Merge upstream/main into pr/overlay-windows (#1027 landed)

GPU handles(): only output-device proofs stay on the CPU; Show All Masks now renders zoom windows (upstream); proof-boundary tolerance in the GPU equivalence test; device-profile draft test.

Co-Authored-By: Claude Sonnet 5.5 <noreply@anthropic.com>

crates/gpu/src/render.rs

@@ -824,7 +824,7 @@ fn linear(cx: &mut Cx<'_>, sampled: &Buf, info: &SourceInfo, plan: &Plan<'_>, ho
 fn denoise(cx: &mut Cx<'_>, img: Buf, plan: &Plan<'_>) -> Buf {
     let (w, h) = (plan.w, plan.h);
     let n = w * h;
-    let (lum, col) = local::nr_params(&plan.settings, plan.src_long, plan.frame.output_long(w, h));
+    let (lum, col) = local::nr_params(&plan.settings, plan.detail_scale);
     let mut img = img;
     if let Some(nr) = lum {
         let l = cx.gpu.buffer(n);
@@ -1015,7 +1015,8 @@ fn masks(cx: &mut Cx<'_>, lin: &Buf, prep: &Prep, plan: &Plan<'_>, host: &mut Ho
                 }
                 MaskShape::Brush { strokes } => {
                     let dabs: Vec<_> = strokes.iter().map(|st| lightcraft_pipeline::masks::brush_dabs(st, frame, w, h)).collect();
-                    if dabs.iter().map(|d| d.dabs.len()).sum::<usize>() > MAX_DABS {
+                    // pen pressure varies radius and flow per dab, which the kernel doesn't take: the CPU rasterizes it
+                    if dabs.iter().map(|d| d.dabs.len()).sum::<usize>() > MAX_DABS || dabs.iter().any(|d| !d.pressure.is_empty()) {
                         None
                     } else if dabs.iter().any(|d| d.auto) {
                         // Auto Mask: stroke by stroke (each one refined before it's combined)