capability-propagation

name	capability-propagation
description	When threading adapter capability flags from device creation through init-time constructors across crate boundaries, add a bool parameter to the constructor chain rather than trying to route through per-frame PassCtx — init-time allocations happen before per-frame dispatch and need the flag at construction time
source	auto-skill
extracted_at	2026-06-15T03:49:26.570Z

Capability flag propagation through init chain

When an adapter capability probed at device creation (e.g. compact format support, f16 storage, ray queries) is needed during init-time resource allocation (e.g. buffer stride selection), the flag must be threaded through the constructor chain, not the per-frame PassCtx.

Why: Init-time constructors (StratumDispatch::new, CellCandidateBuffer::new) run before any frame dispatch. The PassCtx is per-frame and populated with cached capability flags from Renderer3d, but init code has no PassCtx.

How to thread:

1. Store the flag on Renderer3d (probed at device creation)
2. Add a public accessor: pub fn compact_storage_supported(&self) -> bool
3. Pass it through the constructor: StratumDispatch::new(..., compact_storage)
4. Continue threading to leaf allocators: CellCandidateBuffer::new(..., compact_storage)

Call site pattern (editor/lib.rs):

let dispatch = StratumDispatch::new(
    renderer.device(),
    renderer.queue(),
    resolution,
    provider,
    stratum_quality,
    renderer.compact_storage_supported(),
);

Constructor pattern:

pub fn new(
    device: &wgpu::Device,
    // ... existing params ...
    compact_storage: bool,
) -> Self { ... }

When the capability is always true on desktop but needs probing on mobile: Use a runtime probe, not #[cfg] — Titan ships one binary across platforms and the probe is cheap (one get_texture_format_features call at boot). On desktop the probe always passes; on mobile it correctly detects older Mali/ Adreno GPUs that lack STORAGE_BINDING for compact formats.