fix: Emit empty fragment targets instead of [null]

[vende11s]

• Fragment shaders that return only a builtin e.g (out: d.builtin.fragDepth) now create pipelines with fragment.targets: [] instead of [null].
  Both ways are spec-legal, but Firefox rejects [null] with Missing required 'format' member of GPUColorTargetState, so the pipeline never ran.

• Due to this bug Point Light Shadow example now clears the color attachment to [0, 0, 0, 1].

[github-actions]

pkg.pr.new

packages
Ready to be installed by your favorite package manager ⬇️

https://pkg.pr.new/software-mansion/TypeGPU/typegpu@9f0882b43acb55affaa848d3705237ff8c8d9882

benchmark
view benchmark

commit
view commit

[github-actions]

Resolution Time Benchmark

---
config:
  themeVariables:
    xyChart:
      plotColorPalette: "#E63946, #3B82F6, #059669"
---
xychart
  title "Random Branching (🔴 PR | 🔵 main | 🟢 release)"
  x-axis "max depth" [1, 2, 3, 4, 5, 6, 7, 8]
  y-axis "time (ms)"
  line [0.95, 1.84, 4.37, 5.92, 7.00, 13.03, 24.67, 26.04]
  line [0.98, 1.89, 4.47, 6.26, 6.96, 11.42, 23.35, 26.11]
  line [0.93, 1.87, 3.98, 6.54, 7.00, 10.65, 22.93, 24.72]

Loading

---
config:
  themeVariables:
    xyChart:
      plotColorPalette: "#E63946, #3B82F6, #059669"
---
xychart
  title "Linear Recursion (🔴 PR | 🔵 main | 🟢 release)"
  x-axis "max depth" [1, 2, 3, 4, 5, 6, 7, 8]
  y-axis "time (ms)"
  line [0.26, 0.53, 0.71, 0.86, 1.13, 1.19, 1.47, 1.53]
  line [0.31, 0.52, 0.69, 0.79, 1.06, 1.17, 1.42, 1.55]
  line [0.27, 0.47, 0.67, 0.82, 1.14, 1.17, 1.37, 1.55]

Loading

---
config:
  themeVariables:
    xyChart:
      plotColorPalette: "#E63946, #3B82F6, #059669"
---
xychart
  title "Full Tree (🔴 PR | 🔵 main | 🟢 release)"
  x-axis "max depth" [1, 2, 3, 4, 5, 6, 7, 8]
  y-axis "time (ms)"
  line [0.79, 2.10, 3.27, 6.55, 12.18, 25.45, 54.53, 111.06]
  line [0.78, 2.23, 3.26, 6.76, 11.92, 24.85, 55.30, 110.50]
  line [0.87, 1.98, 3.27, 6.55, 12.67, 25.58, 54.27, 110.52]

Loading

[github-actions]

Bundle size comparison (import * as ... in PR vs import * as ... in target):

🟢 Decreased (max -0.00%)	➖ Unchanged	🔴 Increased	❔ Unknown
21	304	0	0

import { ... } in PR vs import * as ... in PR (is the library tree-Shakeable?):

Test	tsdown
tgpu_init.ts	263.09 kB ($${\color{green}-3.5\%}$$)
tgpu_initFromDevice.ts	262.55 kB ($${\color{green}-3.7\%}$$)
tgpu_resolve.ts	161.75 kB ($${\color{green}-40.7\%}$$)
tgpu_resolveWithContext.ts	161.68 kB ($${\color{green}-40.7\%}$$)
tgpu_bindGroupLayout.ts	62.32 kB ($${\color{green}-77.1\%}$$)
tgpu_mutableAccessor.ts	57.04 kB ($${\color{green}-79.1\%}$$)
tgpu_accessor.ts	57.04 kB ($${\color{green}-79.1\%}$$)
tgpu_privateVar.ts	55.73 kB ($${\color{green}-79.6\%}$$)
tgpu_workgroupVar.ts	55.73 kB ($${\color{green}-79.6\%}$$)
tgpu_const.ts	55.15 kB ($${\color{green}-79.8\%}$$)
tgpu_lazy.ts	54.95 kB ($${\color{green}-79.8\%}$$)
tgpu_fragmentFn.ts	39.68 kB ($${\color{green}-85.4\%}$$)
tgpu_fn.ts	39.62 kB ($${\color{green}-85.5\%}$$)
tgpu_vertexFn.ts	39.50 kB ($${\color{green}-85.5\%}$$)
tgpu_computeFn.ts	39.20 kB ($${\color{green}-85.6\%}$$)
tgpu_vertexLayout.ts	28.33 kB ($${\color{green}-89.6\%}$$)
tgpu_comptime.ts	15.93 kB ($${\color{green}-94.2\%}$$)
tgpu_unroll.ts	1.75 kB ($${\color{green}-99.4\%}$$)
tgpu_slot.ts	1.70 kB ($${\color{green}-99.4\%}$$)

If you wish to run a comparison for other, slower bundlers, run the 'Tree-shake test' from the GitHub Actions menu.

[pullfrog]

ℹ️ The fix is correct and spec-legal — one suggestion inline.

Reviewed changes — Single commit (9f0882b) turning [null] fragment target emission into [] for shaders that write only a builtin or nothing, which Firefox rejects with Missing required 'format' member of GPUColorTargetState.

• connectTargetsToShader.ts — void/builtin fragment outputs now produce targets: [] instead of [null].
• renderPipeline.ts — connectedTargets falls back to [] (was [null]) when the fragment output isn't resolvable.
• point-light-shadow example — clear-color used to make the example work correctly.

The change is narrowly scoped and matches the underlying model: a void or builtin-only fragment (e.g. @builtin(frag_depth)) declares no color outputs, so [] is not merely spec-legal but the semantically accurate descriptor. The struct branch that mixes builtins with color outputs is untouched, so pipelines writing both depth and color keep their real targets.

Technical details

# Suggestion: add a regression test

The two changed pipelines reach the `typeof navigator === 'undefined'`-safe
branches before any `navigator.gpu` access, so `connectTargetsToShader(void|builtin)`
can be unit-tested without a device — it returns before the `getPreferredCanvasFormat`
call. Consider pinning the actual regression: a builtin-only fragment (or void) should
yield `fragment.targets: []`, and a struct mixing color + `@builtin(frag_depth)` should
still yield the color targets. Without a test, this subtle spec-compat behavior can
silently regress back to `[null]`.

  ^{｜ Fix all ➔ ｜ Fix 👍s ➔ ｜ View workflow run ｜ Using DeepSeek Flash (free via Pullfrog for OSS) ｜ 𝕏}

[pullfrog]

Minor: worth a regression test pinning targets: [] here? This branch returns before any navigator.gpu access, so connectTargetsToShader(voidOrBuiltin, ...) is unit-testable without a device. The exactly regression — a builtin-only fragment (e.g. @builtin(frag_depth)) yielding targets: [] instead of [null] — would otherwise be easy to silently revert, since the unmanifested difference only shows up on Firefox (null rejects format).