mirui 0.22.1

A lightweight, no_std ECS-driven UI framework for embedded, desktop, and WebAssembly
Documentation
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386
387
388
389
390
391
392
393
394
395
396
397
398
399
400
401
402
403
404
405
406
407
408
409
410
411
412
413
414
415
416
417
418
419
420
421
422
423
424
425
426
427
428
429
430
431
432
433
434
435
436
437
438
439
440
441
442
443
444
445
446
447
448
449
450
451
452
453
454
455
456
457
458
459
460
461
462
463
464
465
466
467
468
469
470
471
472
473
474
475
476
477
478
479
480
481
482
483
484
485
486
487
488
489
490
491
492
493
494
495
496
497
498
499
500
501
502
503
504
505
506
507
508
509
510
511
512
513
514
515
516
517
518
519
520
521
522
523
524
525
526
527
528
529
530
531
532
533
534
535
536
537
538
539
540
541
542
543
544
545
546
547
548
549
550
551
552
553
554
555
556
557
558
559
560
561
562
563
564
565
566
567
568
569
570
571
572
573
574
575
576
577
578
579
580
581
582
583
584
585
586
587
588
589
590
591
592
593
594
595
596
597
598
599
600
601
602
603
604
605
606
607
608
609
610
611
612
613
614
615
616
617
618
619
620
621
622
623
624
625
626
627
628
629
630
631
632
633
634
635
636
637
638
639
640
641
642
643
644
645
646
647
648
649
650
651
652
653
654
655
656
657
658
659
660
661
662
663
664
665
666
667
668
669
670
671
672
673
674
675
676
677
678
679
680
681
682
683
684
685
686
687
688
689
690
691
692
693
694
695
696
697
698
699
700
701
702
703
704
705
706
707
708
709
710
711
712
713
714
715
716
717
718
719
720
721
722
723
724
725
726
727
728
729
730
731
732
733
734
735
736
737
738
739
740
741
742
743
744
745
746
747
748
749
750
751
752
753
754
755
756
757
758
759
760
761
762
763
764
765
766
767
768
769
770
771
772
773
774
775
776
777
778
779
780
781
782
783
784
785
786
787
788
789
790
791
792
793
794
795
796
797
798
799
800
801
802
803
804
805
806
807
808
809
810
811
812
813
814
815
816
817
818
819
820
821
822
823
824
825
826
827
828
829
830
831
832
833
834
835
836
837
838
839
840
841
842
843
844
845
846
847
848
849
850
851
852
853
854
855
856
857
858
859
860
861
862
863
864
865
866
867
868
869
870
871
872
873
874
875
876
877
878
879
880
881
882
883
884
885
886
887
888
889
890
891
892
893
894
895
896
897
898
899
900
901
902
903
904
905
906
907
908
909
910
911
912
913
914
use super::SwRenderer;
use crate::draw::texture::{ColorFormat, Texture};
use crate::types::{Color, Fixed, Rect};

impl SwRenderer<'_> {
    pub(super) fn fill_rect_inner(
        &mut self,
        area: &Rect,
        clip: &Rect,
        color: &Color,
        radius: Fixed,
        opa: u8,
    ) {
        let phys_area = self.viewport.rect_to_physical(*area);
        let phys_clip = self.viewport.rect_to_physical(*clip);
        let radius = radius * self.viewport.scale();
        let area = &phys_area;
        let clip = &phys_clip;

        let screen = Rect::new(0, 0, self.target.width, self.target.height);
        let Some(draw_area) = area.intersect(clip) else {
            return;
        };
        let Some(draw_area) = draw_area.intersect(&screen) else {
            return;
        };

        let r = radius.min(area.w / 2).min(area.h / 2);
        let (px_x0, px_y0, px_x1, px_y1) = draw_area.pixel_bounds();
        // Fold color.a into opa so every downstream blend treats the
        // colour as opaque RGB. Mirrors what blend_pixel does internally
        // before delegating to blend_pixel_int. The fold happens before
        // coverage instead of after, which can shift a half-transparent
        // fully-covered pixel by ±1 alpha vs the pre-eb98ecc path; the
        // difference is below visual threshold and inside the existing
        // ±2 symmetry tolerance.
        let effective_opa = ((color.a as u16 * opa as u16) / 255) as u8;
        let opa_norm =
            Fixed::from_int(effective_opa as i32).map_range((0, 255), (Fixed::ZERO, Fixed::ONE));

        if area.is_aligned() && r == Fixed::ZERO {
            crate::trace_span!("sw.fill_aligned");
            fill_axis_aligned(
                &mut self.target,
                px_x0,
                px_y0,
                px_x1,
                px_y1,
                color,
                effective_opa,
            );
            return;
        }

        // Axis-aligned + rounded: solid fill the inner cross, only run
        // SDF coverage on the four r×r corner bboxes. r is ceil'd so
        // the SDF region fully covers the curved boundary.
        if area.is_aligned() && r > Fixed::ZERO {
            let r_px = r.ceil().to_int();
            let area_x0 = area.x.to_int();
            let area_y0 = area.y.to_int();
            let area_x1 = (area.x + area.w).to_int();
            let area_y1 = (area.y + area.h).to_int();
            // Inner-cross requires the area to be wider/taller than 2r;
            // otherwise the four corner bboxes overlap and there's no
            // straight section to short-circuit.
            if area_x1 - area_x0 > 2 * r_px && area_y1 - area_y0 > 2 * r_px {
                crate::trace_span!("sw.fill_aligned_rounded");
                let inner_x0 = area_x0 + r_px;
                let inner_x1 = area_x1 - r_px;
                let inner_y0 = area_y0 + r_px;
                let inner_y1 = area_y1 - r_px;
                fill_axis_aligned(
                    &mut self.target,
                    inner_x0.max(px_x0),
                    area_y0.max(px_y0),
                    inner_x1.min(px_x1),
                    inner_y0.min(px_y1),
                    color,
                    effective_opa,
                );
                fill_axis_aligned(
                    &mut self.target,
                    inner_x0.max(px_x0),
                    inner_y1.max(px_y0),
                    inner_x1.min(px_x1),
                    area_y1.min(px_y1),
                    color,
                    effective_opa,
                );
                fill_axis_aligned(
                    &mut self.target,
                    area_x0.max(px_x0),
                    inner_y0.max(px_y0),
                    area_x1.min(px_x1),
                    inner_y1.min(px_y1),
                    color,
                    effective_opa,
                );
                let corners = [
                    (area_x0, area_y0, inner_x0, inner_y0), // TL
                    (inner_x1, area_y0, area_x1, inner_y0), // TR
                    (area_x0, inner_y1, inner_x0, area_y1), // BL
                    (inner_x1, inner_y1, area_x1, area_y1), // BR
                ];
                for (cx0, cy0, cx1, cy1) in corners {
                    let bx0 = cx0.max(px_x0);
                    let by0 = cy0.max(px_y0);
                    let bx1 = cx1.min(px_x1);
                    let by1 = cy1.min(px_y1);
                    for py in by0..by1 {
                        for px in bx0..bx1 {
                            let cov = rounded_rect_coverage(
                                Fixed::from_int(px) - area.x,
                                Fixed::from_int(py) - area.y,
                                area.w,
                                area.h,
                                r,
                            );
                            let final_opa = (cov * opa_norm).map01(255).to_int() as u8;
                            if final_opa > 0 {
                                self.target.blend_pixel_int(px, py, color, final_opa);
                            }
                        }
                    }
                }
                return;
            }
        }

        // Y-fractional fast path: a row scrolled to a sub-pixel Y
        // offset is X-axis-aligned with a real integer-aligned middle
        // band, so split the fill into top AA strip + memcpy mid +
        // bottom AA strip and skip per-pixel coverage on the bulk.
        let x_is_aligned = area.x.fract() == Fixed::ZERO && area.w.fract() == Fixed::ZERO;
        if x_is_aligned && r == Fixed::ZERO && area.h >= Fixed::from_int(3) {
            let area_x_end_int = (area.x + area.w).to_int();
            let area_x_int = area.x.to_int();
            let mid_y0_aligned = area.y.ceil().to_int();
            let mid_y1_aligned = (area.y + area.h).floor().to_int();
            if mid_y1_aligned > mid_y0_aligned {
                let clip_x0 = area_x_int.max(px_x0);
                let clip_x1 = area_x_end_int.min(px_x1);
                let clip_mid_y0 = mid_y0_aligned.max(px_y0);
                let clip_mid_y1 = mid_y1_aligned.min(px_y1);
                if clip_x1 > clip_x0 && clip_mid_y1 > clip_mid_y0 {
                    crate::trace_span!("sw.fill_aa_split_mid");
                    fill_axis_aligned(
                        &mut self.target,
                        clip_x0,
                        clip_mid_y0,
                        clip_x1,
                        clip_mid_y1,
                        color,
                        effective_opa,
                    );
                    let aa_top_y0 = px_y0;
                    let aa_top_y1 = mid_y0_aligned.min(px_y1);
                    let aa_bot_y0 = mid_y1_aligned.max(px_y0);
                    let aa_bot_y1 = px_y1;
                    if aa_top_y1 > aa_top_y0 {
                        crate::trace_span!("sw.fill_aa_split_top");
                        fill_aa_band(
                            &mut self.target,
                            *area,
                            aa_top_y0,
                            aa_top_y1,
                            px_x0,
                            px_x1,
                            color,
                            effective_opa,
                            opa_norm,
                        );
                    }
                    if aa_bot_y1 > aa_bot_y0 {
                        crate::trace_span!("sw.fill_aa_split_bot");
                        fill_aa_band(
                            &mut self.target,
                            *area,
                            aa_bot_y0,
                            aa_bot_y1,
                            px_x0,
                            px_x1,
                            color,
                            effective_opa,
                            opa_norm,
                        );
                    }
                    return;
                }
            }
        }

        crate::trace_span!("sw.fill_aa_loop");
        let area_x_end = area.x + area.w;
        let area_y_end = area.y + area.h;
        // Pixels strictly inside the corner-free zone don't need an SDF
        // call; ceil/floor on the Fixed side gives the integer column /
        // row range. When r == 0 the whole interior is corner-free.
        let r_int = if r > Fixed::ZERO {
            r.ceil()
        } else {
            Fixed::ZERO
        };
        let mid_x0 = (area.x + r_int).ceil().to_int();
        let mid_x1 = (area_x_end - r_int).to_int();
        let mid_y0 = (area.y + r_int).ceil().to_int();
        let mid_y1 = (area_y_end - r_int).to_int();
        let has_corners = r > Fixed::ZERO;
        let opa_full = effective_opa == 255;

        for py in px_y0..px_y1 {
            let pixel_top = Fixed::from_int(py);
            let pixel_bot = Fixed::from_int(py + 1);
            let cov_y = if pixel_top >= area.y && pixel_bot <= area_y_end {
                Fixed::ONE
            } else {
                (pixel_bot.min(area_y_end) - pixel_top.max(area.y))
                    .max(Fixed::ZERO)
                    .min(Fixed::ONE)
            };
            let row_full_y = cov_y == Fixed::ONE;
            let in_mid_y = py >= mid_y0 && py < mid_y1;

            for px in px_x0..px_x1 {
                let pixel_left = Fixed::from_int(px);
                let pixel_right = Fixed::from_int(px + 1);
                let col_full_x = pixel_left >= area.x && pixel_right <= area_x_end;
                let in_mid_x = px >= mid_x0 && px < mid_x1;

                // Hot path: straight-zone fully-opaque pixel — skip
                // every Fixed multiply and the SDF call.
                if row_full_y && col_full_x && in_mid_y && in_mid_x && opa_full {
                    self.target.set_pixel(px, py, color);
                    continue;
                }

                let cov_x = if col_full_x {
                    Fixed::ONE
                } else {
                    (pixel_right.min(area_x_end) - pixel_left.max(area.x))
                        .max(Fixed::ZERO)
                        .min(Fixed::ONE)
                };
                let corner_cov = if has_corners && !(in_mid_y && in_mid_x) {
                    rounded_rect_coverage(
                        Fixed::from_int(px) - area.x,
                        Fixed::from_int(py) - area.y,
                        area.w,
                        area.h,
                        r,
                    )
                } else {
                    Fixed::ONE
                };

                let final_opa = (cov_x * cov_y * corner_cov * opa_norm).map01(255).to_int() as u8;
                if final_opa > 0 {
                    // px/py are integers — skip blend_pixel's is_integer dispatch.
                    self.target.blend_pixel_int(px, py, color, final_opa);
                }
            }
        }
    }
}

/// AA loop over `[y0, y1)`, used by the X-axis-aligned split. Caller
/// has verified there are no rounded corners; X coverage is implicitly
/// 1.0 so only Y coverage and `effective_opa` contribute.
#[allow(clippy::too_many_arguments)]
fn fill_aa_band(
    target: &mut Texture,
    area: Rect,
    y0: i32,
    y1: i32,
    x0: i32,
    x1: i32,
    color: &Color,
    effective_opa: u8,
    opa_norm: Fixed,
) {
    let area_y_end = area.y + area.h;
    for py in y0..y1 {
        let pixel_top = Fixed::from_int(py);
        let pixel_bot = Fixed::from_int(py + 1);
        let cov_y = if pixel_top >= area.y && pixel_bot <= area_y_end {
            Fixed::ONE
        } else {
            (pixel_bot.min(area_y_end) - pixel_top.max(area.y))
                .max(Fixed::ZERO)
                .min(Fixed::ONE)
        };
        if cov_y <= Fixed::ZERO {
            continue;
        }
        let final_opa = if cov_y == Fixed::ONE && effective_opa == 255 {
            255
        } else {
            (cov_y * opa_norm).map01(255).to_int() as u8
        };
        if final_opa == 0 {
            continue;
        }
        for px in x0..x1 {
            target.blend_pixel_int(px, py, color, final_opa);
        }
    }
}

fn fill_axis_aligned(
    target: &mut Texture,
    px_x0: i32,
    px_y0: i32,
    px_x1: i32,
    px_y1: i32,
    color: &Color,
    opa: u8,
) {
    // Caller may pass an empty sub-rect when clip only overlaps one of
    // the rounded fast path's bands; the negative width otherwise
    // underflows `(px_x1 - px_x0) as usize` and panics on slice access.
    if px_x1 <= px_x0 || px_y1 <= px_y0 {
        return;
    }
    if opa == 255 {
        let bpp = target.format.bytes_per_pixel();
        let stride = target.stride;
        let row_px = (px_x1 - px_x0) as usize;
        let row_bytes = row_px * bpp;
        let buf = target.buf.as_mut_slice();
        match target.format {
            ColorFormat::RGBA8888 => {
                fill_first_row_then_replicate::<4>(
                    buf,
                    stride,
                    px_x0,
                    px_y0,
                    px_y1,
                    row_bytes,
                    [color.r, color.g, color.b, color.a],
                );
            }
            ColorFormat::RGB565 | ColorFormat::RGB565Swapped => {
                let px16 = ((color.r as u16 >> 3) << 11)
                    | ((color.g as u16 >> 2) << 5)
                    | (color.b as u16 >> 3);
                let pixel = if target.format == ColorFormat::RGB565Swapped {
                    [(px16 >> 8) as u8, px16 as u8]
                } else {
                    [px16 as u8, (px16 >> 8) as u8]
                };
                fill_first_row_then_replicate::<2>(
                    buf, stride, px_x0, px_y0, px_y1, row_bytes, pixel,
                );
            }
            _ => {
                for py in px_y0..px_y1 {
                    for px in px_x0..px_x1 {
                        target.set_pixel(px, py, color);
                    }
                }
            }
        }
    } else {
        for py in px_y0..px_y1 {
            for px in px_x0..px_x1 {
                target.blend_pixel_int(px, py, color, opa);
            }
        }
    }
}

/// Fast-path opaque solid fill: write one scanline, then memcpy
/// it into each subsequent row. Avoids both per-pixel
/// `copy_from_slice` bounds-check overhead and any heap alloc for a
/// repeating-pattern buffer.
#[inline]
fn fill_first_row_then_replicate<const BPP: usize>(
    buf: &mut [u8],
    stride: usize,
    px_x0: i32,
    px_y0: i32,
    px_y1: i32,
    row_bytes: usize,
    pixel: [u8; BPP],
) {
    if px_y0 >= px_y1 {
        return;
    }
    let first_start = px_y0 as usize * stride + px_x0 as usize * BPP;
    let first_row = &mut buf[first_start..first_start + row_bytes];
    for chunk in first_row.chunks_exact_mut(BPP) {
        chunk.copy_from_slice(&pixel);
    }
    for py in (px_y0 + 1)..px_y1 {
        let dst_start = py as usize * stride + px_x0 as usize * BPP;
        let (lo, hi) = buf.split_at_mut(dst_start);
        let src = &lo[first_start..first_start + row_bytes];
        hi[..row_bytes].copy_from_slice(src);
    }
}

pub(super) fn rounded_rect_coverage(px: Fixed, py: Fixed, w: Fixed, h: Fixed, r: Fixed) -> Fixed {
    if r == Fixed::ZERO {
        return Fixed::ONE;
    }

    let (cx, cy) = if px < r && py < r {
        (r, r)
    } else if px >= w - r && py < r {
        (w - r, r)
    } else if px < r && py >= h - r {
        (r, h - r)
    } else if px >= w - r && py >= h - r {
        (w - r, h - r)
    } else {
        return Fixed::ONE;
    };

    // 4×4 supersample on the AA boundary; single-sample looks flat-topped
    // because all the curvature collapses into one pixel row. Inside r-1
    // and outside r+1 short-circuit so only the ~2-px ring pays the 16
    // samples.
    let dx_pc = px - cx + Fixed::ONE / 2;
    let dy_pc = py - cy + Fixed::ONE / 2;
    let dist_sq = dx_pc * dx_pc + dy_pc * dy_pc;
    let r_sq = r * r;
    let r_inner = r - Fixed::ONE;
    if r_inner > Fixed::ZERO {
        let r_inner_sq = r_inner * r_inner;
        if dist_sq <= r_inner_sq {
            return Fixed::ONE;
        }
    }
    let r_outer = r + Fixed::ONE;
    let r_outer_sq = r_outer * r_outer;
    if dist_sq >= r_outer_sq {
        return Fixed::ZERO;
    }
    let mut hits: i32 = 0;
    let step = Fixed::ONE / 4;
    let half_step = step / 2;
    let base_x = px - cx + half_step;
    let base_y = py - cy + half_step;
    for sy in 0..4 {
        let dy = base_y + step * Fixed::from_int(sy);
        for sx in 0..4 {
            let dx = base_x + step * Fixed::from_int(sx);
            if dx * dx + dy * dy <= r_sq {
                hits += 1;
            }
        }
    }
    Fixed::from_int(hits) / 16
}

#[cfg(all(test, feature = "std"))]
mod corner_check {
    extern crate std;
    use super::*;
    use crate::draw::canvas::Canvas;
    use std::string::String;
    use std::vec::Vec;

    fn render_circle(w: i32, h: i32, r: i32) -> Vec<Vec<u8>> {
        let mut buf = std::vec![0u8; (w as usize) * (h as usize) * 4];
        let tex = Texture::new(&mut buf, w as u16, h as u16, ColorFormat::RGBA8888);
        let mut backend = SwRenderer::new(tex);
        let rect = Rect::new(0, 0, w, h);
        let clip = Rect::new(0, 0, w, h);
        backend.fill_rect(
            &rect,
            &clip,
            &Color::rgb(255, 255, 255),
            Fixed::from_int(r),
            255,
        );
        let mut out = std::vec![std::vec![0u8; w as usize]; h as usize];
        for py in 0..h {
            for px in 0..w {
                out[py as usize][px as usize] = backend.target.get_pixel(px, py).r;
            }
        }
        out
    }

    fn ascii(grid: &[Vec<u8>]) -> String {
        let mut s = String::from("\n");
        for row in grid {
            for &a in row {
                s.push_str(if a > 200 {
                    "##"
                } else if a > 100 {
                    ".."
                } else if a > 0 {
                    "::"
                } else {
                    "  "
                });
            }
            s.push('\n');
        }
        s
    }

    #[test]
    fn dump_32x32_r16() {
        let g = render_circle(32, 32, 16);
        std::eprintln!("{}", ascii(&g));
    }

    #[test]
    fn dump_14x14_r7() {
        let g = render_circle(14, 14, 7);
        std::eprintln!("{}", ascii(&g));
    }

    #[test]
    fn dump_50x50_r25() {
        let g = render_circle(50, 50, 25);
        std::eprintln!("{}", ascii(&g));
    }

    #[test]
    fn dump_8x8_r4() {
        let g = render_circle(8, 8, 4);
        std::eprintln!("{}", ascii(&g));
    }

    #[test]
    fn perf_64x64_r32() {
        // Render-time sanity: 64×64 r=32 takes a stable upper bound across
        // 1000 reps. Catches the >100× regression we hit when sqrt was on
        // every pixel; healthy is ~50 µs/frame on a desktop release build.
        use std::time::Instant;
        let mut buf = std::vec![0u8; 64 * 64 * 4];
        let tex = Texture::new(&mut buf, 64, 64, ColorFormat::RGBA8888);
        let mut backend = SwRenderer::new(tex);
        let rect = Rect::new(0, 0, 64, 64);
        let clip = Rect::new(0, 0, 64, 64);
        let t0 = Instant::now();
        for _ in 0..1000 {
            backend.fill_rect(
                &rect,
                &clip,
                &Color::rgb(255, 255, 255),
                Fixed::from_int(32),
                255,
            );
        }
        let elapsed = t0.elapsed();
        let per_frame_us = elapsed.as_secs_f64() * 1e6 / 1000.0;
        std::eprintln!("64x64 r=32: {per_frame_us:.2} µs/frame");
        assert!(
            per_frame_us < 5000.0,
            "corner render too slow: {per_frame_us:.2} µs/frame"
        );
    }

    #[test]
    fn shape_symmetric_horizontal() {
        let g = render_circle(32, 32, 16);
        for (y, row) in g.iter().enumerate() {
            for x in 0..16 {
                let l = row[x];
                let r = row[31 - x];
                assert!(l.abs_diff(r) <= 2, "row {y} x={x}: left {l} vs right {r}",);
            }
        }
    }

    #[test]
    fn shape_symmetric_vertical() {
        let g = render_circle(32, 32, 16);
        for y in 0..16 {
            for x in 0..32 {
                let t = g[y][x];
                let b = g[31 - y][x];
                assert!(t.abs_diff(b) <= 2, "col {x} y={y}: top {t} vs bot {b}",);
            }
        }
    }

    fn count_full_in_row(row: &[u8]) -> usize {
        row.iter().filter(|&&a| a > 200).count()
    }

    fn render_circle_at(
        w: i32,
        h: i32,
        ox: Fixed,
        oy: Fixed,
        fw: Fixed,
        fh: Fixed,
        r: Fixed,
    ) -> Vec<Vec<u8>> {
        let mut buf = std::vec![0u8; (w as usize) * (h as usize) * 4];
        let tex = Texture::new(&mut buf, w as u16, h as u16, ColorFormat::RGBA8888);
        let mut backend = SwRenderer::new(tex);
        let rect = Rect {
            x: ox,
            y: oy,
            w: fw,
            h: fh,
        };
        let clip = Rect::new(0, 0, w, h);
        backend.fill_rect(&rect, &clip, &Color::rgb(255, 255, 255), r, 255);
        let mut out = std::vec![std::vec![0u8; w as usize]; h as usize];
        for py in 0..h {
            for px in 0..w {
                out[py as usize][px as usize] = backend.target.get_pixel(px, py).r;
            }
        }
        out
    }

    #[test]
    fn fractional_origin_rounded_inner_still_solid() {
        // demo-widgets layout often produces fractional area origins
        // (flex SpaceBetween / Center halves a px). Verify the inner
        // straight zone stays fully opaque even when area.x and area.y
        // carry a half-pixel offset.
        let half = Fixed::ONE / 2;
        let ox = Fixed::from_int(2) + half;
        let oy = Fixed::from_int(3) + half;
        let g = render_circle_at(
            48,
            32,
            ox,
            oy,
            Fixed::from_int(40),
            Fixed::from_int(24),
            Fixed::from_int(6),
        );
        // r_int = ceil(6) = 6.
        // mid_x0 = ceil(2.5 + 6) = 9. mid_x1 = (2.5 + 40 - 6).to_int() = 36.
        // mid_y0 = ceil(3.5 + 6) = 10. mid_y1 = (3.5 + 24 - 6).to_int() = 21.
        for y in 10..21 {
            for x in 9..36 {
                assert_eq!(
                    g[y][x], 255,
                    "fractional-origin inner pixel ({x},{y}) not solid: alpha={}",
                    g[y][x]
                );
            }
        }
        // Fractional edges must still anti-alias — top row of the
        // fractional area must be partially transparent (cov_y < 1).
        for x in 9..36 {
            let edge = g[3][x];
            assert!(
                edge > 0 && edge < 255,
                "fractional top-edge pixel ({x},3) should AA but got {edge}"
            );
        }
    }

    #[test]
    fn aligned_rounded_widget_inner_solid() {
        // Typical widget: 40×24 r=6. Exercises the axis-aligned + r>0
        // fast path (inner cross + four corner bboxes). The inner cross
        // (away from any corner bbox) must be fully solid (alpha 255).
        let g = render_circle(40, 24, 6);
        // r_px is ceil(6) = 6, so corner bboxes are [0..6) × [0..6) etc.
        // Inner cross x/y range fully solid: x∈[6, 34), y∈[6, 18).
        for y in 6..18 {
            for x in 6..34 {
                assert_eq!(
                    g[y][x], 255,
                    "inner pixel ({x},{y}) not solid: alpha={}",
                    g[y][x]
                );
            }
        }
        for y in [0, 5, 18, 23] {
            for x in 6..34 {
                assert_eq!(
                    g[y][x], 255,
                    "straight band pixel ({x},{y}) not solid: alpha={}",
                    g[y][x]
                );
            }
        }
        for y in 6..18 {
            for x in [0, 5, 34, 39] {
                assert_eq!(
                    g[y][x], 255,
                    "straight band pixel ({x},{y}) not solid: alpha={}",
                    g[y][x]
                );
            }
        }
    }

    #[test]
    fn aligned_rounded_widget_corner_curve() {
        // 40×24 r=6: top-left corner bbox [0..6) × [0..6). Pixel (0,0)
        // must be transparent (it's the curve's outer extreme), pixel
        // (5,5) must be fully solid (it's on the inside of the curve).
        let g = render_circle(40, 24, 6);
        assert_eq!(g[0][0], 0, "TL corner outer pixel should be empty");
        assert_eq!(g[0][39], 0, "TR corner outer pixel should be empty");
        assert_eq!(g[23][0], 0, "BL corner outer pixel should be empty");
        assert_eq!(g[23][39], 0, "BR corner outer pixel should be empty");
        assert_eq!(g[5][5], 255, "TL corner inner pixel should be solid");
        assert_eq!(g[5][34], 255, "TR corner inner pixel should be solid");
        assert_eq!(g[18][5], 255, "BL corner inner pixel should be solid");
        assert_eq!(g[18][34], 255, "BR corner inner pixel should be solid");
    }

    #[test]
    fn aligned_rounded_is_4way_symmetric() {
        // 40×40 r=8 hits the axis-aligned + r>0 fast path. Output must
        // be 4-way symmetric within ±2 alpha steps so corner SDF logic
        // can't quietly pick up a per-corner asymmetry from clipping
        // arithmetic.
        let g = render_circle(40, 40, 8);
        for y in 0..20 {
            for x in 0..20 {
                let tl = g[y][x];
                let tr = g[y][39 - x];
                let bl = g[39 - y][x];
                let br = g[39 - y][39 - x];
                assert!(
                    tl.abs_diff(tr) <= 2 && tl.abs_diff(bl) <= 2 && tl.abs_diff(br) <= 2,
                    "asymmetry at ({x},{y}): TL={tl} TR={tr} BL={bl} BR={br}"
                );
            }
        }
    }

    #[test]
    fn shape_top_row_narrower_than_middle() {
        // Catches the "circle looks like a flat-top pill" regression: the
        // top row of a 32×32 r=16 circle must be visually narrower than
        // the body, by enough that the curvature is perceivable (≥4 px
        // on each side).
        let g = render_circle(32, 32, 16);
        let top = count_full_in_row(&g[0]);
        let mid = count_full_in_row(&g[16]);
        assert!(
            top < mid,
            "top={top} mid={mid} — top row wider/equal to middle"
        );
        assert!(
            mid >= top + 8,
            "top={top} mid={mid} — corner curvature too flat (mid-top={})",
            mid - top
        );
    }

    fn render_with_color(w: i32, h: i32, r: Fixed, color: Color, opa: u8) -> Vec<Vec<Color>> {
        let mut buf = std::vec![0u8; (w as usize) * (h as usize) * 4];
        let tex = Texture::new(&mut buf, w as u16, h as u16, ColorFormat::RGBA8888);
        let mut backend = SwRenderer::new(tex);
        let rect = Rect::new(0, 0, w, h);
        let clip = Rect::new(0, 0, w, h);
        backend.fill_rect(&rect, &clip, &color, r, opa);
        let mut out = std::vec![std::vec![Color::rgba(0,0,0,0); w as usize]; h as usize];
        for py in 0..h {
            for px in 0..w {
                out[py as usize][px as usize] = backend.target.get_pixel(px, py);
            }
        }
        out
    }

    fn render_at_with_color(
        w: i32,
        h: i32,
        ox: Fixed,
        oy: Fixed,
        fw: Fixed,
        fh: Fixed,
        r: Fixed,
        color: Color,
        opa: u8,
    ) -> Vec<Vec<Color>> {
        let mut buf = std::vec![0u8; (w as usize) * (h as usize) * 4];
        let tex = Texture::new(&mut buf, w as u16, h as u16, ColorFormat::RGBA8888);
        let mut backend = SwRenderer::new(tex);
        let rect = Rect {
            x: ox,
            y: oy,
            w: fw,
            h: fh,
        };
        let clip = Rect::new(0, 0, w, h);
        backend.fill_rect(&rect, &clip, &color, r, opa);
        let mut out = std::vec![std::vec![Color::rgba(0,0,0,0); w as usize]; h as usize];
        for py in 0..h {
            for px in 0..w {
                out[py as usize][px as usize] = backend.target.get_pixel(px, py);
            }
        }
        out
    }

    #[test]
    fn rounded_fill_respects_color_alpha_fast_path() {
        // Half-transparent colour at opa=255 must blend at ~half RGB
        // against an empty canvas (color.a folded into the alpha sent
        // to blend_pixel_int / fill_axis_aligned).
        // 40×24 r=6 + integer origin → axis-aligned + r>0 fast path.
        let half = Color::rgba(255, 255, 255, 128);
        let g = render_with_color(40, 24, Fixed::from_int(6), half, 255);
        // Inner-cross pixel — hits fill_axis_aligned RGBA8888 splat.
        let inner = g[12][20];
        assert!(
            (110..=145).contains(&inner.r),
            "inner pixel r {} should be ~128, not 255",
            inner.r
        );
    }

    #[test]
    fn rounded_fill_respects_color_alpha_aa_loop() {
        // Force the general aa_loop by using a fractional origin: that
        // skips the axis-aligned fast paths and exercises the corner
        // SDF + edge AA branches.
        let half = Color::rgba(255, 255, 255, 128);
        let half_px = Fixed::ONE / 2;
        let g = render_at_with_color(
            48,
            32,
            Fixed::from_int(2) + half_px,
            Fixed::from_int(3) + half_px,
            Fixed::from_int(40),
            Fixed::from_int(24),
            Fixed::from_int(6),
            half,
            255,
        );
        // (15, 15) sits in the straight zone (mid_x=[9,36), mid_y=[10,21))
        // and hits the aa_loop short-circuit set_pixel — but only when
        // opa_full is true. With color.a < 255, opa_full must be false,
        // and the pixel falls through to the Fixed multiply path that
        // writes color.r * effective_opa onto the buffer.
        let inner = g[15][20];
        assert!(
            (110..=145).contains(&inner.r),
            "aa_loop straight-zone pixel r {} should be ~128, not 255",
            inner.r
        );
        // (4, 4) sits inside a corner bbox (mid_x0=9, mid_y0=10), so
        // corner_cov is computed via the SDF and final_opa is
        // cov × effective_opa. Where cov ≈ 1 (well inside the corner
        // disk), r should still be ~128, not 255.
        let corner = g[8][8];
        assert!(
            (90..=145).contains(&corner.r),
            "aa_loop SDF-corner pixel r {} should be near 128, not 255",
            corner.r
        );
    }

    #[test]
    fn rounded_fill_respects_color_alpha_general_loop() {
        // A circle (2r==w==h) doesn't satisfy the inner-cross condition
        // and falls through to the general aa_loop. Different code
        // path from the fractional case above.
        let half = Color::rgba(255, 255, 255, 128);
        let g = render_with_color(32, 32, Fixed::from_int(16), half, 255);
        // Centre of the circle — fully covered, just the alpha fold matters.
        let center = g[16][16];
        assert!(
            (110..=145).contains(&center.r),
            "circle centre r {} should be ~128, not 255",
            center.r
        );
    }

    #[test]
    fn rounded_fill_clip_covers_only_corner() {
        // 40×24 r=6 with clip = (0,0,3,24) leaves only the TL corner
        // bbox visible. Two of the inner-cross fill_axis_aligned calls
        // collapse to negative-width sub-rects; the entry guard must
        // catch that before the usize cast underflows.
        let mut buf = std::vec![0u8; 40 * 24 * 4];
        let tex = Texture::new(&mut buf, 40, 24, ColorFormat::RGBA8888);
        let mut backend = SwRenderer::new(tex);
        let rect = Rect::new(0, 0, 40, 24);
        let clip = Rect::new(0, 0, 3, 24);
        backend.fill_rect(
            &rect,
            &clip,
            &Color::rgb(255, 255, 255),
            Fixed::from_int(6),
            255,
        );
        // The visible vertical strip x∈[0,3) must still draw. (1, 8) is
        // in the middle slab band (away from corners); guard must not
        // have eaten it together with the empty top/bottom bands.
        assert_eq!(
            backend.target.get_pixel(1, 8).a,
            255,
            "clipped middle slab pixel (1,8) should be drawn"
        );
        // (2, 4) is inside the TL corner bbox AND inside the clip;
        // SDF coverage should write at least a partial pixel here.
        let corner = backend.target.get_pixel(2, 4);
        assert!(
            corner.a > 0,
            "TL corner pixel (2,4) should be at least partially drawn, got a={}",
            corner.a
        );
        // Outside clip — must remain untouched.
        assert_eq!(
            backend.target.get_pixel(20, 12).a,
            0,
            "pixel outside clip should not be written"
        );
    }
}