiding this comment.
It's another backend like bgfx, I ended up going with bgfx, but if anyone else wants to run with Diligent it would work the same. Cleaned the commit anyway
| // Implementations (DX8Backend.cpp, future BgfxBackend.cpp, etc.) can use | ||
| // whatever C++ features the project's main build allows. | ||
|
|
||
| class IRenderBackend |
There was a problem hiding this comment.
How will we cover DX8Caps related stuff?
There was a problem hiding this comment.
Capability queries become a set of narrow virtuals with safe defaults eg Supports_Texture_Format(WW3DFormat), Supports_Compressed_Textures(), Get_Texture_Limits(), Supports_Texture_Op(...), etc.
DX8Backend forwards each to DX8Wrapper::Get_Current_Caps(), so DX8Caps stays the reference implementation and is never exposed directly. The few sites that read raw D3DCAPS8 bitfields today (COLORWRITEENABLE in W3DScene/W3DVolumetricShadow, TextureOpCaps in shader.cpp) and the Voodoo3 vendor check get promoted to neutral predicates (Supports_Color_Write_Mask(), Supports_Texture_Op(), Is_Legacy_Voodoo3()) so shared engine code carries no D3D types or #ifdefs. I've kept these out of this scaffolding PR deliberately, each lands with the call-site it unblocks, but the shape is good.
| // Implementations (DX8Backend.cpp, future BgfxBackend.cpp, etc.) can use | ||
| // whatever C++ features the project's main build allows. | ||
|
|
||
| class IRenderBackend |
There was a problem hiding this comment.
What will we do with direct calls to DX8, for example _Get_D3D_Device8 or Get_DX8_Texture_Stage_State_Value_Name ?
There was a problem hiding this comment.
This took a lot of doing, but has been worth it. In my run ahead branches for bgfx (still local), I have it so there are two classes, handled differently. The high-frequency render/texture-stage-state calls aren't re-exposed as raw D3DRS_/D3DTSS_; they route through a backend-neutral fixed-function state cache (keyed by the same ordinals) plus typed semantic setters, so a non-DX8 backend reads intent rather than D3D enums. The genuinely DX8-only entry points (_Get_D3D_Device8, Create_DX8*, raw SetRenderTarget) migrate to named high-level methods (Set_Render_Target_With_Z, a view-capture primitive, …); the irreducible cases e.g. hand-written water pixel-shader bytecode go behind a named-enum hatch (Create_Legacy_Pixel_Shader(kind)) rather than a raw device pointer. The end state has no _Get_D3D_Device8 left in the engine subsystems; the raw device stays inside DX8Backend. Pure DX8 diagnostics eg Get_DX8_Texture_Stage_State_Value_Name simply stay on DX8Wrapper, as they're debug-only and not part of the abstraction.
| proxy.h | ||
| rddesc.h | ||
| RenderBackend.cpp | ||
| RenderBackend.h |
There was a problem hiding this comment.
Can put the backends except IRenderBackend.h into a subfolder so they are tucked away a bit.
|
I've already got the bgfx backend working on my Mac and Windows machines, a lot was figured along the way, and some stuff like the lifecycle I'm happy to add here and inherit from. I'm still ironing out some quirks, will share more when things feel polished enough. |
| g_renderBackend->Shutdown(); | ||
| Shutdown_Render_Backend(); |
There was a problem hiding this comment.
g_renderBackend->Shutdown() is called unconditionally before Shutdown_Render_Backend(), which contains the null guard. If Do_Onetime_Device_Dependent_Shutdowns() is ever reached without a paired Do_Onetime_Device_Dependent_Inits() — for example through an error-recovery path that resets the device partway through initialization — this is a null-pointer dereference crash. The guard should wrap both calls.
| g_renderBackend->Shutdown(); | |
| Shutdown_Render_Backend(); | |
| if (g_renderBackend != nullptr) | |
| { | |
| g_renderBackend->Shutdown(); | |
| } | |
| Shutdown_Render_Backend(); |
Prompt To Fix With AI
This is a comment left during a code review.
Path: Core/Libraries/Source/WWVegas/WW3D2/dx8wrapper.cpp
Line: 460-461
Comment:
`g_renderBackend->Shutdown()` is called unconditionally before `Shutdown_Render_Backend()`, which contains the null guard. If `Do_Onetime_Device_Dependent_Shutdowns()` is ever reached without a paired `Do_Onetime_Device_Dependent_Inits()` — for example through an error-recovery path that resets the device partway through initialization — this is a null-pointer dereference crash. The guard should wrap both calls.
```suggestion
if (g_renderBackend != nullptr)
{
g_renderBackend->Shutdown();
}
Shutdown_Render_Backend();
```
How can I resolve this? If you propose a fix, please make it concise.
Summary
First PR in a planned multi-step refactor introducing an
IRenderBackendinterface in WW3D2. This PR adds only the scaffolding — interface, adapter, lifecycle wiring, and one isolated proof-of-concept call site. zero behavior change. The DX8 path remains the only renderer;g_renderBackendis constructed as aDX8Backendthat forwards every method to the existingDX8Wrapperstatic facade.What this PR adds
IRenderBackend.h— abstract interface covering the W3D-facing subset ofDX8Wrapper's public API. Only high-level methods (those takingShaderClass/TextureBaseClass/Matrix4x4/etc.) are virtualized. D3D8-typed low-level methods stay onDX8WrapperasDX8Backend-specific escape hatches — see the doc for the rationale.DX8Backend.{h,cpp}— concrete adapter that forwards every virtual method to the existingDX8Wrapper::static functions. Pure forwarding, no new rendering logic.RenderBackend.{h,cpp}— exposes the globalIRenderBackend* g_renderBackend.Init_Render_Backend()constructsnew DX8Backend()fromDX8Wrapper::Do_Onetime_Device_Dependent_Inits();Shutdown_Render_Backend()tears it down fromDX8Wrapper::Do_Onetime_Device_Dependent_Shutdowns().Test plan