fix: X4500 overlapping registers for FACEGEN + TRUE_PBR

[Copilot]

When FACEGEN and TRUE_PBR are both defined, the HLSL compiler throws error X4500: overlapping register semantics because both feature blocks unconditionally claim registers s4/t4 — FACEGEN for detail sampler/texture, TRUE_PBR for parallax sampler/texture.

Changes

• Register declarations (s4/t4): Guard SampParallaxSampler/TexParallaxSampler with !defined(FACEGEN). Split the TRUE_PBR block so SampRMAOSSampler/TexRMAOSSampler (s5/t5) remain declared for FACEGEN+TRUE_PBR — those registers are free when FACEGEN is active (the ENVMAP block already has the same !defined(FACEGEN) guard).

// Before — both claim s4/t4 when FACEGEN+TRUE_PBR:
#        if defined(TRUE_PBR)
SamplerState SampParallaxSampler : register(s4);
SamplerState SampRMAOSSampler    : register(s5);
#        endif

// After — parallax guarded, RMAOS kept for FACEGEN+TRUE_PBR:
#        if defined(TRUE_PBR) && !defined(FACEGEN)
SamplerState SampParallaxSampler : register(s4);
#        endif
#        if defined(TRUE_PBR)
SamplerState SampRMAOSSampler    : register(s5);
#        endif

• Parallax displacement block (EMAT/TRUE_PBR section): Wrap the HasDisplacement parallax path with #if !defined(FACEGEN) — face geometry doesn't use parallax displacement, and TexParallaxSampler/SampParallaxSampler are undeclared for this combination. PBRParallax stays declared (remains false) so downstream code compiles correctly.

• Shadow multiplier blocks (directional + point light): Add !defined(FACEGEN) to the #elif defined(TRUE_PBR) && !defined(LODLANDSCAPE) branches that pass TexParallaxSampler/SampParallaxSampler to GetParallaxSoftShadowMultiplier — those resources are not declared for FACEGEN+TRUE_PBR.

---

✨ Let Copilot coding agent set things up for you — coding agent works faster and does higher quality work when set up for your repo.

[github-actions]

Actionable Suggestions

• Weather Editor (Dlizzio): Needs version bump to 1-2-0

[github-actions]

✅ A pre-release build is available for this PR:
Download

[alandtse]

@copilot please review comments critically and address or explain why not.

[alandtse]

@copilot shouldn't this be an elseif?

[Copilot]

No — using #elif here would be incorrect. When TRUE_PBR && !FACEGEN, we need both SampParallaxSampler and SampRMAOSSampler declared. If the second block were #elif defined(TRUE_PBR) instead of a fresh #if, SampRMAOSSampler would only be declared in the TRUE_PBR + FACEGEN case, not in the normal TRUE_PBR (non-FACEGEN) case — breaking the RMAOS sample at line 1794 (rawRMAOS = TexRMAOSSampler.SampleBias(SampRMAOSSampler, ...)) which has no FACEGEN guard of its own. The two separate #if blocks are necessary.

[alandtse]

Same question about elseif

[Copilot]

Same reasoning as the sampler block above — TexRMAOSSampler is sampled at line 1794 inside a plain #if defined(TRUE_PBR) block (no FACEGEN exclusion), so it must be declared for all TRUE_PBR variants. An #elif would make TexRMAOSSampler only visible when TRUE_PBR && FACEGEN, breaking the non-FACEGEN path. Two separate #if blocks are required.

[Copilot]

@copilot please review comments critically and address or explain why not.

Both review comments have been addressed. The two separate #if defined(TRUE_PBR) blocks (instead of #elif) are intentional — SampRMAOSSampler/TexRMAOSSampler must be declared for all TRUE_PBR variants because they are sampled at line 1794 with no FACEGEN guard. Using #elif would omit those declarations for the non-FACEGEN path and break RMAOS sampling there.