Repository navigation
fix: preserve widget identity across graph compatibility paths - #14233
claude[bot] wants to merge 33 commits into
Conversation
📦 Bundle: 11.9 MB gzip 🔴 +939 BDetailsSummary
Category Glance App Entry Points — 6.62 kB (baseline 6.65 kB) • 🟢 -27 BMain entry bundles and manifests
Status: 1 added / 1 removed Graph Workspace — 2.56 MB (baseline 2.56 MB) • 🟢 -185 BGraph editor runtime, canvas, workflow orchestration
Status: 2 added / 2 removed / 1 unchanged Views & Navigation — 149 kB (baseline 149 kB) • ⚪ 0 BTop-level views, pages, and routed surfaces
Status: 11 added / 11 removed / 7 unchanged Panels & Settings — 872 kB (baseline 872 kB) • 🟢 -27 BConfiguration panels, inspectors, and settings screens
Status: 8 added / 8 removed / 27 unchanged User & Accounts — 154 kB (baseline 154 kB) • 🟢 -59 BAuthentication, profile, and account management bundles
Status: 9 added / 9 removed / 8 unchanged Editors & Dialogs — 502 kB (baseline 502 kB) • 🟢 -54 BModals, dialogs, drawers, and in-app editors
Status: 11 added / 11 removed / 6 unchanged UI Components — 340 kB (baseline 340 kB) • ⚪ 0 BReusable component library chunks
Status: 8 added / 8 removed / 22 unchanged Data & Services — 1.83 MB (baseline 1.83 MB) • 🟢 -142 BStores, services, APIs, and repositories
Status: 26 added / 26 removed / 15 unchanged Utilities & Hooks — 2.45 MB (baseline 2.44 MB) • 🔴 +6.86 kBHelpers, composables, and utility bundles
Status: 42 added / 42 removed / 43 unchanged Vendor & Third-Party — 9.69 MB (baseline 9.69 MB) • ⚪ 0 BExternal libraries and shared vendor chunks Status: 16 unchanged Other — 34.8 MB (baseline 34.8 MB) • 🟢 -1.59 kBBundles that do not match a named category
Status: 95 added / 96 removed / 648 unchanged ⚡ Performance Report
Scenario timings
Show regressions
All metrics
Historical variance (last 15 runs)
Trend (last 15 commits on main)
Raw data{
"timestamp": "2026-10-28T19:08:09.820Z",
"gitSha": "93b63f45bcbcf5e080c0b307ff39959f0607d1dd",
"branch": "report-assertion-failures-to-datadog-rum",
"measurements": [
{
"name": "canvas-idle",
"durationMs": 2007.0430000000101,
"styleRecalcs": 11,
"styleRecalcDurationMs": 3.035,
"layouts": 10,
"layoutDurationMs": 2.7880000000000003,
"taskDurationMs": 469.999,
"heapDeltaBytes": 4185660,
"heapUsedBytes": 74314100,
"domNodes": 0,
"jsHeapTotalBytes": 9437184,
"scriptDurationMs": 9.817,
"eventListeners": 0,
"totalBlockingTimeMs": 0,
"frameDurationMs": 16.66611570247934,
"p95FrameDurationMs": 16.700000000000728
},
{
"name": "canvas-idle",
"durationMs": 2008.6009999999987,
"styleRecalcs": 10,
"styleRecalcDurationMs": 2.6220000000000003,
"layouts": 10,
"layoutDurationMs": 2.575,
"taskDurationMs": 424.60699999999997,
"heapDeltaBytes": -1968296,
"heapUsedBytes": 66211164,
"domNodes": 0,
"jsHeapTotalBytes": 3407872,
"scriptDurationMs": 8.189,
"eventListeners": 0,
"totalBlockingTimeMs": 0,
"frameDurationMs": 16.666115702479342,
"p95FrameDurationMs": 16.700000000000273
},
{
"name": "canvas-mouse-sweep",
"durationMs": 1695.8609999999794,
"styleRecalcs": 73,
"styleRecalcDurationMs": 12.459000000000001,
"layouts": 10,
"layoutDurationMs": 2.6060000000000003,
"taskDurationMs": 739.595,
"heapDeltaBytes": 17603340,
"heapUsedBytes": 79139324,
"domNodes": 0,
"jsHeapTotalBytes": 3407872,
"scriptDurationMs": 124.90799999999999,
"eventListeners": 6,
"totalBlockingTimeMs": 0,
"frameDurationMs": 16.666666666666668,
"p95FrameDurationMs": 16.700000000000273
},
{
"name": "canvas-mouse-sweep",
"durationMs": 1727.6710000001003,
"styleRecalcs": 74,
"styleRecalcDurationMs": 11.345,
"layouts": 10,
"layoutDurationMs": 2.528,
"taskDurationMs": 709.2460000000001,
"heapDeltaBytes": 21732488,
"heapUsedBytes": 83276340,
"domNodes": 0,
"jsHeapTotalBytes": 8650752,
"scriptDurationMs": 123.77700000000002,
"eventListeners": 6,
"totalBlockingTimeMs": 0,
"frameDurationMs": 16.666346153846156,
"p95FrameDurationMs": 16.700000000000273
},
{
"name": "canvas-zoom-sweep",
"durationMs": 1723.9480000000071,
"styleRecalcs": 18,
"styleRecalcDurationMs": 4.6899999999999995,
"layouts": 12,
"layoutDurationMs": 3.055,
"taskDurationMs": 372.965,
"heapDeltaBytes": 19555504,
"heapUsedBytes": 80665732,
"domNodes": 57,
"jsHeapTotalBytes": 3407872,
"scriptDurationMs": 14.462000000000002,
"eventListeners": 43,
"totalBlockingTimeMs": 0,
"frameDurationMs": 16.666346153846156,
"p95FrameDurationMs": 16.700000000000728
},
{
"name": "canvas-zoom-sweep",
"durationMs": 1726.261999999906,
"styleRecalcs": 19,
"styleRecalcDurationMs": 4.615000000000001,
"layouts": 11,
"layoutDurationMs": 2.3939999999999997,
"taskDurationMs": 324.441,
"heapDeltaBytes": 23932684,
"heapUsedBytes": 85423928,
"domNodes": 57,
"jsHeapTotalBytes": 6815744,
"scriptDurationMs": 10.607,
"eventListeners": 43,
"totalBlockingTimeMs": 0,
"frameDurationMs": 16.665384615384614,
"p95FrameDurationMs": 16.700000000000728
},
{
"name": "dom-widget-clipping",
"durationMs": 564.943000000028,
"styleRecalcs": 14,
"styleRecalcDurationMs": 2.6939999999999995,
"layouts": 9,
"layoutDurationMs": 2.3640000000000003,
"taskDurationMs": 420.036,
"heapDeltaBytes": 3539628,
"heapUsedBytes": 65015632,
"domNodes": 0,
"jsHeapTotalBytes": 7602176,
"scriptDurationMs": 139.292,
"eventListeners": 0,
"totalBlockingTimeMs": 0,
"frameDurationMs": 16.66470588235295,
"p95FrameDurationMs": 16.800000000000182
},
{
"name": "dom-widget-clipping",
"durationMs": 561.2669999999298,
"styleRecalcs": 14,
"styleRecalcDurationMs": 2.8739999999999997,
"layouts": 10,
"layoutDurationMs": 2.34,
"taskDurationMs": 405.19300000000004,
"heapDeltaBytes": 9181764,
"heapUsedBytes": 84295124,
"domNodes": 0,
"jsHeapTotalBytes": 10485760,
"scriptDurationMs": 133.71099999999998,
"eventListeners": 0,
"totalBlockingTimeMs": 0,
"frameDurationMs": 16.664705882352937,
"p95FrameDurationMs": 16.700000000000273
},
{
"name": "dom-widget-order-dense-n200-w200-i120",
"durationMs": 1055.720000000008,
"styleRecalcs": 0,
"styleRecalcDurationMs": 0,
"layouts": 0,
"layoutDurationMs": 0,
"taskDurationMs": 741.817,
"heapDeltaBytes": -14001268,
"heapUsedBytes": 75126628,
"domNodes": -203,
"jsHeapTotalBytes": 20717568,
"scriptDurationMs": 50.22599999999999,
"eventListeners": -106,
"totalBlockingTimeMs": 381,
"frameDurationMs": 29.16666666666668,
"p95FrameDurationMs": 16.800000000000182
},
{
"name": "dom-widget-order-dense-n200-w200-i120",
"durationMs": 1069.3899999999985,
"styleRecalcs": 0,
"styleRecalcDurationMs": 0,
"layouts": 0,
"layoutDurationMs": 0,
"taskDurationMs": 747.993,
"heapDeltaBytes": -12465648,
"heapUsedBytes": 74981796,
"domNodes": -203,
"jsHeapTotalBytes": 20717568,
"scriptDurationMs": 56.918,
"eventListeners": -106,
"totalBlockingTimeMs": 394,
"frameDurationMs": 29.630555555555564,
"p95FrameDurationMs": 16.700000000000728
},
{
"name": "dom-widget-order-sparse-n1000-w20-i120",
"durationMs": 847.801000000004,
"styleRecalcs": 0,
"styleRecalcDurationMs": 0,
"layouts": 0,
"layoutDurationMs": 0,
"taskDurationMs": 479.35700000000026,
"heapDeltaBytes": -25343640,
"heapUsedBytes": 66712824,
"domNodes": -203,
"jsHeapTotalBytes": 4325376,
"scriptDurationMs": 132.812,
"eventListeners": -106,
"totalBlockingTimeMs": 162,
"frameDurationMs": 22.36842105263158,
"p95FrameDurationMs": 16.800000000000182
},
{
"name": "dom-widget-order-sparse-n1000-w20-i120",
"durationMs": 813.0550000000767,
"styleRecalcs": 0,
"styleRecalcDurationMs": 0,
"layouts": 0,
"layoutDurationMs": 0,
"taskDurationMs": 464.67899999999986,
"heapDeltaBytes": -26740364,
"heapUsedBytes": 66440840,
"domNodes": -203,
"jsHeapTotalBytes": 4587520,
"scriptDurationMs": 125.22500000000001,
"eventListeners": -106,
"totalBlockingTimeMs": 171,
"frameDurationMs": 23.33142857142858,
"p95FrameDurationMs": 16.800000000000182
},
{
"name": "extension-clean-combined-draw",
"durationMs": 8300.980999999978,
"styleRecalcs": 2,
"styleRecalcDurationMs": 0.543,
"layouts": 2,
"layoutDurationMs": 0.5419999999999999,
"taskDurationMs": 7445.396999999999,
"heapDeltaBytes": 73202660,
"heapUsedBytes": 153215292,
"domNodes": -269,
"jsHeapTotalBytes": 62255104,
"scriptDurationMs": 5790.214,
"eventListeners": -68,
"totalBlockingTimeMs": 30,
"frameDurationMs": 24.29032258064516,
"p95FrameDurationMs": 33.400000000001455
},
{
"name": "extension-clean-combined-draw",
"durationMs": 8420.26199999998,
"styleRecalcs": 1,
"styleRecalcDurationMs": 0.21200000000000038,
"layouts": 1,
"layoutDurationMs": 0.19799999999999982,
"taskDurationMs": 7539.6810000000005,
"heapDeltaBytes": 65369328,
"heapUsedBytes": 145566528,
"domNodes": -269,
"jsHeapTotalBytes": 66711552,
"scriptDurationMs": 5920.1230000000005,
"eventListeners": -68,
"totalBlockingTimeMs": 92,
"frameDurationMs": 24.372543352601156,
"p95FrameDurationMs": 33.400000000000546
},
{
"name": "extension-rgthree-prototype-draw",
"durationMs": 8419.50799999995,
"styleRecalcs": 0,
"styleRecalcDurationMs": 0,
"layouts": 0,
"layoutDurationMs": 0,
"taskDurationMs": 7521.012999999999,
"heapDeltaBytes": 63585824,
"heapUsedBytes": 143830148,
"domNodes": -269,
"jsHeapTotalBytes": 66187264,
"scriptDurationMs": 5876.749,
"eventListeners": -68,
"totalBlockingTimeMs": 158,
"frameDurationMs": 24.465988372093022,
"p95FrameDurationMs": 33.400000000001455
},
{
"name": "extension-rgthree-prototype-draw",
"durationMs": 7877.527999999984,
"styleRecalcs": 1,
"styleRecalcDurationMs": 0.21600000000000005,
"layouts": 1,
"layoutDurationMs": 0.196,
"taskDurationMs": 7013.208,
"heapDeltaBytes": 51869308,
"heapUsedBytes": 131644672,
"domNodes": -269,
"jsHeapTotalBytes": 59633664,
"scriptDurationMs": 5433.6269999999995,
"eventListeners": -68,
"totalBlockingTimeMs": 71,
"frameDurationMs": 24.255384615384614,
"p95FrameDurationMs": 33.400000000000546
},
{
"name": "large-combo-open",
"durationMs": 171.5080000000171,
"styleRecalcs": 21,
"styleRecalcDurationMs": 11.660000000000002,
"layouts": 10,
"layoutDurationMs": 2.585,
"taskDurationMs": 130.97699999999998,
"heapDeltaBytes": 14617516,
"heapUsedBytes": 72874588,
"domNodes": 936,
"jsHeapTotalBytes": 0,
"scriptDurationMs": 6.889999999999999,
"eventListeners": 205,
"totalBlockingTimeMs": 0,
"frameDurationMs": 23.80000000000005,
"p95FrameDurationMs": 66.80000000000018
},
{
"name": "large-combo-open",
"durationMs": 192.97899999992296,
"styleRecalcs": 23,
"styleRecalcDurationMs": 12.981,
"layouts": 10,
"layoutDurationMs": 2.707,
"taskDurationMs": 161.303,
"heapDeltaBytes": 14510528,
"heapUsedBytes": 90896752,
"domNodes": 936,
"jsHeapTotalBytes": 1572864,
"scriptDurationMs": 7.582,
"eventListeners": 205,
"totalBlockingTimeMs": 0,
"frameDurationMs": 22.211111111111123,
"p95FrameDurationMs": 66.69999999999982
},
{
"name": "large-graph-idle",
"durationMs": 2024.575000000027,
"styleRecalcs": 9,
"styleRecalcDurationMs": 2.371,
"layouts": 8,
"layoutDurationMs": 2.0460000000000003,
"taskDurationMs": 616.8260000000001,
"heapDeltaBytes": -3135180,
"heapUsedBytes": 84231572,
"domNodes": -269,
"jsHeapTotalBytes": 3309568,
"scriptDurationMs": 34.711999999999996,
"eventListeners": -68,
"totalBlockingTimeMs": 0,
"frameDurationMs": 16.6655737704918,
"p95FrameDurationMs": 16.800000000000182
},
{
"name": "large-graph-idle",
"durationMs": 2008.6760000000368,
"styleRecalcs": 8,
"styleRecalcDurationMs": 2.1010000000000004,
"layouts": 8,
"layoutDurationMs": 1.9760000000000004,
"taskDurationMs": 546.3019999999999,
"heapDeltaBytes": -25851636,
"heapUsedBytes": 61645304,
"domNodes": -269,
"jsHeapTotalBytes": 3833856,
"scriptDurationMs": 30.264,
"eventListeners": -68,
"totalBlockingTimeMs": 0,
"frameDurationMs": 16.666115702479335,
"p95FrameDurationMs": 16.700000000000728
},
{
"name": "large-graph-pan",
"durationMs": 2538.9519999999948,
"styleRecalcs": 70,
"styleRecalcDurationMs": 9.298,
"layouts": 8,
"layoutDurationMs": 1.8370000000000002,
"taskDurationMs": 1451.7820000000002,
"heapDeltaBytes": -18660832,
"heapUsedBytes": 71208172,
"domNodes": -269,
"jsHeapTotalBytes": 4059136,
"scriptDurationMs": 686.016,
"eventListeners": -66,
"totalBlockingTimeMs": 0,
"frameDurationMs": 16.66578947368421,
"p95FrameDurationMs": 16.700000000000728
},
{
"name": "large-graph-pan",
"durationMs": 2369.1670000000613,
"styleRecalcs": 69,
"styleRecalcDurationMs": 8.516,
"layouts": 8,
"layoutDurationMs": 1.9820000000000002,
"taskDurationMs": 1389.802,
"heapDeltaBytes": -11640624,
"heapUsedBytes": 76992332,
"domNodes": -269,
"jsHeapTotalBytes": 4845568,
"scriptDurationMs": 685.0319999999999,
"eventListeners": -66,
"totalBlockingTimeMs": 0,
"frameDurationMs": 16.666197183098593,
"p95FrameDurationMs": 16.700000000000728
},
{
"name": "large-graph-zoom",
"durationMs": 3788.6140000000523,
"styleRecalcs": 72,
"styleRecalcDurationMs": 10.269,
"layouts": 66,
"layoutDurationMs": 10.944,
"taskDurationMs": 1573.7939999999999,
"heapDeltaBytes": -10648240,
"heapUsedBytes": 79218664,
"domNodes": -245,
"jsHeapTotalBytes": 4096000,
"scriptDurationMs": 681.4720000000001,
"eventListeners": -68,
"totalBlockingTimeMs": 0,
"frameDurationMs": 16.665638766519827,
"p95FrameDurationMs": 16.700000000000728
},
{
"name": "large-graph-zoom",
"durationMs": 3615.7590000000255,
"styleRecalcs": 71,
"styleRecalcDurationMs": 9.769,
"layouts": 64,
"layoutDurationMs": 10.341,
"taskDurationMs": 1530.9580000000003,
"heapDeltaBytes": -19463992,
"heapUsedBytes": 70766520,
"domNodes": -245,
"jsHeapTotalBytes": 5144576,
"scriptDurationMs": 672.296,
"eventListeners": -68,
"totalBlockingTimeMs": 0,
"frameDurationMs": 16.666359447004606,
"p95FrameDurationMs": 16.700000000000728
},
{
"name": "legacy-node-drag",
"durationMs": 2784.6490000000017,
"styleRecalcs": 47,
"styleRecalcDurationMs": 5.7509999999999994,
"layouts": 7,
"layoutDurationMs": 1.703,
"taskDurationMs": 1729.8969999999997,
"heapDeltaBytes": -13035216,
"heapUsedBytes": 78689584,
"domNodes": 0,
"jsHeapTotalBytes": 5541888,
"scriptDurationMs": 853.802,
"eventListeners": 182,
"totalBlockingTimeMs": 0,
"frameDurationMs": 16.766265060240965,
"p95FrameDurationMs": 16.700000000000728
},
{
"name": "legacy-node-drag",
"durationMs": 2645.216000000005,
"styleRecalcs": 45,
"styleRecalcDurationMs": 5.258000000000001,
"layouts": 6,
"layoutDurationMs": 1.349,
"taskDurationMs": 1652.67,
"heapDeltaBytes": 19814940,
"heapUsedBytes": 97428640,
"domNodes": 0,
"jsHeapTotalBytes": 5398528,
"scriptDurationMs": 807.3750000000001,
"eventListeners": 182,
"totalBlockingTimeMs": 0,
"frameDurationMs": 16.771974522292997,
"p95FrameDurationMs": 16.799999999999272
},
{
"name": "minimap-geometry-cadence",
"durationMs": 2969.763999999998,
"styleRecalcs": 7,
"styleRecalcDurationMs": 1.8809999999999998,
"layouts": 7,
"layoutDurationMs": 1.6079999999999999,
"taskDurationMs": 659.9620000000001,
"heapDeltaBytes": -9130460,
"heapUsedBytes": 80538804,
"domNodes": -269,
"jsHeapTotalBytes": 3833856,
"scriptDurationMs": 295.68899999999996,
"eventListeners": -68,
"totalBlockingTimeMs": 0,
"frameDurationMs": 16.66629213483146,
"p95FrameDurationMs": 16.799999999999272
},
{
"name": "minimap-geometry-cadence",
"durationMs": 2998.9570000000185,
"styleRecalcs": 8,
"styleRecalcDurationMs": 1.966,
"layouts": 8,
"layoutDurationMs": 1.939,
"taskDurationMs": 679.56,
"heapDeltaBytes": -9961524,
"heapUsedBytes": 79492344,
"domNodes": -269,
"jsHeapTotalBytes": 3833856,
"scriptDurationMs": 301.399,
"eventListeners": -68,
"totalBlockingTimeMs": 0,
"frameDurationMs": 16.666666666666668,
"p95FrameDurationMs": 16.700000000000728
},
{
"name": "minimap-idle",
"durationMs": 2034.4629999999597,
"styleRecalcs": 8,
"styleRecalcDurationMs": 2.053,
"layouts": 8,
"layoutDurationMs": 1.9940000000000002,
"taskDurationMs": 606.0850000000002,
"heapDeltaBytes": -27329908,
"heapUsedBytes": 63054528,
"domNodes": -269,
"jsHeapTotalBytes": 3833856,
"scriptDurationMs": 31.218999999999998,
"eventListeners": -68,
"totalBlockingTimeMs": 0,
"frameDurationMs": 16.666393442622947,
"p95FrameDurationMs": 16.800000000000182
},
{
"name": "minimap-idle",
"durationMs": 2035.0429999999733,
"styleRecalcs": 8,
"styleRecalcDurationMs": 2.0950000000000006,
"layouts": 8,
"layoutDurationMs": 1.896,
"taskDurationMs": 539.218,
"heapDeltaBytes": -9033808,
"heapUsedBytes": 83380036,
"domNodes": -269,
"jsHeapTotalBytes": 4620288,
"scriptDurationMs": 30.544,
"eventListeners": -68,
"totalBlockingTimeMs": 0,
"frameDurationMs": 16.665573770491804,
"p95FrameDurationMs": 16.700000000000273
},
{
"name": "minimap-progress-execution-cadence",
"durationMs": 2777.3520000000076,
"styleRecalcs": 26,
"styleRecalcDurationMs": 10.917,
"layouts": 7,
"layoutDurationMs": 1.713,
"taskDurationMs": 519.996,
"heapDeltaBytes": -20182080,
"heapUsedBytes": 69561332,
"domNodes": -266,
"jsHeapTotalBytes": 4620288,
"scriptDurationMs": 161.885,
"eventListeners": -68,
"totalBlockingTimeMs": 0,
"frameDurationMs": 16.665868263473055,
"p95FrameDurationMs": 16.799999999999272
},
{
"name": "minimap-progress-execution-cadence",
"durationMs": 2745.284999999967,
"styleRecalcs": 27,
"styleRecalcDurationMs": 10.753,
"layouts": 8,
"layoutDurationMs": 1.8460000000000003,
"taskDurationMs": 512.6909999999999,
"heapDeltaBytes": -19991168,
"heapUsedBytes": 72852668,
"domNodes": -259,
"jsHeapTotalBytes": 4620288,
"scriptDurationMs": 153.57899999999998,
"eventListeners": -68,
"totalBlockingTimeMs": 0,
"frameDurationMs": 16.665454545454548,
"p95FrameDurationMs": 16.800000000000182
},
{
"name": "minimap-topology-cadence",
"durationMs": 3128.6380000000236,
"styleRecalcs": 8,
"styleRecalcDurationMs": 2.1950000000000007,
"layouts": 8,
"layoutDurationMs": 2.213,
"taskDurationMs": 817.076,
"heapDeltaBytes": -3934892,
"heapUsedBytes": 87849580,
"domNodes": -269,
"jsHeapTotalBytes": 5931008,
"scriptDurationMs": 328.783,
"eventListeners": -68,
"totalBlockingTimeMs": 0,
"frameDurationMs": 16.665957446808513,
"p95FrameDurationMs": 16.700000000000728
},
{
"name": "minimap-topology-cadence",
"durationMs": 3079.0120000000343,
"styleRecalcs": 7,
"styleRecalcDurationMs": 1.7830000000000008,
"layouts": 7,
"layoutDurationMs": 1.55,
"taskDurationMs": 804.176,
"heapDeltaBytes": -18620600,
"heapUsedBytes": 71495488,
"domNodes": -269,
"jsHeapTotalBytes": 4620288,
"scriptDurationMs": 303.392,
"eventListeners": -68,
"totalBlockingTimeMs": 0,
"frameDurationMs": 16.665945945945946,
"p95FrameDurationMs": 16.700000000000728
},
{
"name": "subgraph-dom-widget-clipping",
"durationMs": 573.3599999999797,
"styleRecalcs": 46,
"styleRecalcDurationMs": 4.781000000000001,
"layouts": 9,
"layoutDurationMs": 2.1710000000000003,
"taskDurationMs": 381.36,
"heapDeltaBytes": -1015852,
"heapUsedBytes": 61284488,
"domNodes": 0,
"jsHeapTotalBytes": 8388608,
"scriptDurationMs": 123.279,
"eventListeners": 46,
"totalBlockingTimeMs": 0,
"frameDurationMs": 16.66571428571429,
"p95FrameDurationMs": 16.700000000000273
},
{
"name": "subgraph-dom-widget-clipping",
"durationMs": 565.26800000006,
"styleRecalcs": 47,
"styleRecalcDurationMs": 4.752,
"layouts": 10,
"layoutDurationMs": 2.3390000000000004,
"taskDurationMs": 360.819,
"heapDeltaBytes": 676984,
"heapUsedBytes": 63210960,
"domNodes": 0,
"jsHeapTotalBytes": 8912896,
"scriptDurationMs": 117.38200000000002,
"eventListeners": 46,
"totalBlockingTimeMs": 0,
"frameDurationMs": 16.664705882352937,
"p95FrameDurationMs": 16.700000000000273
},
{
"name": "subgraph-idle",
"durationMs": 2009.5360000000255,
"styleRecalcs": 10,
"styleRecalcDurationMs": 2.905,
"layouts": 10,
"layoutDurationMs": 2.652,
"taskDurationMs": 470.23900000000003,
"heapDeltaBytes": 2501548,
"heapUsedBytes": 64971724,
"domNodes": 0,
"jsHeapTotalBytes": 8126464,
"scriptDurationMs": 9.655999999999999,
"eventListeners": 0,
"totalBlockingTimeMs": 0,
"frameDurationMs": 16.666942148760327,
"p95FrameDurationMs": 16.799999999999727
},
{
"name": "subgraph-idle",
"durationMs": 2008.7149999999383,
"styleRecalcs": 11,
"styleRecalcDurationMs": 2.544,
"layouts": 10,
"layoutDurationMs": 2.364,
"taskDurationMs": 424.275,
"heapDeltaBytes": 2824136,
"heapUsedBytes": 64615308,
"domNodes": 0,
"jsHeapTotalBytes": 9437184,
"scriptDurationMs": 8.034999999999998,
"eventListeners": 0,
"totalBlockingTimeMs": 0,
"frameDurationMs": 16.66528925619835,
"p95FrameDurationMs": 16.700000000000728
},
{
"name": "subgraph-mouse-sweep",
"durationMs": 1677.5480000000016,
"styleRecalcs": 75,
"styleRecalcDurationMs": 13.325,
"layouts": 9,
"layoutDurationMs": 2.2279999999999993,
"taskDurationMs": 694.6909999999999,
"heapDeltaBytes": 13369996,
"heapUsedBytes": 75553756,
"domNodes": 0,
"jsHeapTotalBytes": 3932160,
"scriptDurationMs": 95.129,
"eventListeners": 6,
"totalBlockingTimeMs": 0,
"frameDurationMs": 16.666336633663363,
"p95FrameDurationMs": 16.700000000000273
},
{
"name": "subgraph-mouse-sweep",
"durationMs": 1677.090000000021,
"styleRecalcs": 76,
"styleRecalcDurationMs": 13.378999999999998,
"layouts": 10,
"layoutDurationMs": 2.3080000000000003,
"taskDurationMs": 690.183,
"heapDeltaBytes": 19432688,
"heapUsedBytes": 81424616,
"domNodes": 0,
"jsHeapTotalBytes": 10223616,
"scriptDurationMs": 91.243,
"eventListeners": 6,
"totalBlockingTimeMs": 0,
"frameDurationMs": 16.666336633663366,
"p95FrameDurationMs": 16.700000000000728
},
{
"name": "subgraph-transition-enter",
"durationMs": 1371.5449999999691,
"styleRecalcs": 19,
"styleRecalcDurationMs": 18.498,
"layouts": 5,
"layoutDurationMs": 10.839999999999998,
"taskDurationMs": 960.3970000000002,
"heapDeltaBytes": 38271272,
"heapUsedBytes": 112588524,
"domNodes": 16029,
"jsHeapTotalBytes": 12718080,
"scriptDurationMs": 27.651999999999997,
"eventListeners": 3246,
"totalBlockingTimeMs": 133,
"frameDurationMs": 19.21111111111111,
"p95FrameDurationMs": 16.800000000000182
},
{
"name": "viewport-pan-sweep",
"durationMs": 10397.212000000025,
"styleRecalcs": 251,
"styleRecalcDurationMs": 30.762999999999998,
"layouts": 9,
"layoutDurationMs": 2.129,
"taskDurationMs": 5114.172,
"heapDeltaBytes": -14663016,
"heapUsedBytes": 72160448,
"domNodes": -269,
"jsHeapTotalBytes": 4812800,
"scriptDurationMs": 2309.87,
"eventListeners": -66,
"totalBlockingTimeMs": 0,
"frameDurationMs": 16.66602564102564,
"p95FrameDurationMs": 16.700000000000728
},
{
"name": "viewport-pan-sweep",
"durationMs": 8992.407999999956,
"styleRecalcs": 251,
"styleRecalcDurationMs": 29.162000000000003,
"layouts": 9,
"layoutDurationMs": 2.176,
"taskDurationMs": 4825.488,
"heapDeltaBytes": -4586088,
"heapUsedBytes": 85599448,
"domNodes": -269,
"jsHeapTotalBytes": 4321280,
"scriptDurationMs": 2215.405,
"eventListeners": -66,
"totalBlockingTimeMs": 0,
"frameDurationMs": 16.66611111111111,
"p95FrameDurationMs": 16.700000000000728
},
{
"name": "vue-large-graph-idle",
"durationMs": 19802.344000000005,
"styleRecalcs": 0,
"styleRecalcDurationMs": 0,
"layouts": 0,
"layoutDurationMs": 0,
"taskDurationMs": 19137.941,
"heapDeltaBytes": -9201936,
"heapUsedBytes": 274257016,
"domNodes": -3343,
"jsHeapTotalBytes": 10477568,
"scriptDurationMs": 265.804,
"eventListeners": -653,
"totalBlockingTimeMs": 0,
"frameDurationMs": 35.93321234119783,
"p95FrameDurationMs": 66.70000000000073
},
{
"name": "vue-large-graph-idle",
"durationMs": 19754.57299999994,
"styleRecalcs": 0,
"styleRecalcDurationMs": 0,
"layouts": 0,
"layoutDurationMs": 0,
"taskDurationMs": 19133.792,
"heapDeltaBytes": 3730556,
"heapUsedBytes": 281206120,
"domNodes": -3343,
"jsHeapTotalBytes": 12312576,
"scriptDurationMs": 277.401,
"eventListeners": -653,
"totalBlockingTimeMs": 0,
"frameDurationMs": 35.8426497277677,
"p95FrameDurationMs": 66.70000000000073
},
{
"name": "vue-large-graph-pan",
"durationMs": 23757.550999999978,
"styleRecalcs": 180,
"styleRecalcDurationMs": 22.38799999999999,
"layouts": 0,
"layoutDurationMs": 0,
"taskDurationMs": 23497.926,
"heapDeltaBytes": -12973564,
"heapUsedBytes": 279037412,
"domNodes": -3343,
"jsHeapTotalBytes": 16769024,
"scriptDurationMs": 743.825,
"eventListeners": -651,
"totalBlockingTimeMs": 0,
"frameDurationMs": 45.583685220729365,
"p95FrameDurationMs": 83.40000000000146
},
{
"name": "vue-large-graph-pan",
"durationMs": 23432.85400000002,
"styleRecalcs": 179,
"styleRecalcDurationMs": 20.595000000000002,
"layouts": 0,
"layoutDurationMs": 0,
"taskDurationMs": 23120.976,
"heapDeltaBytes": -2958156,
"heapUsedBytes": 279362076,
"domNodes": -3343,
"jsHeapTotalBytes": 17293312,
"scriptDurationMs": 751.591,
"eventListeners": -651,
"totalBlockingTimeMs": 3,
"frameDurationMs": 44.21188679245284,
"p95FrameDurationMs": 83.40000000000146
},
{
"name": "vue-node-resize-workload",
"durationMs": 7160.0160000000415,
"styleRecalcs": 20,
"styleRecalcDurationMs": 51.247000000000014,
"layouts": 20,
"layoutDurationMs": 21.999999999999993,
"taskDurationMs": 7046.365999999999,
"heapDeltaBytes": -19189712,
"heapUsedBytes": 263174024,
"domNodes": -3343,
"jsHeapTotalBytes": 24109056,
"scriptDurationMs": 104.957,
"eventListeners": -653,
"totalBlockingTimeMs": 11,
"frameDurationMs": 42.65714285714286,
"p95FrameDurationMs": 83.40000000000146
},
{
"name": "vue-node-resize-workload",
"durationMs": 7129.597999999987,
"styleRecalcs": 20,
"styleRecalcDurationMs": 57.284,
"layouts": 20,
"layoutDurationMs": 26.673000000000002,
"taskDurationMs": 7055.849,
"heapDeltaBytes": 9036808,
"heapUsedBytes": 282796656,
"domNodes": -3343,
"jsHeapTotalBytes": 24109056,
"scriptDurationMs": 100.75900000000001,
"eventListeners": -653,
"totalBlockingTimeMs": 0,
"frameDurationMs": 41.568604651162794,
"p95FrameDurationMs": 83.39999999999964
},
{
"name": "vue-zoom-out-idle",
"durationMs": 9837.725999999975,
"styleRecalcs": 0,
"styleRecalcDurationMs": 0,
"layouts": 0,
"layoutDurationMs": 0,
"taskDurationMs": 9486.802,
"heapDeltaBytes": -15770356,
"heapUsedBytes": 264802548,
"domNodes": 0,
"jsHeapTotalBytes": 0,
"scriptDurationMs": 123.35499999999999,
"eventListeners": 0,
"totalBlockingTimeMs": 0,
"frameDurationMs": 35.94744525547445,
"p95FrameDurationMs": 66.70000000000073
},
{
"name": "vue-zoom-out-idle",
"durationMs": 9774.269000000004,
"styleRecalcs": 0,
"styleRecalcDurationMs": 0,
"layouts": 0,
"layoutDurationMs": 0,
"taskDurationMs": 9431.197999999999,
"heapDeltaBytes": 349620,
"heapUsedBytes": 272874832,
"domNodes": 0,
"jsHeapTotalBytes": 0,
"scriptDurationMs": 125.12600000000002,
"eventListeners": 0,
"totalBlockingTimeMs": 0,
"frameDurationMs": 34.81459074733095,
"p95FrameDurationMs": 66.70000000000073
},
{
"name": "workflow-execution",
"durationMs": 109.95900000000347,
"styleRecalcs": 8,
"styleRecalcDurationMs": 8.805,
"layouts": 8,
"layoutDurationMs": 3.0509999999999997,
"taskDurationMs": 81.716,
"heapDeltaBytes": 4748568,
"heapUsedBytes": 66274280,
"domNodes": 127,
"jsHeapTotalBytes": 0,
"scriptDurationMs": 7.857000000000001,
"eventListeners": 3,
"totalBlockingTimeMs": 0,
"frameDurationMs": 16.671428571428546,
"p95FrameDurationMs": 16.799999999999727
},
{
"name": "workflow-execution",
"durationMs": 100.67900000001373,
"styleRecalcs": 8,
"styleRecalcDurationMs": 7.785,
"layouts": 7,
"layoutDurationMs": 2.486,
"taskDurationMs": 74.24399999999999,
"heapDeltaBytes": -12930640,
"heapUsedBytes": 63305932,
"domNodes": 125,
"jsHeapTotalBytes": 3670016,
"scriptDurationMs": 4.756,
"eventListeners": 5,
"totalBlockingTimeMs": 0,
"frameDurationMs": 16.666666666666668,
"p95FrameDurationMs": 16.700000000000273
},
{
"name": "agent-composer-empty-conversation-typing",
"durationMs": 282.2569999999587,
"styleRecalcs": 5,
"styleRecalcDurationMs": 0.615,
"layouts": 76,
"layoutDurationMs": 11.759,
"taskDurationMs": 250.51199999999994,
"heapDeltaBytes": -6252284,
"heapUsedBytes": 35971424,
"domNodes": 4,
"jsHeapTotalBytes": 2621440,
"scriptDurationMs": 111.14200000000001,
"eventListeners": 292,
"totalBlockingTimeMs": 0,
"frameDurationMs": 16.666666666666668,
"p95FrameDurationMs": 16.699999999999818
},
{
"name": "agent-composer-empty-conversation-typing",
"durationMs": 308.8270000000648,
"styleRecalcs": 5,
"styleRecalcDurationMs": 0.6719999999999999,
"layouts": 76,
"layoutDurationMs": 12.884,
"taskDurationMs": 266.34400000000005,
"heapDeltaBytes": -5852880,
"heapUsedBytes": 36537036,
"domNodes": 4,
"jsHeapTotalBytes": 3407872,
"scriptDurationMs": 122.31599999999999,
"eventListeners": 292,
"totalBlockingTimeMs": 0,
"frameDurationMs": 16.66111111111109,
"p95FrameDurationMs": 16.700000000000728
},
{
"name": "agent-composer-long-conversation-typing",
"durationMs": 411.39399999997295,
"styleRecalcs": 5,
"styleRecalcDurationMs": 2.247,
"layouts": 76,
"layoutDurationMs": 13.360000000000001,
"taskDurationMs": 366.75600000000003,
"heapDeltaBytes": -910972,
"heapUsedBytes": 92935128,
"domNodes": 4,
"jsHeapTotalBytes": 528384,
"scriptDurationMs": 133.92399999999998,
"eventListeners": 292,
"totalBlockingTimeMs": 0,
"frameDurationMs": 16.666666666666668,
"p95FrameDurationMs": 16.699999999999818
},
{
"name": "agent-composer-long-conversation-typing",
"durationMs": 378.680999999915,
"styleRecalcs": 5,
"styleRecalcDurationMs": 2.177,
"layouts": 76,
"layoutDurationMs": 11.610000000000001,
"taskDurationMs": 349.723,
"heapDeltaBytes": -768380,
"heapUsedBytes": 88074320,
"domNodes": 4,
"jsHeapTotalBytes": 0,
"scriptDurationMs": 126.794,
"eventListeners": 292,
"totalBlockingTimeMs": 0,
"frameDurationMs": 16.660869565217382,
"p95FrameDurationMs": 16.800000000000182
}
]
}🔬 E2E Coverage
🎭 Playwright: ✅ 3078 passed, 0 failed · 1 flaky📊 Browser Reports
🎨 Storybook: ✅ Built — View Storybook🧩 Custom Nodes Ecosystem Matrix: 🟡 Withheld — could not measureExecutes the frontend JS of every registry pack that ships any, against this change, in both renderers. The eight
Findings and check detailsThe matrix runs on same-repository version-bump PRs with relevant changes, main-branch pushes, and manual dispatches. This candidate does not qualify, so every expensive stage was skipped. Reproduce locally# build one shard's fixture, then run it
python3 scripts/registry-census/build_matrix.py --shard 1/4
MATRIX_RENDERER=vue MATRIX_VUE=1 \
npx vitest run --config vitest.matrix.config.mts
Full docs: scripts/registry-census/README.md 🌾 Fallow: ✅ PassedFallow audits the diff for duplication, complexity and dead code. It gates on new findings only, so when the audit completes, a red is something this PR introduced or moved — never inherited backlog. A red can also mean the audit itself did not finish; the details below say which happened. Findings and check detailsNo new findings in the changed files. Reproduce locally# the base the gate compared against
npx fallow audit --root . --quiet --format json \
--changed-since 79bffa34870215c37c2c3fed586a1ee6af980f55
Full docs: ADR-DEVEX-LINT-0015 🏛️ Architecture: ✅ PassedFindings and check detailsWarnings (269)
Reproduce locallypnpm architecture:checkFull docs: Frontend domain catalog |
Codecov Report❌ Patch coverage is @@ Coverage Diff @@
## main #14233 +/- ##
==========================================
+ Coverage 86.66% 87.63% +0.96%
==========================================
Files 3910 2752 -1158
Lines 182231 151882 -30349
Branches 47504 41032 -6472
==========================================
- Hits 157936 133095 -24841
+ Misses 21840 18112 -3728
+ Partials 2455 675 -1780
Flags with carried forward coverage won't be shown. Click here to find out more.
... and 1159 files with indirect coverage changes 🚀 New features to boost your workflow:
|
|
Important Review skippedBot user detected. To trigger a single review, invoke the
|
f49970a to
153e2ae
Compare
153e2ae to
659bed9
Compare
There was a problem hiding this comment.
🔍 Cursor Review — Consolidated panel
Triggered by @christian-byrne.
Found 8 finding(s).
| Severity | Count |
|---|---|
| 🟠 High | 1 |
| 🟡 Medium | 3 |
| 🟢 Low | 3 |
| ⚪ Nit | 1 |
Panel: 6/6 reviewers contributed findings.
Addresses review feedback: #14233 (comment)
Addresses review feedback: #14233 (comment)
Addresses review feedback: #14233 (comment)
There was a problem hiding this comment.
Verdict: changes required. The original identity failures are fixed, but type refresh now loses wrapper-local state, leaves stale input bindings, and can register the wrong widget type in the store.
Reviewed 6df6c72, following the previous review. The request-changes verdict is for reproduced widget defects, not the cleanup nits.
- P2: hiding or retyping the installed wrapper is mistaken for a change to the source; re-adding the source replaces the wrapper and loses that state.
- P2: replacement skips removal cleanup and leaves an input's
_widgetpointing at a detached wrapper. - P2:
syncWidgetOrderbinds its pre-deduplication copy. After a source changes to toggle and is unshifted, the installed widget is toggle but the store contains number. - P2: the new reorder regression test passes against the previous broken implementation. A splice-out/reinsert variant fails on the old code and passes on this head.
1. Still open: the title and description describe a different diff
The previous P2 description finding remains unanswered, and the body is unchanged. This diff does not add main.ts telemetry wiring or the LGraph.add guard. The named reportAssertionFailure.ts helper does not exist. The listed tests and Review Focus questions are stale.
Describe the widget identity and refresh behavior, subgraph link-id repair, and slot normalization actually present here. Remove claims of new layout assertions: the relevant layout no-ops already exist on base. The removeInput(-1) corruption claim was also wrong; the old guard already returned, unlike removeOutput.
"Breaking: none" omits both asserts condition becoming void and accessors that now throw TypeError in production instead of reporting then returning undefined. This round removes the assertion reporter path from useApp as well as useExtensionDialogService. State the runtime-only digit-string compatibility policy from the author's reply.
2. Still open: the test-boundary reply addresses only the public signature
I accept the reply that the public methods should remain slot: number. Runtime compatibility need not become supported TypeScript input.
The remaining parts of the finding still apply. The tests retain copied input/output bodies and redundant not.toThrow wrappers, and this round increases double assertions from ten to eleven. Model the untyped caller explicitly, for example Reflect.apply(testNode.removeOutput, testNode, [slot]) with slot: unknown, rather than using as unknown as number. Table the variants and assert the resulting slots and callbacks. This does not require widening the production API.
3. Previous inline findings at this head
All 19 earlier inline comments have author replies. There are no addressed-but-unanswered threads requiring another acknowledgment, so I have not added duplicate replies or resolved any thread.
| Earlier finding | Disposition |
|---|---|
| Insertion-order duplicate | Addressed by 760bc98. Existing-wrapper membership is considered across the full list. |
| Identity lost during reorder | Production fix confirmed in 0b8433d. The new regression test needs the stronger sequence reported inline. |
| Stale cached type | Original case fixed by 4e932f3. New refresh regressions are separate findings, not a claim that this original fix failed. |
| Dialog import timeout | Addressed by dff6ab8. The focused suite now passes. |
| App registry double-report/type inconsistency | Addressed by 54227a8. Single TypeError, with a test. |
| Repeated caller bounds checks | Addressed by 7c8f4b4. One optional simplification inside the helper is inline. |
| Repeated membership scan | Addressed by 1183a4a. |
| Pre-pushed positional restoration coverage | Addressed by 68f1bf6. Wrong early-return and count-all mutations fail the new test. |
| Repeated widget fixture | Addressed in LGraphNode tests by 16d089a, but six copies now exist in widgetsView tests. Follow-through reported inline. |
| View test placement | Addressed by fc5d23a. |
| Double casts and copied slot tests | Partially answered; see section 2. |
| Connected digit-string removal | Addressed by 70984a0. Reverting the reindex expression fails the test. |
| Subgraph link preconditions | Addressed by 244f946. |
| Missing-node operation table | Addressed by 3e9a166; redundant re-registration wrapper also removed. |
| Replay-test narrative | Addressed by b8ebca9. |
| Layout no-op comment | Addressed by 4e6c496. |
| Membership comment | Addressed by 8944966. |
| Dialog one-use message | Addressed by 03c31a2. |
| Privacy versus grouping wording | Addressed by 6df6c72. |
The earlier 39 root threads from DrJKL and the automated review rounds were also checked, without a time filter. Their addressed/superseded dispositions from the previous review still hold. The previously unverified multiple-live-link question is now a deletion finding based on link-store ingress validation, reported inline. I accept the author's assertion-signature explanation; void is sound, and a rename is not required for this review.
4. Remaining policy questions
- P3: raw reinsertion keeps the installed position when the source type is unchanged, but takes the insertion position when the source type changed. A raw push moves the refreshed widget to the end; unshift moves it to the front. This also changes positional widgets_values order. Prefer refreshing in the existing slot, as addCustomWidget does, or explicitly define and test the different policy.
- The prior filtered-link lookup question remains unanswered. connect filters the mirror while inherited getLinks still serves the SubgraphNode consumers. Keep filtering local to repair unless changing those consumers is justified; a global override is not an automatic cleanup.
5. Verification, rejected candidates, and limitations
- Focused Vitest run over nine affected suites: 163/163 passed. Broader litegraph control: 850/850 passed.
- A new behavioral probe on this head failed: after source number-to-toggle mutation and raw unshift, installed type was toggle but the store returned number. This is a product-state assertion, not just a mocked call count.
- Cross-commit probes showed the hidden wrapper retained on the prior head but replaced and visible on this head. Replacement also left
_widgetdetached and called onRemove zero times. Those probes recorded observed state through deliberate sentinel failures; their nonzero exit alone is not the evidence. - Mutation checks confirm the new pre-pushed restoration and connected-slot tests reject the old implementations. The new reorder test does not; splice-move and destructuring-swap variants do.
- Scoped oxfmt passed all 14 delta files. Scoped ESLint exited zero but ignored seven test files. Scoped type-aware oxlint covered the paths and exited zero with ten pre-existing architecture-import warnings. Fallow's four unchanged widgetMap boundary findings remain baseline noise.
- Rejected: exploiting method bivariance as a replacement for double casts; reverting progressive widget conversion without preserving deduplication; deleting the supported raw refresh path merely because its test is missing; re-raising the unchanged CRDT helper guard; imposing an extra import ordering convention despite the formatter passing.
- All five review lanes examined the delta and the whole PR. Some maintainability principle lenses were applied without loading their skill text, so this is not a claim of exhaustive specialist coverage. No browser, full application build, or repository-wide typecheck was run under the repository's scoped-check policy.
- CI was still running when inspected: 85 successful checks, 40 skipped, 13 without a conclusion. This review does not claim a completed green CI run.
- No new owner comments exposed a matching review-skill gap. No gap catch or miss is inferred from silence.
|
Addressed the remaining review summary at 1f02468:
The focused widget, node, subgraph, and widget-map suites pass 126/126. The guarded commit and push also passed formatting, type-aware lint, ESLint, app typecheck, Knip, and changed-file Fallow. |
All substantive findings from this automated round were fixed in 1f02468. Every inline thread has a commit-linked reply and is resolved. Red-before regressions now cover wrapper-local type retention and direct refresh lifecycle, input rebinding, installed position, and final store type; 192 changed-area tests and the 2,230-test LiteGraph control pass. The stale PR metadata and runtime compatibility notes were also corrected.
There was a problem hiding this comment.
Verdict: changes required. Normal refresh is fixed, but extension teardown can now corrupt the widget list or leave an input unbound; the reorder regression test still misses the original bug.
Reviewed 1f02468 after the previous review. The request-changes verdict rests on the reproduced callback failures. Cleanup nits are non-blocking.
- P2: throwing
onRemoveduring raw refresh leaves two toggle wrappers and the old number store entry. ThroughaddCustomWidget, it clears_widgetwithout restoring the binding. - P2: an
onRemovethat splices another widget runs before reconciliation finishes. Raw refresh leaves duplicate widget IDs;addCustomWidgetuses a stale index and overwrites an unrelated widget. - P2: the splice/reinsert edit went into the neighboring test that never re-adds the source. Both committed reorder tests still pass against the pre-fix implementation.
1. Previous findings and author replies
All 13 previous inline threads have author replies, so none needs an acknowledgment reply. I have not resolved any thread.
| Previous finding | Disposition at this head |
|---|---|
| Wrapper-local hidden state | Fixed. Recorded source type now decides refresh. The hidden-state probe passes here and fails before. |
| Outgoing cleanup and input binding | Fixed for nonthrowing, nonmutating callbacks on both entry paths. The new exception/reentrancy failures are inline below. |
| Wrong store type after raw refresh | Fixed on the normal path. Both push and unshift preserve the toggle type and installed order. |
| Reorder regression test | Still open. The suggested act block was applied to the wrong test. The inline pair puts it in the test with addCustomWidget. |
| Nested reverse map | Map flattened. The one-caller rememberConcreteWidget helper remains despite the reply saying it was replaced; residual inline. |
| Redundant membership arm | Fixed. Removing the source-equality arm is equivalent for adopted, pass-through and wrapped widgets. |
| Cached/installed naming | Fixed. |
| Slot helper duplication | Fixed. Prior/head implementations agree on 27 boundary inputs. |
| Partial dedupe narrative | Deleted. No production comment blocks added this round. |
| Duplicate link precondition | Fixed with assert.exists. |
| Repeated fixture | Consolidated. The new unused override parameter is an inline nit. |
| Missing invalid-slot callback coverage | Fixed. Removing the upper-bound guard now fails output rows 2 and '2'. The cross-product and duplicate assertion can be reduced. |
| Unreachable extra-live-link recovery | Removed; target uniqueness remains enforced at link-store ingress. The now-unnecessary array collector is a separate simplification. |
The previous body findings also changed: the title and description now describe this PR; all eleven double casts are gone; raw refresh preserves the installed slot and positional values; filtering remains local to SubgraphOutput.connect. Earlier addressed/superseded findings retain their dispositions recorded in the previous review. I re-read the fetched review and comment history without excluding older unresolved discussions.
2. Remaining description corrections and scope
- P3: replace "No UI change; no browser or Agent behavior changed." This PR deliberately changes widget visibility/identity and production singleton failures. No template or CSS change is not the same as no browser behavior change. Say that no browser or Agent end-to-end verification was performed, and name the unit-tested behavior instead. Agent consumers read the changed widget lists; that is not proof of an Agent regression, but it rules out a blanket unchanged-behavior assurance.
- P3: bound "across direct array mutation" to re-adding a source while its wrapper remains installed. Assigning a raw source over its own wrapper, replacing it with splice, or replacing the whole list still loses wrapper-local state. These probes fail identically on base and head, so I am not asking this PR to fix them; the description should not promise otherwise.
- P3: compatibility notes should include invalid output indices becoming no-ops and accepted digit strings reaching removal callbacks as numbers. On base,
removeOutput(-1)removed the last output, out-of-range output removal fired a callback, and digit-string callbacks received a string. - P3: the template asks for a one- or two-sentence Summary. Keep the widget behavior there and move secondary changes to compatibility notes. Optional scope reduction: the
assertsignature and singleton TypeErrors can be reviewed separately from graph/widget compatibility. They share no implementation path with refresh.
3. Cleanup contract still needs a decision
P3: clearing each input's _widget before teardown and setting it again afterward has no demonstrated in-repository consumer. A single rebind passes 103 focused tests. External callbacks can inspect it, however, so that search does not prove the intermediate value is irrelevant. Define whether callbacks see the old, new, or absent binding, then implement that once alongside the exception/reentrancy fixes. Do not add a test that merely freezes the current transient undefined state without a contract.
The inline source-owned-callback question is related but distinct: a replacement inherits the same source callback and resources. My previous request for outgoing cleanup was too broad if interpreted as permission to dispose resources still owned by the replacement. Separate outgoing wrapper state from source-owned resources.
4. Verification and rejected candidates
NODE_OPTIONS=--no-experimental-webstorage pnpm test:unitover the four widget/node/subgraph suites: 126/126 passed. The other five affected assertion/accessor/layout/Agent suites: 57/57 passed. Total 183 tests across nine suites.- Additional probes verified the normal-path fixes on head and the failures before. Callback probes observed duplicate toggle widgets with a number store entry, a cleared input, and loss of the unrelated
afterwidget. Some probe assertions deliberately expose diagnostic state; the actual received values, not an exit code alone, support the findings. - Both committed reorder tests pass against 760bc98, before the identity fix. The splice-then-readd probe fails there and passes here. Mutation checks also verified the refresh/input/callback/visibility tests and exposed the missing renamed-source store-release assertion.
- Simplification experiments passed 103 tests for source deduplication, 126 for the identity predicate and helper removal, and 374 for single-link lookup. Negative controls fail when deduplication, refresh or link filtering is removed. These experiments do not settle callback reentrancy.
- Scoped formatting, oxlint and ESLint exited zero. ESLint ignores the two test files; oxlint emitted nine existing architecture warnings. Scratch type proofs covered the test narrowing. Fallow reported only the four previously recorded import-boundary findings.
- Rejected as pre-existing: slot-replacing raw-source identity loss and unmatched
afterChangeon an otherwise empty subgraph reconnect. Rejected a demand to retract the PR's red-before claim merely because an earlier assertion stops a committed test: independent prior-round probes support those failures. The PR explicitly lists focused checks, so omission of other suites is not a false claim; I ran those suites above. Type casts that remain are supported by the recorded source type.Reflect.applyintentionally exercises the untyped boundary; I am not reopening that choice. - Standard-depth correctness, maintainability, tests, idiomaticity and documentation lanes ran. The documentation lane applied prose checks without loading every specialist; no claim of exhaustive lens coverage. No browser, full build or repository-wide typecheck was run. CI snapshot: 72 success, 28 skipped, 26 without a conclusion.
- No new owner comment exposed a skill gap. No catch or miss is inferred from silence.
Superseded by c64509c: all 16 findings were addressed with thread-specific replies and resolved. The two P2 callback/reentrancy paths now have red-before regressions for both addCustomWidget and raw-array refresh; 187 changed-area tests and the 2,225-test LiteGraph control pass. Awaiting a fresh exact-head automated verdict.
There was a problem hiding this comment.
Verdict: changes required. Callback-safe refresh is fixed, but cleanup attached to the returned wrapper is now lost and never runs on removal.
Reviewed c64509c after the previous review. The P2 lifecycle finding is the reason for requesting changes; the other items are non-blocking.
The chosen boundary is sound: refresh does not execute extension teardown. The missing case is an extension attaching or chaining onRemove on the installed wrapper after addCustomWidget returns it. Refresh reconstructs from the source, discards the wrapper-local handler, and real removal cannot reach it. Independent probes found zero local cleanup calls through both entry paths, versus one before. A chain over a source-owned handler runs the source handler once but loses the local half.
1. Prior findings at this head
All 16 previous inline threads have author replies. No acknowledgment replies are needed, and no threads were resolved.
| Previous finding | Disposition |
|---|---|
| Throwing teardown | Fixed by eliminating callbacks during refresh. Existing list, store and binding failures no longer occur. |
| Reentrant teardown | Fixed by the same boundary. Callback mutations cannot invalidate in-flight indices. |
| Source-owned resource teardown | Answered and implemented: zero calls at refresh, one at real removal. Wrapper-local cleanup is the remaining case reported inline. |
| Source-identity regression | Fixed. The current splice/reinsert/readd test fails on the pre-identity-fix code and passes here. |
| In-place reorder coverage | Restored. A dedupe-identical-entries mutation fails this test. |
| Duplicate source Set | Removed; the map now tracks accepted sources. |
| Repeated stale predicate | Removed; returned wrapper identity answers the question. |
| One-caller provenance helper | Inlined and deleted. |
| Full link collector | Simplified to first matching link with the post-hook slot lookup retained. |
| Original/source naming | Corrected. |
| Slot-test cross-product | Reduced. Every retained slot class has a killing mutation, including the input wiring case. |
| Duplicate call-count assertion | Deleted. |
| Duplicate widget inequality | Deleted. |
| Unused fixture overrides | Replaced with a typed callback parameter that three callers use. |
| Store-release coverage | Fixed. Removing releaseRegisteredState now fails the previous-ID assertion. |
| Array narrowing/imports | Fixed in the existing raw-refresh test. Fold the new callback table into the same narrowed setup, as noted inline. |
The previous body items are also mostly settled: the Summary is two sentences; slot-replacing and whole-list mutations are explicitly excluded; no-browser/Agent-E2E wording is honest; the transient unbinding step is gone. The optional split of assertion/accessor changes remains a non-blocking scope suggestion, not a demand to restart the PR. Earlier rounds' addressed findings retain their recorded dispositions.
2. Remaining policy and description points
- P3 question: a renamed-and-retyped source can take a sibling's name because its replacement occupies the earlier slot. Reproduced list:
other:toggle,other#1:number; the previously installed number widget loses its name. Before in-place refresh, the incoming widget was suffixed instead. If first-in-array naming is the intended rule during refresh too, document that consequence or exclude collisions when renaming a live source. This is not a reason to add a second naming system. - P3 clarity: explicitly name the
removeOutputchange in the compatibility notes. Invalid numeric output indices now do nothing, including typed calls; input removal already ignored them. The current phrase "runtime indices" includes these cases, so this is clarification rather than another false behavior claim. - P3: restore "in production" to "instead of reporting and returning undefined." Development access already threw on base; production is where the fail-closed behavior changes.
- P3: delete "this change is covered at the shared LiteGraph compatibility boundary" or scope it to widget/graph changes. The assertion and accessor tests are outside LiteGraph and are included in the ten-suite count.
Do not claim preservation of all wrapper-local state until the teardown case is addressed. Source-owned callbacks surviving refresh does not establish that callbacks attached to the returned wrapper survive it.
3. Checks, cleared candidates, and limitations
- Parent focused Vitest run: 121/121 across four suites. The claimed ten-suite run was independently reproduced: 187/187. Broader LiteGraph: 2,225 passed, 8 expected failures, 5 skipped across 143 files.
- The two original callback-failure paths are fixed under the new no-callback-at-refresh contract. Tests run against the previous production code fail six refresh cases. Old probes expecting teardown specifically during refresh are not treated as requirements.
- Wrapper-local teardown probes observed zero total calls on head and one on prior/base. The independent test-lane probe fails with "expected vi.fn() to be called once, but got 0 times." A scratch callback-transfer experiment preserved the local chain without executing it during refresh; its four-suite control stayed at 121/121. That experiment does not justify blindly overwriting a newly changed source callback.
- The proposed two-row refresh test consolidation preserves the mutation failures for position, binding, source teardown, real removal, store release and raw reconciliation. Each entry path still has a distinct failure it catches. The title-format probe confirms
%jdistinguishes2,"2",""and" ". - Scoped oxlint, ESLint and formatting exited zero; ESLint ignored the two test files. Scratch typechecking reported no errors in the touched tests but had unrelated Vue-resolution environment errors, so it is not a clean application typecheck. Fallow findings were baseline findings. No production comment blocks were added or changed this round.
- Rejected: deleting the installed-map update merely because repeated transfer currently looks idempotent; the map should describe the replacement it accepted. Also rejected routing all of
addCustomWidgetthrough raw array insertion: it changes returned-versus-installed identity while widgets temporarily uses a plain array. The existing separate path has a consumer. - Rejected the proposed deletion of "lifecycle" from the earlier red-before evidence solely because callbacks now stay deferred. Registration release and input binding are lifecycle behavior too; the historical failures were independently demonstrated. No need to rewrite that evidence as if it concerned only callback timing.
- Standard-depth correctness, maintainability, tests, idiomaticity and documentation lanes ran. No browser or Agent E2E was run. CI snapshot: 80 successful, 27 skipped, 18 without a conclusion. No full-build or clean repo-wide typecheck claim.
- No new owner comment exposed a review-skill gap. No gap catch or miss is inferred from silence.
Superseded by 47c07f3: the P2 wrapper-local teardown loss is covered red-before for both entry paths and repaired by rebasing nested callback chains onto the replacement current source callback. All five inline findings are answered and resolved; 188 changed-area tests and the 2,226-test broader control pass, along with pre-commit and pre-push gates. Awaiting a fresh exact-head automated verdict.
There was a problem hiding this comment.
Verdict: changes required. Known teardown chains are preserved, but hand-written chains now double-call an unchanged source teardown and can prevent widget removal.
Reviewed 47c07f3, following round 4. The P2 opaque-chain defect is blocking; the type/interface and cleanup findings are non-blocking.
An extension can capture const original = widget.onRemove and install a function that calls original then its own cleanup. That closure has no WeakMap metadata. When the source callback is unchanged, the fallback now calls it directly and again inside the closure. Both refresh paths reproduce source=2, local=1. With a non-repeatable unsubscribe, the second call throws and removeWidget leaves the widget and its store entry installed. Returning the current callback unchanged when the source callback did not change fixes these probes without running any cleanup during refresh.
1. Previous findings and replies
All five round-4 inline threads have author replies; no acknowledgment replies are needed.
| Finding | Disposition |
|---|---|
| Lost wrapper-local cleanup | Fixed for direct assignment and known chains, including repeated refreshes and changing the source callback. The hand-written-chain case is the new failure above. |
| Six refresh tests | Consolidated to two entry-path rows. The harmful-callback scaffolding remains redundant; inline deletion keeps the same mutation sensitivity. |
| Ambiguous slot-test titles | Fixed with %j; the nine invalid-slot rows are distinguishable. |
| Numeric operation selection | Fixed with named operations. |
| Array representation of one link | Scalar comparison implemented with conditional writes retained. The new two-line repair helper can be inlined without violating the complexity gate. |
The compatibility notes now name invalid numeric output indices, production accessor behavior, the E2E limitation, and the first-in-array collision policy. The r4 naming question is answered, subject to the wording clarification below. Earlier fixed/rejected findings keep their recorded dispositions. No threads have been resolved.
2. State the limits of callback rebasing
- P3: only chains created by
chainCallback/useChainCallbackhave inspectable metadata. When the source callback changes, a hand-written closure can still call its captured old callback. The observed result is old source=1, new source=1, local=1. Narrow "rebases locally attached teardown chains" to known chains and state the opaque fallback policy. Do not add machinery pretending it can rewrite arbitrary JavaScript closures. - P3 question: transferred callbacks now receive the replacement as
this. A callback attached to the old wrapper may read fields that were assigned only there. Core consumers often close over their widget and avoid this issue, but the probe confirms the receiver change. Specify the ownership/receiver contract for local teardown; preserving a function reference is not proof that its resources moved. - P3: clarify that the naming collision occurs when a source is both renamed and type-refreshed. A rename alone leaves the installed wrapper's name unchanged. On refresh, the earlier array entry keeps the name and the later entry gets the suffix, whichever is the pre-existing sibling.
- P3 scope: the optional split suggestion is still open, and production additions are now about 301 lines. The assertion/accessor changes can be split from graph/widget compatibility; the callback utility move is another independently reviewable step. This is a scope recommendation, not an additional correctness blocker.
3. Verification and rejected alternatives
- Focused run:
NODE_OPTIONS=--no-experimental-webstorage pnpm test:unitover widgetsView, LGraphNode, SubgraphIO, widgetMap and useChainCallback: 122/122 passed. The author reports 188 across eleven suites and 2,226 in the broad control; those totals were not independently rerun this round. The changed-suite count is consistent with the previously verified totals. - Repeated refresh, local additions between refreshes, source chaining, same-type re-add and duplicate raw insertion probes pass for known chains. The opaque unchanged-source probe fails with source cleanup called twice. A separate non-idempotent cleanup probe throws
unsubscribe is not a functionand prevents removal. - The proposed unchanged-source fast path makes those probes pass, including opaque replacement/disable cases; the five-suite control remains 122/122. The remaining changed-source opaque probe still invokes its captured old callback, which is the documented limitation requested above, not a claim that this guard solves opaque closure introspection.
- A bare local-cleanup spy retains the red-before failure and the call-during-refresh mutation failures, including swallowed exceptions. Type probes show the new unconstrained rebase API accepts non-functions and can claim a fresh function still has a caller's attached properties. The production caller uses a void teardown, so the type finding is P3 here.
- Simplification experiments: flat-chain reconstruction passed 122 tests with a failing known-chain-disabled control; inlining only mirror repair passed 374 tests and the Fallow gate. Inlining all three SubgraphOutput helpers failed the complexity gate, so I am not asking for that. Flattening must retain callback order, receiver semantics, the opaque fallback and the unchanged-source fast path.
- Rejected replacing the old composable wrapper with another alias re-export: repository guidance forbids new internal barrel re-exports. Migrate internal imports if removing that wrapper. The move into LiteGraph itself is justified because the layer rules prohibit LiteGraph importing a composable.
- Kept the source-callback provenance and registry, which known-chain rebasing needs. Kept the widgetMap equality guard because it avoids writing an unchanged replacement callback. The type finding concerns the newly introduced rebase casts, not the two casts moved verbatim from the old chain helper.
- Standard-depth correctness, maintainability, tests, idiomaticity and documentation lanes ran. Both added/moved production docblocks were inventoried and have deletion comments. Scoped lint/format checks exited zero; ESLint ignores the three test files. Scratch typechecking has unrelated Vue/global resolution errors and is not a clean full typecheck. CI snapshot: 60 successful, 25 skipped, 31 without a conclusion. No browser/Agent E2E or full build was run.
- No new owner comment exposed a review-skill gap. No catch or miss is inferred from silence.
| let refreshComplete = false | ||
| const localCleanup = vi.fn(() => { | ||
| if (refreshComplete) return | ||
| widgets.splice(0, 1) | ||
| throw new Error('extension teardown ran during refresh') | ||
| }) |
There was a problem hiding this comment.
[P3] Nit: Replace this body with a bare vi.fn and remove refreshComplete and its later assignment. not.toHaveBeenCalled already rejects every invocation during refresh. The simpler spy still fails both call-during-refresh mutants, including swallowed exceptions, and the previous production implementation. The throw/splice body adds no unique signal and can obscure the failure with a secondary graph error. This needs the separate assignment deleted too, so no incomplete suggestion.
There was a problem hiding this comment.
Addressed in 527e7cc: the refreshComplete flag and mutation/throw callback were removed. The receiver check now verifies the replacement-widget contract without executing harmful cleanup during refresh.
| } | ||
| } | ||
| ] as const)( | ||
| 'refreshes wrapper state and defers teardown through $entry', |
There was a problem hiding this comment.
[P3] Nit: The rows vary both entry path and attachment form, but only the path is named. Add an attachment label such as direct or known-chain and include it in the title. Mutation checks fail different rows based on attachment form, so that distinction belongs in the failure output. This requires labels on both rows and a title change.
There was a problem hiding this comment.
Fixed in 840dae874e: Added cleanup-attachment labels to the table and title, also addressing the carried round-6 finding.
| widget.onRemove = chainCallback<IBaseWidget, IBaseWidget['onRemove']>( | ||
| widget.onRemove, | ||
| cleanup | ||
| ) |
There was a problem hiding this comment.
[P3] Nit: These explicit type arguments are inferable from the callback and assignment; the plain call was checked in the type probe.
| widget.onRemove = chainCallback<IBaseWidget, IBaseWidget['onRemove']>( | |
| widget.onRemove, | |
| cleanup | |
| ) | |
| widget.onRemove = chainCallback(widget.onRemove, cleanup) |
There was a problem hiding this comment.
Fixed in 840dae874e: Removed the explicit callback generics, also addressing the carried round-6 finding.
Superseded by 527e7cc: the P2 opaque unchanged-source double teardown is covered red-before at the utility and both widget refresh entry paths, then fixed with the unchanged-source fast path. All eight findings are answered and resolved; 192 changed-area tests, the 2,230-test broader control, pre-commit, and pre-push pass. Awaiting a fresh exact-head automated verdict.
There was a problem hiding this comment.
Verdict: ready for human review
The blocking unchanged-source teardown bug is fixed at 527e7cc. Both widget refresh paths now preserve the opaque callback and complete removal. The current six-suite run passes 126/126; the three new opaque regressions fail against the previous head. Remaining inline findings are non-blocking P3 type, test and maintainability cleanup. This is a COMMENT, not an approval.
Non-blocking compatibility and scope notes
- P3: the cleared-handler case is not explicit in compatibility note 2. When a wrapper's
onRemoveisundefined, unchanged-source refresh keeps teardown disabled, but changed-source refresh installs the new source callback. Both paths reproduced that distinction. Add that case alongside opaque closures and direct replacements. - P3: the optional split recommendation remains open at 303 added production lines. Assertion/accessor changes and widget compatibility remain separable. This is not a condition for human review, and there is no recorded author commitment to split.
- The existing
useChainCallbackcompatibility entry point is accepted. The unnecessary re-export is gone and the rebase tests now import their owner. The new private/public composer indirection is a separate inline cleanup suggestion.
Verified behavior and rejected candidates
- The original duplicate-cleanup reproduction now passes for both entry paths, including non-idempotent teardown, repeated refreshes, local overrides and cleared handlers. Three nested known-chain levels, multiple callbacks per level, two rebases and callbacks appended between rebases preserve order and receiver.
- The changed-source opaque limitation remains real. If the new source callback and an opaque local closure both call the old source callback, the old callback can run twice and a non-idempotent cleanup can still abort removal. This follows the policy now disclosed in the PR: run the new callback, then the opaque closure, which may retain old captures. I am not treating it as a fix for arbitrary closures. The proposed shortcut of removing the old source from the new chain is unsafe when the opaque callback does not call it; metadata cannot prove what the closure captures. Human review should accept this documented limit explicitly.
- Rejected the truthy-nonfunction regression candidate. Its probe bypassed
Callablewith an assertion, and the consumer search found no reachable untyped caller or extension export. No new guard is requested for that invalid internal input. - Accepted the exact
chainedrename requested last round. I am not reopening it to rename the pre-existingChainedCallbacktype. - Kept the new integration test's deferral and removal/store assertions. A callback invoked early and then cleared can preserve total call counts while violating deferral. Surviving store-release mutations without identifying the remaining cleanup path do not prove the store assertion redundant. The test simplifications retain these outcomes.
- The callback composer now returns the original callback's result. Traced consumers ignore the result or use void hooks; no production behavior regression was found. The inline type finding concerns an absent original returning undefined. Type probes reject noncallables and strip attached callable properties as intended.
- The type-only test is enforced by the configured TypeScript checks, not Vitest. A wrong return type passes Vitest but fails the scoped type check. No runtime-test claim is made for it.
Checks and coverage
NODE_OPTIONS=--no-experimental-webstorage pnpm test:unitover widgetsView, LGraphNode, SubgraphIO, widgetMap, useChainCallback and chainCallback, with--retry=0: 6 files, 126 passed.- Red-before at 47c07f3: 3 opaque regressions fail with
already unsubscribed; 21 other tests in those two suites pass. - Simplification experiment: 158 passed across seven suites; disabling known-chain handling causes two failures. Fallow reports no new findings across 18 changed files. Scoped TypeScript comparisons have the same unrelated ambient-resolution diagnostics before and after, not a clean full typecheck.
- Type and test mutations verify the return-type, inferred-generics, receiver and duplicate-call findings. The author's 192-test and 2,230-test totals are consistent with the previous verified counts plus this delta, but were not independently rerun in this round. Hook-success claims are author-reported.
- Standard-depth correctness, maintainability, idiomaticity, tests and documentation lanes ran. The delta adds no comments and removes the two previously flagged docblocks. Whole-PR shape checks and dependent/open-PR contract searches found no additional conflict. No browser/Agent E2E, full build or full typecheck was run.
- All existing external-review findings have dispositions in the record below. No new owner comment exposed a review-skill gap. Threads remain unresolved by this reviewer.
Review record
Earlier rounds and external reviewers
A1. DrJKL (human)
| # | Sev | Finding | Source | Final disposition and evidence |
|---|---|---|---|---|
| T0 | q | LGraph.add duplicate guard: throw or only log in production? | r3669742819 (review 4802485428) | superseded: LGraph.ts not in the diff at r1 or head; main owns the guard. Author reply r4239699360, claude[bot] r3669774729 r4052115585. Verified r1 doc lane T0 and r1 body §2. |
| T1 | q | "Calmate": comment tone on duplicate-node check | r3669805286 (review 4802590843) | superseded (moot): file outside the diff. Only claude[bot] replied (r3669811385, tone dialed down); no christian-byrne reply. Verified r1 doc lane T1. Thread resolved in r1 snapshot. The bot author already replied; no duplicate acknowledgment. |
| T2 | q | Dislikes single-function module reportAssertionFailure.ts |
r3669809172 (review 4802597248) | superseded: module merged away (r3670388970, claude[bot]) and later the whole telemetry change left the PR; no such file at r1 or head. No christian-byrne reply. Verified r1 doc lane T2 and r1 body §2. Thread resolved in r1 snapshot. |
| T38 | q | "assert no longer asserts?" (asserts condition → void) |
r4240036962 (review 5481662393) | contested→accepted: author explains runtime unchanged and narrowing unsound (r4242618928); r1 judged void sound, r2 body §3 accepted the explanation, no rename required. Head assert.ts docblock documents it. Thread was unresolved in r1 snapshot; current resolution state is not in REST data. |
| : | : | DrJKL review review 4802604208 (DISMISSED, empty body, no comments) | review 4802604208 | no finding content. |
A2. github-actions[bot] Cursor review panel (35 inline roots, five rounds)
Panel bodies (review 5481239787, review 5481295354, review 5481342515, review 5481384733, review 5481430206) contain only counts; every finding is inline. Round-cap issue comment issuecomment 6104069397 says the last round left nothing open; this record agrees after independent checks. All 35 have christian-byrne replies; replies are not used as proof. Base evidence: r1 doc lane "Prior thread dispositions" table and r1 body §2 (verified at c83741a); later rewrites are cited where they changed the code.
| # | Sev | Finding | Source | Final disposition and evidence |
|---|---|---|---|---|
| T3 | High | removeInput out-of-range now hits assert on an extension hot path | r4239636219 | addressed: quiet no-op (r1 T3); head normalizeLegacySlotIndex → return. |
| T4 | Medium | SubgraphOutput assert runs after beforeChange/topology/disconnect | r4239636222 | addressed then superseded: r1 pre-hook check; report removed in 1f02468 after r2 r4243604608, verified r3 body §1. Head has no assert. |
| T5 | Medium | Interpolated values in new assert messages (removeOutput/removeInput/addCustomWidget…) break grouping | r4239636223 | addressed: no assert added in LGraphNode at head (diff checked); r1 T5. |
| T6 | Medium | layoutStore move/resize assert while create tolerates stale ops | r4239636226 | addressed: missing layouts return false as on base; layoutStore.ts not in head diff (r1 T6). |
| T7 | Low | User-renamable output name in telemetry message | r4239636229 | superseded: no SubgraphOutput telemetry at head (r1 T7; r3 removal). |
| T8 | Low | Duplicate check compares toConcreteWidget result, not passed widget | r4239636231 | addressed: head matches candidate === widget or source provenance === custom_widget (r1 T8; r2/r3 verified identity fixes). |
| T9 | Low | Bounds check accepts fractional index 0.5 | r4239636233 | addressed: Number.isInteger (r1 T9). |
| T10 | Nit | dialogService throws different error type in DEV vs prod | r4239636236 | addressed: single TypeError (r1 T10); its timeout test fixed via r1 r4240113826, verified r2. |
| T11 | High | removeInput assert on hot path (round-2 repeat of T3) | r4239686447 | addressed: as T3. |
| T12 | Medium | SubgraphOutput assert after disconnect/topology (repeat of T4) | r4239686449 | addressed then superseded: as T4. |
| T13 | Medium | layout move/resize assert vs create tolerance (repeat of T6) | r4239686450 | addressed: as T6. |
| T14 | Medium | Interpolated assert messages (repeat of T5) | r4239686452 | addressed: as T5. |
| T15 | Low | this.name in telemetry (repeat of T7) |
r4239686455 | superseded: as T7. |
| T16 | Low | Fractional slot accepted (repeat of T9) | r4239686460 | addressed: as T9. |
| T17 | Low | Duplicate check on toConcreteWidget result (repeat of T8) | r4239686463 | addressed: as T8. |
| T18 | Low | dialogService assert-plus-throw double report in prod | r4239686467 | addressed: dialogService fixed (r1 T18); same pattern in appRegistry fixed after r1 r4240113831 (54227a8), verified r2 body §3. Head: both throw TypeError, no assert. |
| T19 | Nit | agentCrdt test helper continues past missing widgets with void assert | r4239686473 | addressed: helper throws Error (r1 T19; head diff). |
| T20 | High | Rewrite removed !this.widgets.includes(widget) recheck; refused widget still bound |
r4239723953 | addressed: head indexOf(widget) === -1 → return widget after push/replace (r1 T20). |
| T21 | Medium | Guard counts raw serialized linkIds | r4239723958 | addressed then superseded: live-link derivation (r1 T21); cardinality guard deleted r3; head findExistingLink filters live links by target. |
| T22 | Low | candidate === custom_widget misses non-extensible source pushed directly first |
r4239723961 | addressed: provenance lookup; insertion-order residual r1 r4240113813 fixed 760bc98, verified r2. |
| T23 | Low | removeOutput still asserts where removeInput is quiet | r4239723964 | addressed: both quiet (r1 T23; head). |
| T24 | Low | removeWidget never deletes source-map entry; false duplicate later | r4239723966 | superseded: node-local map removed; later flattened provenance WeakMap (r2 r4243604578 → r3), helper inlined (r3 r4243732935 → r4). Duplicate detection reads installed list. r1 T24 notes stale reply text. |
| T25 | Nit | Output name moved into user-renamable context | r4239723970 | superseded: no telemetry context at head (r1 T25). |
| T26 | Medium | existingLink from pre-hook snapshot | r4239765697 | addressed: head re-reads after onConnectOutput (r1 T26). |
| T27 | Medium | getLinks lacks dedupe/target check | r4239765699 | addressed: target filter in findExistingLink, mirror reset to one id (r1 T27; r3 r4243732939 → r4 simplification). |
| T28 | Low | Mirror repair before veto, outside change | r4239765701 | addressed: head repair after hook inside beforeChange (r1 T28). |
| T29 | Low | Source→wrapper cache never invalidated; stale snapshot | r4239765703 | superseded: forward cache removed; refresh keyed by recorded source type (r2 r4243604565 → r3 verified). |
| T30 | Low | Numeric-string slots rejected by Number.isInteger |
r4239765704 | addressed: digit-string conversion; connected-graph test via r1 r4240113851, verified r2. |
| T31 | Nit | assert doc reads as "move user data into context" | r4239765707 | addressed: head docblock forbids user data in message and context; wording follow-up r1 r4240113867 verified r2. |
| T32 | Nit | layoutStore comment points to nonexistent handleCreateGroup | r4239765710 | superseded: comment deleted after r1 r4240113861 (4e6c496, verified r2); file not in head diff. |
| T33 | Medium | Number('')/whitespace/hex/null → 0 |
r4239807033 | addressed: /^\d+$/ gate; tests cover '', ' ', '0x0' (r1 T33; r3/r4 table). |
| T34 | Medium | Dedupe on wrapper source collapses transient reorder duplicate | r4239807038 | addressed: pre-conversion entries (r1 T34); in-place reorder test restored after r3 r4243732922, dedupe-identical mutation fails (r4 body §1). |
| T35 | Low | targetSlot captured before hook | r4239807041 | addressed: head recomputes after hook (r1 T35). |
| T36 | Low | Two distinct live links → permanent refusal; disconnect extras instead | r4239807043 | contested: r2 r4243604608 proved one live link per target through all link-store ingress paths (replaceLink/registerLink/updateEndpoints, r2 correctness C3/F2); recovery deleted 1f02468, verified r3. Residual: the bot's "older saves" premise relies on registerLink validation; no separate load-fixture probe on record. |
| T37 | Low | Cache cleared only in removeWidget/refusal; view-removed wrappers stale | r4239807046 | superseded: committed-set invalidation replaced (it caused r1 r4240113817); provenance per wrapper, verified r2/r3. |
A3. Other external bot output (informational, not findings)
| Source | Content | Disposition |
|---|---|---|
| claude[bot] reviews review 4802540523 review 4802600854 review 4803301260 review 5254421852 | empty bodies; replies on T0-T2 only | no findings. |
| christian-byrne reviews review 4802477934 review 5260114758 (DISMISSED, empty) and reply-container reviews | replies only | no findings. |
| github-actions PR report issuecomment 5110503003 | bundle/perf metrics (+1.16 kB gzip) | informational. |
| codecov issuecomment 5110877347 | patch coverage 91.13%, 11 lines missing (widgetMap 5, LGraphNode 4, widgetsView 1, SubgraphOutput 1) | informational; head commit of report not stated in body. |
| datadog issuecomment 6103782038 | 11 failed jobs at an older head linked in that report | stale; not evidence about this head. |
| coderabbit issuecomment 5739204237 | review skipped (bot user) | informational. |
| christian-byrne issuecomment 6115082010 | author summary of r2 body fixes at 1f02468 | context only; claims verified in r3 body §1. |
B. mrseek-reviews rounds
R1 : review 5481738317 at c83741a (COMMENTED)
Inline (19 roots, all with author replies; dispositions verified in r2 body §3 unless noted):
| # | Sev | Finding | Source | Final disposition and evidence |
|---|---|---|---|---|
| R1-1 | P2 | Dedupe depends on insertion order; raw source before wrapper leaves duplicate | r4240113813 | addressed: 760bc98, verified r2. |
| R1-2 | P2 | Reorder invalidates source identity; re-register duplicates | r4240113817 | addressed: production 0b8433d verified r2; weak test → R2-4 → R3-4 → addressed r4. |
| R1-3 | P2 | Cache ignores writable source type; cast hides mismatch |
r4240113822 | addressed: 4e932f3 verified r2; refresh side effects tracked separately (R2-1..3). |
| R1-4 | P2 | dialogService.extension test dynamic import times out | r4240113826 | addressed: dff6ab8, focused suite passes (r2). |
| R1-5 | P3 | appRegistry repeats assert-plus-throw | r4240113831 | addressed: 54227a8, single TypeError with test (r2). |
| R1-6 | P3 | Both callers repeat bounds checks after parser | r4240113833 | addressed: 7c8f4b4 (r2); helper residual R2-8 → addressed r3. |
| R1-7 | P3 | includes then indexOf double scan | r4240113835 | addressed: 1183a4a (r2). |
| R1-8 | P3 | Missing pre-push positional restoration regression | r4240113839 | addressed: 68f1bf6, mutations fail (r2). |
| R1-9 | P3 | Non-extensible fixture copied six times (LGraphNode tests) | r4240113841 | addressed in that file (16d089a); copies moved → R2-11 → addressed r3. |
| R1-10 | P3 | widgetsView cases live in LGraphNode tests | r4240113845 | addressed: fc5d23a (r2). |
| R1-11a | P3 | Numeric-string boundary absent from signature; ten double casts | r4240113848 | contested: author keeps slot: number (r4243541745); r2 body §2 accepted the signature; casts eliminated via Reflect.apply at 1f02468 (r3 body §1: all eleven casts gone). |
| R1-11b | P3 | Copied input/output tests and not.toThrow wrappers; table them |
r4240113848 | addressed: carried as R2-B2, table at 1f02468 (r3); cross-product trimmed after R3-11 (r4). |
| R1-12 | P3 | Connected-graph numeric-string output removal untested | r4240113851 | addressed: 70984a0, reindex mutation fails (r2). |
| R1-13 | P3 | Non-null assertions on connect results | r4240113853 | addressed: 244f946 (r2). |
| R1-14 | P3 | Missing-node move/resize tests duplicated | r4240113854 | addressed: 3e9a166 (r2). |
| R1-15 | P3 | Narrative test comment in layoutStore.test | r4240113857 | addressed: b8ebca9 (r2). |
| R1-16 | P3 | Only production change in layoutStore is a redundant comment | r4240113861 | addressed: 4e6c496 (r2); file out of diff at head. |
| R1-17 | P3 | Redundant refusal comment before membership guard | r4240113864 | addressed: 8944966 (r2). |
| R1-18 | P3 | One-use dialog error message | r4240113865 | addressed: 03c31a2 (r2). |
| R1-19 | P3 | assert docblock conflates privacy and grouping | r4240113867 | addressed: 6df6c72 (r2); head docblock checked. |
Body-only (r1 body):
| # | Sev | Finding | Source | Final disposition and evidence |
|---|---|---|---|---|
| R1-B1 | P2 | Title/description describe an obsolete diff (main.ts, LGraph.add, reportAssertionFailure, absent asserts, "Breaking: none") | review 5481738317 §1 | addressed: carried R2-B1; rewritten at 1f02468, verified r3 body §1; follow-up wording R3-B1..B4, R4-B2..B5, R5-B1..B3. |
| R1-B2 | P2 | Wrapper identity uses parallel nested WeakMaps, snapshot, three invalidations; resolve identity from the list | review 5481738317 lead | addressed: snapshot removed, installed lookup from list (r2 ledger M5), maps flattened (R2-5 → r3), helper inlined (R3-8 → r4). |
| R1-B3 | P3 q | Multi-live-link recovery reachability unverified | review 5481738317 §3 | addressed: converted to R2-13 deletion finding; recovery removed, verified r3. |
| R1-B4 | P3 q | Filtered link lookup local to connect vs global getLinks contract | review 5481738317 §3.1 | addressed: kept local (author issuecomment 6115082010); verified r3 body §1. |
| R1-B5 | P3 q | Rename report-only assertion? | review 5481738317 §3.2 | contested→accepted: r2 accepted author explanation (r4242618928); no rename required. |
R2 : review 5485392232 at 6df6c72 (now DISMISSED)
Inline (13 roots, all with author replies; dispositions verified in r3 body §1):
| # | Sev | Finding | Source | Final disposition and evidence |
|---|---|---|---|---|
| R2-1 | P2 | Wrapper/source type mismatch misread as source change; wrapper-local hidden state lost | r4243604565 | addressed: recorded source type, probe passes (r3). |
| R2-2 | P2 | Replacement by assignment skips removal lifecycle; input _widget detached |
r4243604567 | addressed for normal callbacks (r3); exception/reentrancy follow-ups R3-1/R3-2 → addressed r4. |
| R2-3 | P2 | Bind loop iterates pre-dedupe array; store holds wrong type | r4243604568 | addressed: final list bound (r3). |
| R2-4 | P2 | Reorder regression test passes on old buggy code | r4243604572 | addressed after R3-4: splice/readd test fails pre-fix (r4 body §1). |
| R2-5 | P3 | Reverse-map node dimension and remember helper survived | r4243604578 | addressed: flattened (r3); helper residual R3-8 → addressed r4. |
| R2-6 | P3 | Redundant candidate === custom_widget arm |
r4243604581 | addressed (r3). |
| R2-7 | P3 | "cached" name for installed lookup | r4243604586 | addressed (r3). |
| R2-8 | P3 | Slot helper repeats bounds branch | r4243604593 | addressed: 27 boundary inputs agree (r3). |
| R2-9 | P3 | Dedupe comment stale | r4243604598 | addressed: deleted (r3). |
| R2-10 | P3 | toBeDefined plus if/throw duplicate precondition |
r4243604600 | addressed: assert.exists (r3). |
| R2-11 | P3 | Fixture copied six times in widgetsView tests | r4243604605 | addressed (r3); unused override residual R3-14 → addressed r4. |
| R2-12 | P3 | Bounds test checks only length; guard removal stays green | r4243604607 | addressed: upper-bound mutation fails (r3); cross-product R3-11 → addressed r4. |
| R2-13 | P3 | Multi-live-link recovery unreachable; remove it | r4243604608 | addressed: removed 1f02468 (r3); collector residual R3-9 → addressed r4. |
Body-only (r2 body):
| # | Sev | Finding | Source | Final disposition and evidence |
|---|---|---|---|---|
| R2-B1 | P2 | Description still stale (carry of R1-B1), plus useApp reporter removal and digit-string policy | review 5485392232 §1 | addressed: r3 body §1. |
| R2-B2 | P3 | Test boundary: copied tests, not.toThrow, casts rose to eleven (carry of R1-11b) |
review 5485392232 §2 | addressed: Reflect.apply table, casts gone (r3). |
| R2-B3 | P3 q | Raw refresh position depends on type change (push→end, unshift→front) | review 5485392232 §4 | addressed: refresh in installed slot (r3 body §1). |
| R2-B4 | q | Filtered-link lookup still unanswered (carry of R1-B4) | review 5485392232 §4 | addressed: as R1-B4. |
R3 : review 5485541105 at 1f02468 (now DISMISSED)
Inline (16 roots, all with author replies; dispositions verified in r4 body §1):
| # | Sev | Finding | Source | Final disposition and evidence |
|---|---|---|---|---|
| R3-1 | P2 | Throwing extension onRemove aborts refresh halfway | r4243732902 | addressed: no callbacks during refresh (r4). |
| R3-2 | P2 | Reentrant onRemove mutates list during index-dependent reconciliation | r4243732903 | addressed: same boundary (r4). |
| R3-3 | P3 q | Which resources does refresh dispose (source-owned onRemove copied)? | r4243732907 | addressed: 0 calls at refresh, 1 at removal (r4); wrapper-local case → R4-1. |
| R3-4 | P2 | Splice/reinsert suggestion applied to the wrong test | r4243732910 | addressed: fails pre-fix, passes head (r4). |
| R3-5 | P3 | Restore in-place reorder in neighboring test | r4243732922 | addressed: dedupe mutation fails (r4). |
| R3-6 | P3 | acceptedRawSources duplicates installedBySource | r4243732927 | addressed (r4). |
| R3-7 | P3 | Stale predicate duplicates toConcreteWidget identity | r4243732930 | addressed (r4). |
| R3-8 | P3 | rememberConcreteWidget one-caller wrapper | r4243732935 | addressed: inlined (r4). |
| R3-9 | P3 | Full link collector when only first link used | r4243732939 | addressed: findExistingLink (r4; head checked). |
| R3-10 | P3 | originalWidgets[index] misnamed source |
r4243732943 | addressed (r4). |
| R3-11 | P3 | 2×9 invalid-slot cross-product | r4243732946 | addressed: each retained class has a killing mutation (r4). |
| R3-12 | P3 | Redundant call-count assertion | r4243732950 | addressed (r4). |
| R3-13 | P3 | Redundant widget inequality | r4243732952 | addressed (r4). |
| R3-14 | P3 | Unused fixture override parameter | r4243732956 | addressed (r4). |
| R3-15 | P3 | Store release not protected by tests | r4243732960 | addressed: removing release fails (r4). |
| R3-16a | P3 | Narrow node.widgets once with assert.exists; drop non-null assertions |
r4243732961 | addressed (r4). |
| R3-16b | P3 | Explicit assert/vi imports for consistency |
r4243732961 | addressed (r4; author r4243761948). |
Body-only (r3 body):
| # | Sev | Finding | Source | Final disposition and evidence |
|---|---|---|---|---|
| R3-B1 | P3 | "No UI change; no browser or Agent behavior changed" is false | review 5485541105 §2 | addressed: honest no-E2E wording (r4 body §1). |
| R3-B2 | P3 | Bound "direct array mutation" claim; slot-replace/splice/whole-list still lose state (pre-existing) | review 5485541105 §2 | addressed: exclusions documented (r4). |
| R3-B3 | P3 | Compat notes: invalid output indices become no-ops; digit strings reach callbacks as numbers | review 5485541105 §2 | addressed: r4 partial, completed r5 (r5 doc lane D-r4-2). |
| R3-B4 | P3 | Summary exceeds template's 1-2 sentences | review 5485541105 §2 | addressed (r4). |
| R3-B5 | P3 | Optional split: assert signature and accessors separate from graph/widget work | review 5485541105 §2 | open, non-blocking: carried R4, R5-B4; r6 maintainability finding 6 (303 prod lines). No owner reply on record. |
| R3-B6 | P3 q | Transient _widget unbinding contract undefined |
review 5485541105 §3 | addressed: direct rebind, transient step removed (r4). |
R4 : review 5485620268 at c64509c (now DISMISSED)
Inline (5 roots, all with author replies; dispositions verified in r5 body §1):
| # | Sev | Finding | Source | Final disposition and evidence |
|---|---|---|---|---|
| R4-1 | P2 | Cleanup attached to returned wrapper lost on refresh | r4243811745 | addressed for direct and known chains (r5); opaque residual R5-1 → addressed r6 (fast path). |
| R4-2a | P3 | Six refresh tests prove two entry paths; collapse to table | r4243811750 | addressed: two-row table (r5). |
| R4-2b | P3 | Use const refreshed instead of let plus not.toThrow |
r4243811750 | addressed in r5 rewrite (r5 tests lane). |
| R4-2c | P3 | Narrow widgets once instead of optional no-op calls | r4243811750 | addressed in r5 rewrite. |
| R4-2d | P3 | Inspect installed slot, not find-BooleanWidget-then-assert | r4243811750 | addressed in r5 rewrite. |
| R4-3 | P3 | $0 titles make 2/'2'/''/' ' indistinguishable |
r4243811751 | addressed: %j (r5). |
| R4-4 | P3 | Operations selected by [1]/[0] | r4243811757 | addressed: named destructuring (r5). |
| R4-5 | P3 | Array representation of at most one link | r4243811760 | addressed: scalar comparison (r5); new helper → R5-5 → addressed r6. |
Body-only (r4 body):
| # | Sev | Finding | Source | Final disposition and evidence |
|---|---|---|---|---|
| R4-B1 | P3 q | Renamed+retyped source takes sibling's name | review 5485620268 §2 | addressed: first-in-array policy documented (r5); trigger clarified after R5-B3 (r6). |
| R4-B2 | P3 | Name the removeOutput behavior change explicitly | review 5485620268 §2 | addressed (r5 doc lane). |
| R4-B3 | P3 | Restore "in production" qualifier for accessor TypeError | review 5485620268 §2 | addressed (r5 doc lane). |
| R4-B4 | P3 | Delete or scope "covered at shared LiteGraph boundary" | review 5485620268 §2 | addressed: scoped to widget/graph (r5 doc lane). |
| R4-B5 | P3 | Do not claim preservation of all wrapper-local state | review 5485620268 §2 | addressed: phrase removed (r5 doc lane). |
| R4-B6 | P3 | Optional split (carry of R3-B5) | review 5485620268 §1 | open, non-blocking: as R3-B5. |
R5 : review 5485704684 at 47c07f3 (now DISMISSED)
Inline (11 roots; 8 have author replies "Addressed in 527e7cc19c"; dispositions from r6 lanes):
| # | Sev | Finding | Source | Final disposition and evidence |
|---|---|---|---|---|
| R5-1 | P2 | Unchanged-source opaque hand chain runs source teardown twice; removal can abort | r4243893941 | addressed: previousOriginal === nextOriginal fast path; r5 probes pass, K1 control fails 3 tests (r6 correctness, tests). Related new r6 candidate R6-C1 (changed source that extends old) is parent-owned. |
| R5-2 | P3 | Unconstrained T: accepts non-callables, promises mock attrs, metadata erased/cast |
r4243893946 | addressed for the posted claims (r6 idiomaticity R-1..R-4); residual return-type over-promise is new r6 I6-1/I6-2, parent-owned. |
| R5-3 | P3 | Recursive callback/found bag rebuilds per level | r4243893948 | addressed: iterative walk (r6 maintainability); residual duplicate exit (maint finding 2) and function expression (I6-6) are new r6 candidates. |
| R5-4a | P3 | Test-only rebaseChainCallback re-export; move tests beside owner |
r4243893950 | addressed: re-export deleted, tests in utils/chainCallback.test.ts (r6 maintainability, idiomaticity R-7). |
| R5-4b | P3 | One-line useChainCallback proxy leaves two names; migrate callers if removing | r4243893950 | retained-intentional: owner keeps it as compatibility surface (r4243915214); the comment's ask was conditional; r6 maintainability notes signature drift only. |
| R5-4c | P3 | Rename opaque test's vague "replacement" label | r4243893950 | addressed: renamed with "local" (r6 tests T3 row). |
| R5-5 | P3 | Two-line repair helper with one caller; inline | r4243893951 | addressed: inlined, Fallow clean (r6 maintainability); method/local name echo is maint finding 5 (below bar). |
| R5-6 | P3 | chainedCallback value vs ChainedCallback type |
r4243893954 | addressed: renamed chained; stem echo moved to type (r6 maint finding 4, new). |
| R5-7 | P3 | Moved chainCallback docblock repeats names | r4243893957 | addressed: deleted (r6 maintainability). |
| R5-8 | P3 | Rebase docblock over-promises for opaque closures | r4243893960 | addressed: deleted; body distinguishes metadata-backed vs opaque (r6 maintainability, documentation). |
| R5-9 | P3 | Harmful-callback body: use bare vi.fn, drop refreshComplete/throw/splice |
r4243893965 | addressed at 527e7cc: spy only records receiver (r6 tests T1 row). Reviewer acknowledged the fix. The new subscribed-flag test cleanup is non-blocking feedback in this review. |
| R5-10 | P3 | Title names entry path but not attachment form | r4243893967 | open: title still through $entry (head widgetsView.test.ts :275-276); no author reply (r6 tests). |
| R5-11 | P3 | Explicit type arguments inferable; suggestion chainCallback(widget.onRemove, cleanup) |
r4243893971 | open: now <IBaseWidget, NonNullable<IBaseWidget['onRemove']>>; plain call compiles (r6 idiomaticity I6-4, type-proofs Q1); no author reply. |
Body-only (r5 body):
| # | Sev | Finding | Source | Final disposition and evidence |
|---|---|---|---|---|
| R5-B1 | P3 | State limits of callback rebasing: only metadata-backed chains; changed-source opaque closure may call stale callback | review 5485704684 §2 | addressed: body N2a/N2b (r6 documentation); behavior unchanged and disclosed (r6 correctness R5-C2 settled). r6 doc lane has a wording candidate for cleared callbacks; parent-owned. |
| R5-B2 | P3 q | Receiver/ownership contract for transferred teardown | review 5485704684 §2 | addressed: body N3a/N3b; test asserts receiver is replacement (r6 correctness, documentation). |
| R5-B3 | P3 | Collision occurs only with rename plus type refresh; describe array position | review 5485704684 §2 | addressed: body N4, probe rows match (r6 documentation). |
| R5-B4 | P3 | Optional split, ~301 prod lines (carry of R3-B5) | review 5485704684 §2 | open, non-blocking: 303 lines at head; split remains optional. |
Round 5 harmful-callback finding 4243893965 is addressed at this head; reviewer acknowledgment.
Round 6 dispositions
-
[P3] Nit: Include undefined in the returned call signature. Open, non-blocking; inline in this review at
src/lib/litegraph/src/utils/chainCallback.ts:16. -
[P3] Nit: The new Callable constraint lets these conditional helpers use the standard utilities. Open, non-blocking; inline in this review at
src/lib/litegraph/src/utils/chainCallback.ts:50. -
[P3] Nit: chainCallback is now a one-call wrapper around chainCallbacks, whose name differs by one letter. Open, non-blocking; inline in this review at
src/lib/litegraph/src/utils/chainCallback.ts:50. -
[P3] Nit: The undefined-link case can take the same fallback as an unknown chain. Open, non-blocking; inline in this review at
src/lib/litegraph/src/utils/chainCallback.ts:82. -
[P3] Nit: The cast that required a function expression is gone. Open, non-blocking; inline in this review at
src/lib/litegraph/src/utils/chainCallback.ts:34. -
[P3] Nit: The method and its result are both named mirrorNeedsRepair. Open, non-blocking; inline in this review at
src/lib/litegraph/src/subgraph/SubgraphOutput.ts:77. -
[P3] Nit: Carried from round 5: the type parameters still infer under the new Callable constraint. Open, non-blocking; inline in this review at
src/lib/litegraph/src/node/widgetsView.test.ts:267. -
[P3] Nit: Carried from round 5: these rows vary cleanup attachment as well as entry path, but the title names only $entry. Open, non-blocking; inline in this review at
src/lib/litegraph/src/node/widgetsView.test.ts:275. -
[P3] Nit: Once identity holds, invoking this function and counting its calls tests the closure constructed in the test. Open, non-blocking; inline in this review at
src/lib/litegraph/src/utils/chainCallback.test.ts:54. -
[P3] Nit: The first refresh table already proves both entry paths reach the shared transfer. Open, non-blocking; inline in this review at
src/lib/litegraph/src/node/widgetsView.test.ts:340. -
[P3] Nit: Vitest already records receivers in localCleanup.mock.contexts. Open, non-blocking; inline in this review at
src/lib/litegraph/src/node/widgetsView.test.ts:290. -
P3 cleared-handler compatibility note: open, non-blocking; described above.
-
P3 optional split: open, non-blocking; described above.
-
Changed-source opaque double-run candidate: retained documented limitation; no unsafe closure-deduplication fix requested.
-
Truthy nonfunction input: rejected, excluded by the callable type and traced consumers.
-
Rename ChainedCallback after the requested result rename: rejected, the requested naming change is complete.
-
Delete deferral/removal/store assertions: rejected; preserve observable boundary outcomes.
-
Generic overload boundary: merged into the absent-original return-type finding; specialization to void remains an alternative.
-
Mutable result/function expression findings from two lanes: merged into one inline suggestion.
-
Throwing cleanup spies: merged into the low-level identity and integration-test simplifications.
Checks at this head: six focused suites, 126/126 passed; three opaque regressions red against the previous head; scoped type proofs and mutation tests run; Fallow no new findings. No full typecheck, full build or browser E2E.
| this: infer O, | ||
| ...args: infer P | ||
| ) => infer R | ||
| ? (this: O, ...args: P) => R |
There was a problem hiding this comment.
[P3] Nit: Include undefined in the returned call signature. With a known or opaque current callback and nextOriginal undefined, the new function returns undefined even when T returns number. The overload currently hides that mismatch. The only production caller uses void, so this is non-blocking. The widened signature adds no diagnostics in the scoped consumer check. Specializing rebasing to the actual void teardown contract would also remove this generic return boundary.
| ? (this: O, ...args: P) => R | |
| ? (this: O, ...args: P) => R | undefined |
There was a problem hiding this comment.
Fixed in 840dae874e: Specialized rebasing to its production void cleanup contract and removed the unchecked generic overload.
| type Args = NonNullable<T> extends (...args: infer P) => unknown ? P : never | ||
| type Ret = NonNullable<T> extends (...args: unknown[]) => infer R ? R : never | ||
|
|
||
| return chainCallbacks(originalCallback, callbacks) as ( | ||
| this: O, | ||
| ...args: Args | ||
| ) => Ret |
There was a problem hiding this comment.
[P3] Nit: The new Callable constraint lets these conditional helpers use the standard utilities. Ret currently becomes never for a function with typed parameters because it checks against unknown[]; that defect is inherited, but this rewritten implementation now returns the original's actual result. The replacement also accounts for an absent original. Scoped consumer checks add no diagnostics.
| type Args = NonNullable<T> extends (...args: infer P) => unknown ? P : never | |
| type Ret = NonNullable<T> extends (...args: unknown[]) => infer R ? R : never | |
| return chainCallbacks(originalCallback, callbacks) as ( | |
| this: O, | |
| ...args: Args | |
| ) => Ret | |
| return chainCallbacks(originalCallback, callbacks) as ( | |
| this: O, | |
| ...args: Parameters<T> | |
| ) => ReturnType<T> | undefined |
There was a problem hiding this comment.
Fixed in 840dae874e: Replaced the conditional return type and cast with inferred argument and result types, including undefined. The return-type regression fails before the fix and passes now.
| export function chainCallback<O, T extends Callable>( | ||
| originalCallback: T | undefined, | ||
| ...callbacks: ChainedCallback<O, T>[] | ||
| ) { | ||
| type Args = NonNullable<T> extends (...args: infer P) => unknown ? P : never | ||
| type Ret = NonNullable<T> extends (...args: unknown[]) => infer R ? R : never | ||
|
|
||
| return chainCallbacks(originalCallback, callbacks) as ( | ||
| this: O, | ||
| ...args: Args | ||
| ) => Ret |
There was a problem hiding this comment.
[P3] Nit: chainCallback is now a one-call wrapper around chainCallbacks, whose name differs by one letter. Fold the private body into the public function and have rebase call chainCallback<unknown, Callable> directly. That removes one indirection without adding a cast; the experiment passed 158 tests and Fallow, with identical scoped TypeScript diagnostics. Apply the return-type correction above while doing it. The body and three separated call sites must change together, so this is a coordinated refactor rather than an isolated suggestion.
There was a problem hiding this comment.
Fixed in 840dae874e: Folded the private composer into chainCallback; rebasing now calls that implementation directly.
| const appendedCallbacks: Callable[] = [] | ||
| let original: Callable | undefined = currentCallback | ||
| while (original !== previousOriginal) { | ||
| if (!original) break | ||
| const chain = callbackChains.get(original) | ||
| if (!chain) { | ||
| return chainCallbacks(nextOriginal, [currentCallback]) | ||
| } | ||
| appendedCallbacks.unshift(...chain.callbacks) | ||
| original = chain.originalCallback | ||
| } | ||
|
|
||
| if (original === previousOriginal) { | ||
| return chainCallbacks(nextOriginal, appendedCallbacks) | ||
| } | ||
|
|
||
| return chainCallbacks(nextOriginal, [currentCallback]) |
There was a problem hiding this comment.
[P3] Nit: The undefined-link case can take the same fallback as an unknown chain. This removes the break, repeated loop-condition check and duplicate fallback. The annotation is needed to avoid TS7022 through the loop assignment. The combined simplification passed 158 tests; disabling known-chain handling still makes two tests fail.
| const appendedCallbacks: Callable[] = [] | |
| let original: Callable | undefined = currentCallback | |
| while (original !== previousOriginal) { | |
| if (!original) break | |
| const chain = callbackChains.get(original) | |
| if (!chain) { | |
| return chainCallbacks(nextOriginal, [currentCallback]) | |
| } | |
| appendedCallbacks.unshift(...chain.callbacks) | |
| original = chain.originalCallback | |
| } | |
| if (original === previousOriginal) { | |
| return chainCallbacks(nextOriginal, appendedCallbacks) | |
| } | |
| return chainCallbacks(nextOriginal, [currentCallback]) | |
| const appendedCallbacks: Callable[] = [] | |
| let link: Callable | undefined = currentCallback | |
| while (link !== previousOriginal) { | |
| const chain: CallbackChain | undefined = link && callbackChains.get(link) | |
| if (!chain) { | |
| return chainCallbacks(nextOriginal, [currentCallback]) | |
| } | |
| appendedCallbacks.unshift(...chain.callbacks) | |
| link = chain.originalCallback | |
| } | |
| return chainCallbacks(nextOriginal, appendedCallbacks) |
There was a problem hiding this comment.
Fixed in 840dae874e: Unified the missing-link and opaque-link fallback and removed the duplicate branch.
| const chained = function (this: unknown, ...args: never[]) { | ||
| let result: unknown | ||
| if (originalCallback) { | ||
| result = Reflect.apply(originalCallback, this, args) | ||
| } | ||
| for (const callback of callbacks) { | ||
| Reflect.apply(callback, this, args) | ||
| } | ||
| return result | ||
| } |
There was a problem hiding this comment.
[P3] Nit: The cast that required a function expression is gone. Use the repository's function-declaration convention and assign result once. This also folds into the wrapper-collapse suggestion.
| const chained = function (this: unknown, ...args: never[]) { | |
| let result: unknown | |
| if (originalCallback) { | |
| result = Reflect.apply(originalCallback, this, args) | |
| } | |
| for (const callback of callbacks) { | |
| Reflect.apply(callback, this, args) | |
| } | |
| return result | |
| } | |
| function chained(this: unknown, ...args: never[]) { | |
| const result = originalCallback | |
| ? Reflect.apply(originalCallback, this, args) | |
| : undefined | |
| for (const callback of callbacks) { | |
| Reflect.apply(callback, this, args) | |
| } | |
| return result | |
| } |
There was a problem hiding this comment.
Fixed in 840dae874e: The callback is now a function declaration with a single const result and inferred return type.
| widget.onRemove = chainCallback< | ||
| IBaseWidget, | ||
| NonNullable<IBaseWidget['onRemove']> | ||
| >(widget.onRemove, cleanup) |
There was a problem hiding this comment.
[P3] Nit: Carried from round 5: the type parameters still infer under the new Callable constraint. NonNullable is needed only because this call supplies them explicitly. The plain call compiles and assigns to widget.onRemove at this head.
| widget.onRemove = chainCallback< | |
| IBaseWidget, | |
| NonNullable<IBaseWidget['onRemove']> | |
| >(widget.onRemove, cleanup) | |
| widget.onRemove = chainCallback(widget.onRemove, cleanup) |
There was a problem hiding this comment.
Fixed in 840dae874e: Removed the explicit callback type parameters; the call now infers them and passes the app typecheck.
| } | ||
| } | ||
| ] as const)( | ||
| 'refreshes wrapper state and defers teardown through $entry', |
There was a problem hiding this comment.
[P3] Nit: Carried from round 5: these rows vary cleanup attachment as well as entry path, but the title names only $entry. Add an attachment label, direct or known-chain, and include it in the title so a failure identifies both varying inputs. The row labels and title need coordinated edits.
There was a problem hiding this comment.
Fixed in 840dae874e: Added direct and known-chain attachment labels and included them in the parameterized test title.
| it('retains an opaque hand-written chain when the original is unchanged', () => { | ||
| let subscribed = true | ||
| const original = vi.fn(() => { | ||
| if (!subscribed) throw new Error('already unsubscribed') | ||
| subscribed = false | ||
| }) | ||
| const local = vi.fn() | ||
| const handWrittenChain = () => { | ||
| original() | ||
| local() | ||
| } | ||
|
|
||
| const rebased = rebaseChainCallback(handWrittenChain, original, original) | ||
| rebased?.() | ||
|
|
||
| expect(rebased).toBe(handWrittenChain) | ||
| expect(original).toHaveBeenCalledOnce() | ||
| expect(local).toHaveBeenCalledOnce() | ||
| }) |
There was a problem hiding this comment.
[P3] Nit: Once identity holds, invoking this function and counting its calls tests the closure constructed in the test. Keep the documented identity contract here; the widget regression checks invocation and removal. This smaller case still fails on the previous head and when the unchanged-source guard is removed.
| it('retains an opaque hand-written chain when the original is unchanged', () => { | |
| let subscribed = true | |
| const original = vi.fn(() => { | |
| if (!subscribed) throw new Error('already unsubscribed') | |
| subscribed = false | |
| }) | |
| const local = vi.fn() | |
| const handWrittenChain = () => { | |
| original() | |
| local() | |
| } | |
| const rebased = rebaseChainCallback(handWrittenChain, original, original) | |
| rebased?.() | |
| expect(rebased).toBe(handWrittenChain) | |
| expect(original).toHaveBeenCalledOnce() | |
| expect(local).toHaveBeenCalledOnce() | |
| }) | |
| it('retains an opaque hand-written chain when the original is unchanged', () => { | |
| const original = vi.fn() | |
| const handWrittenChain = () => original() | |
| expect(rebaseChainCallback(handWrittenChain, original, original)).toBe( | |
| handWrittenChain | |
| ) | |
| }) |
There was a problem hiding this comment.
Fixed in 840dae874e: Reduced the utility regression to the callback-identity contract; integration coverage still verifies deferred cleanup and completed removal.
| it.for([ | ||
| { | ||
| entry: 'addCustomWidget', | ||
| refresh(node: LGraphNode, _widgets: IBaseWidget[], source: IBaseWidget) { | ||
| return node.addCustomWidget(source) | ||
| } | ||
| }, | ||
| { | ||
| entry: 'widgets.unshift', | ||
| refresh(_node: LGraphNode, widgets: IBaseWidget[], source: IBaseWidget) { | ||
| widgets.unshift(source) | ||
| return widgets[0] | ||
| } | ||
| } | ||
| ] as const)( | ||
| 'preserves an opaque teardown chain through $entry when its source callback is unchanged', |
There was a problem hiding this comment.
[P3] Nit: The first refresh table already proves both entry paths reach the shared transfer. This second table can be one addCustomWidget integration case with a bare sourceCleanup spy instead of the subscribed flag and throw. The single-path version still catches each tested missing-transfer, nonidentical-callback, duplicate-call and premature-call mutant; duplicate cleanup then fails with a call-count error rather than a secondary throw. Keep the deferral and completed-removal/store assertions: they state the boundary behavior requested last round. The table, body and spy need a coordinated rewrite, so I am not supplying a partial table deletion.
There was a problem hiding this comment.
Fixed in 840dae874e: Reduced the opaque-chain regression to one addCustomWidget case with a bare spy, retaining deferred cleanup and completed removal/store assertions. Removing the unchanged-source guard still fails this case.
| const cleanupReceivers: IBaseWidget[] = [] | ||
| const localCleanup = vi.fn(function (this: IBaseWidget) { | ||
| cleanupReceivers.push(this) | ||
| }) |
There was a problem hiding this comment.
[P3] Nit: Vitest already records receivers in localCleanup.mock.contexts. Use a bare vi.fn here and replace the later call-count plus cleanupReceivers equality with expect(localCleanup.mock.contexts).toEqual([refreshed]). That equality also proves one call. Both receiver mutants still fail with the native record, and the repo uses mock.contexts elsewhere. Setup and assertions are separated and must change together.
There was a problem hiding this comment.
Fixed in 840dae874e: Replaced manual receiver bookkeeping and its duplicate count assertion with mock.contexts equality. Dropping the callback receiver still fails the regression.
There was a problem hiding this comment.
Verdict: ready for human review
I pushed 840dae874e at the maintainer's request. This is an implementation hand-back, not an independent re-review or approval. All eleven round-6 inline findings and the two carried round-5 threads have fixes and replies. The description now states the cleared-handler policy. The optional split remains unnecessary for this follow-up: the simplification reduces added production lines from 303 to 263 without rewriting the author's commits.
Verification: 159 tests passed across seven focused suites; strict callback return-type checking passed after failing before the fix; removing the unchanged-source guard failed two tests and dropping the appended callback receiver failed four. Both mutations were reverted. Commit/push hooks passed formatting, type-aware lint, app typechecking, Knip and Fallow. Hosted checks remain pending; no browser E2E was run.
The disclosed changed-source opaque-closure limitation remains: arbitrary closures may invoke an old captured cleanup, including one also captured by the new source. Metadata cannot safely deduplicate those captures. No threads were resolved and no approval was submitted.
Review record
Takeover outcomes at 840dae8
-
4243995009: fixed at 840dae8. Specialized rebasing to its production void cleanup contract and removed the unchecked generic overload.
-
4243995015: fixed at 840dae8. Replaced the conditional return type and cast with inferred argument and result types, including undefined. The return-type regression fails before the fix and passes now.
-
4243995018: fixed at 840dae8. Folded the private composer into chainCallback; rebasing now calls that implementation directly.
-
4243995019: fixed at 840dae8. Unified the missing-link and opaque-link fallback and removed the duplicate branch.
-
4243995021: fixed at 840dae8. The callback is now a function declaration with a single const result and inferred return type.
-
4243995028: fixed at 840dae8. Renamed the local decision to repairMirror at its declaration and both uses.
-
4243995031: fixed at 840dae8. Removed the explicit callback type parameters; the call now infers them and passes the app typecheck.
-
4243995033: fixed at 840dae8. Added direct and known-chain attachment labels and included them in the parameterized test title.
-
4243995034: fixed at 840dae8. Reduced the utility regression to the callback-identity contract; integration coverage still verifies deferred cleanup and completed removal.
-
4243995046: fixed at 840dae8. Reduced the opaque-chain regression to one addCustomWidget case with a bare spy, retaining deferred cleanup and completed removal/store assertions. Removing the unchanged-source guard still fails this case.
-
4243995048: fixed at 840dae8. Replaced manual receiver bookkeeping and its duplicate count assertion with mock.contexts equality. Dropping the callback receiver still fails the regression.
-
4243893971: fixed at 840dae8. Removed the explicit callback generics, also addressing the carried round-6 finding.
-
4243893967: fixed at 840dae8. Added cleanup-attachment labels to the table and title, also addressing the carried round-6 finding.
-
Cleared-handler documentation: addressed in the PR description.
-
Optional split: closed for this follow-up after reducing added production code to 263 lines; no history rewrite.
-
Changed-source opaque captures: retained as the disclosed limitation, not claimed fixed.
-
Verification: 159/159 focused tests and enabled hooks passed; hosted checks pending.
Historical record through round 6
The following records the prior head; the takeover outcomes above supersede its remaining open dispositions and check counts.
Earlier rounds and external reviewers
A1. DrJKL (human)
| # | Sev | Finding | Source | Final disposition and evidence |
|---|---|---|---|---|
| T0 | q | LGraph.add duplicate guard: throw or only log in production? | r3669742819 (review 4802485428) | superseded: LGraph.ts not in the diff at r1 or head; main owns the guard. Author reply r4239699360, claude[bot] r3669774729 r4052115585. Verified r1 doc lane T0 and r1 body §2. |
| T1 | q | "Calmate": comment tone on duplicate-node check | r3669805286 (review 4802590843) | superseded (moot): file outside the diff. Only claude[bot] replied (r3669811385, tone dialed down); no christian-byrne reply. Verified r1 doc lane T1. Thread resolved in r1 snapshot. The bot author already replied; no duplicate acknowledgment. |
| T2 | q | Dislikes single-function module reportAssertionFailure.ts |
r3669809172 (review 4802597248) | superseded: module merged away (r3670388970, claude[bot]) and later the whole telemetry change left the PR; no such file at r1 or head. No christian-byrne reply. Verified r1 doc lane T2 and r1 body §2. Thread resolved in r1 snapshot. |
| T38 | q | "assert no longer asserts?" (asserts condition → void) |
r4240036962 (review 5481662393) | contested→accepted: author explains runtime unchanged and narrowing unsound (r4242618928); r1 judged void sound, r2 body §3 accepted the explanation, no rename required. Head assert.ts docblock documents it. Thread was unresolved in r1 snapshot; current resolution state is not in REST data. |
| : | : | DrJKL review review 4802604208 (DISMISSED, empty body, no comments) | review 4802604208 | no finding content. |
A2. github-actions[bot] Cursor review panel (35 inline roots, five rounds)
Panel bodies (review 5481239787, review 5481295354, review 5481342515, review 5481384733, review 5481430206) contain only counts; every finding is inline. Round-cap issue comment issuecomment 6104069397 says the last round left nothing open; this record agrees after independent checks. All 35 have christian-byrne replies; replies are not used as proof. Base evidence: r1 doc lane "Prior thread dispositions" table and r1 body §2 (verified at c83741a); later rewrites are cited where they changed the code.
| # | Sev | Finding | Source | Final disposition and evidence |
|---|---|---|---|---|
| T3 | High | removeInput out-of-range now hits assert on an extension hot path | r4239636219 | addressed: quiet no-op (r1 T3); head normalizeLegacySlotIndex → return. |
| T4 | Medium | SubgraphOutput assert runs after beforeChange/topology/disconnect | r4239636222 | addressed then superseded: r1 pre-hook check; report removed in 1f02468 after r2 r4243604608, verified r3 body §1. Head has no assert. |
| T5 | Medium | Interpolated values in new assert messages (removeOutput/removeInput/addCustomWidget…) break grouping | r4239636223 | addressed: no assert added in LGraphNode at head (diff checked); r1 T5. |
| T6 | Medium | layoutStore move/resize assert while create tolerates stale ops | r4239636226 | addressed: missing layouts return false as on base; layoutStore.ts not in head diff (r1 T6). |
| T7 | Low | User-renamable output name in telemetry message | r4239636229 | superseded: no SubgraphOutput telemetry at head (r1 T7; r3 removal). |
| T8 | Low | Duplicate check compares toConcreteWidget result, not passed widget | r4239636231 | addressed: head matches candidate === widget or source provenance === custom_widget (r1 T8; r2/r3 verified identity fixes). |
| T9 | Low | Bounds check accepts fractional index 0.5 | r4239636233 | addressed: Number.isInteger (r1 T9). |
| T10 | Nit | dialogService throws different error type in DEV vs prod | r4239636236 | addressed: single TypeError (r1 T10); its timeout test fixed via r1 r4240113826, verified r2. |
| T11 | High | removeInput assert on hot path (round-2 repeat of T3) | r4239686447 | addressed: as T3. |
| T12 | Medium | SubgraphOutput assert after disconnect/topology (repeat of T4) | r4239686449 | addressed then superseded: as T4. |
| T13 | Medium | layout move/resize assert vs create tolerance (repeat of T6) | r4239686450 | addressed: as T6. |
| T14 | Medium | Interpolated assert messages (repeat of T5) | r4239686452 | addressed: as T5. |
| T15 | Low | this.name in telemetry (repeat of T7) |
r4239686455 | superseded: as T7. |
| T16 | Low | Fractional slot accepted (repeat of T9) | r4239686460 | addressed: as T9. |
| T17 | Low | Duplicate check on toConcreteWidget result (repeat of T8) | r4239686463 | addressed: as T8. |
| T18 | Low | dialogService assert-plus-throw double report in prod | r4239686467 | addressed: dialogService fixed (r1 T18); same pattern in appRegistry fixed after r1 r4240113831 (54227a8), verified r2 body §3. Head: both throw TypeError, no assert. |
| T19 | Nit | agentCrdt test helper continues past missing widgets with void assert | r4239686473 | addressed: helper throws Error (r1 T19; head diff). |
| T20 | High | Rewrite removed !this.widgets.includes(widget) recheck; refused widget still bound |
r4239723953 | addressed: head indexOf(widget) === -1 → return widget after push/replace (r1 T20). |
| T21 | Medium | Guard counts raw serialized linkIds | r4239723958 | addressed then superseded: live-link derivation (r1 T21); cardinality guard deleted r3; head findExistingLink filters live links by target. |
| T22 | Low | candidate === custom_widget misses non-extensible source pushed directly first |
r4239723961 | addressed: provenance lookup; insertion-order residual r1 r4240113813 fixed 760bc98, verified r2. |
| T23 | Low | removeOutput still asserts where removeInput is quiet | r4239723964 | addressed: both quiet (r1 T23; head). |
| T24 | Low | removeWidget never deletes source-map entry; false duplicate later | r4239723966 | superseded: node-local map removed; later flattened provenance WeakMap (r2 r4243604578 → r3), helper inlined (r3 r4243732935 → r4). Duplicate detection reads installed list. r1 T24 notes stale reply text. |
| T25 | Nit | Output name moved into user-renamable context | r4239723970 | superseded: no telemetry context at head (r1 T25). |
| T26 | Medium | existingLink from pre-hook snapshot | r4239765697 | addressed: head re-reads after onConnectOutput (r1 T26). |
| T27 | Medium | getLinks lacks dedupe/target check | r4239765699 | addressed: target filter in findExistingLink, mirror reset to one id (r1 T27; r3 r4243732939 → r4 simplification). |
| T28 | Low | Mirror repair before veto, outside change | r4239765701 | addressed: head repair after hook inside beforeChange (r1 T28). |
| T29 | Low | Source→wrapper cache never invalidated; stale snapshot | r4239765703 | superseded: forward cache removed; refresh keyed by recorded source type (r2 r4243604565 → r3 verified). |
| T30 | Low | Numeric-string slots rejected by Number.isInteger |
r4239765704 | addressed: digit-string conversion; connected-graph test via r1 r4240113851, verified r2. |
| T31 | Nit | assert doc reads as "move user data into context" | r4239765707 | addressed: head docblock forbids user data in message and context; wording follow-up r1 r4240113867 verified r2. |
| T32 | Nit | layoutStore comment points to nonexistent handleCreateGroup | r4239765710 | superseded: comment deleted after r1 r4240113861 (4e6c496, verified r2); file not in head diff. |
| T33 | Medium | Number('')/whitespace/hex/null → 0 |
r4239807033 | addressed: /^\d+$/ gate; tests cover '', ' ', '0x0' (r1 T33; r3/r4 table). |
| T34 | Medium | Dedupe on wrapper source collapses transient reorder duplicate | r4239807038 | addressed: pre-conversion entries (r1 T34); in-place reorder test restored after r3 r4243732922, dedupe-identical mutation fails (r4 body §1). |
| T35 | Low | targetSlot captured before hook | r4239807041 | addressed: head recomputes after hook (r1 T35). |
| T36 | Low | Two distinct live links → permanent refusal; disconnect extras instead | r4239807043 | contested: r2 r4243604608 proved one live link per target through all link-store ingress paths (replaceLink/registerLink/updateEndpoints, r2 correctness C3/F2); recovery deleted 1f02468, verified r3. Residual: the bot's "older saves" premise relies on registerLink validation; no separate load-fixture probe on record. |
| T37 | Low | Cache cleared only in removeWidget/refusal; view-removed wrappers stale | r4239807046 | superseded: committed-set invalidation replaced (it caused r1 r4240113817); provenance per wrapper, verified r2/r3. |
A3. Other external bot output (informational, not findings)
| Source | Content | Disposition |
|---|---|---|
| claude[bot] reviews review 4802540523 review 4802600854 review 4803301260 review 5254421852 | empty bodies; replies on T0-T2 only | no findings. |
| christian-byrne reviews review 4802477934 review 5260114758 (DISMISSED, empty) and reply-container reviews | replies only | no findings. |
| github-actions PR report issuecomment 5110503003 | bundle/perf metrics (+1.16 kB gzip) | informational. |
| codecov issuecomment 5110877347 | patch coverage 91.13%, 11 lines missing (widgetMap 5, LGraphNode 4, widgetsView 1, SubgraphOutput 1) | informational; head commit of report not stated in body. |
| datadog issuecomment 6103782038 | 11 failed jobs at an older head linked in that report | stale; not evidence about this head. |
| coderabbit issuecomment 5739204237 | review skipped (bot user) | informational. |
| christian-byrne issuecomment 6115082010 | author summary of r2 body fixes at 1f02468 | context only; claims verified in r3 body §1. |
B. mrseek-reviews rounds
R1 : review 5481738317 at c83741a (COMMENTED)
Inline (19 roots, all with author replies; dispositions verified in r2 body §3 unless noted):
| # | Sev | Finding | Source | Final disposition and evidence |
|---|---|---|---|---|
| R1-1 | P2 | Dedupe depends on insertion order; raw source before wrapper leaves duplicate | r4240113813 | addressed: 760bc98, verified r2. |
| R1-2 | P2 | Reorder invalidates source identity; re-register duplicates | r4240113817 | addressed: production 0b8433d verified r2; weak test → R2-4 → R3-4 → addressed r4. |
| R1-3 | P2 | Cache ignores writable source type; cast hides mismatch |
r4240113822 | addressed: 4e932f3 verified r2; refresh side effects tracked separately (R2-1..3). |
| R1-4 | P2 | dialogService.extension test dynamic import times out | r4240113826 | addressed: dff6ab8, focused suite passes (r2). |
| R1-5 | P3 | appRegistry repeats assert-plus-throw | r4240113831 | addressed: 54227a8, single TypeError with test (r2). |
| R1-6 | P3 | Both callers repeat bounds checks after parser | r4240113833 | addressed: 7c8f4b4 (r2); helper residual R2-8 → addressed r3. |
| R1-7 | P3 | includes then indexOf double scan | r4240113835 | addressed: 1183a4a (r2). |
| R1-8 | P3 | Missing pre-push positional restoration regression | r4240113839 | addressed: 68f1bf6, mutations fail (r2). |
| R1-9 | P3 | Non-extensible fixture copied six times (LGraphNode tests) | r4240113841 | addressed in that file (16d089a); copies moved → R2-11 → addressed r3. |
| R1-10 | P3 | widgetsView cases live in LGraphNode tests | r4240113845 | addressed: fc5d23a (r2). |
| R1-11a | P3 | Numeric-string boundary absent from signature; ten double casts | r4240113848 | contested: author keeps slot: number (r4243541745); r2 body §2 accepted the signature; casts eliminated via Reflect.apply at 1f02468 (r3 body §1: all eleven casts gone). |
| R1-11b | P3 | Copied input/output tests and not.toThrow wrappers; table them |
r4240113848 | addressed: carried as R2-B2, table at 1f02468 (r3); cross-product trimmed after R3-11 (r4). |
| R1-12 | P3 | Connected-graph numeric-string output removal untested | r4240113851 | addressed: 70984a0, reindex mutation fails (r2). |
| R1-13 | P3 | Non-null assertions on connect results | r4240113853 | addressed: 244f946 (r2). |
| R1-14 | P3 | Missing-node move/resize tests duplicated | r4240113854 | addressed: 3e9a166 (r2). |
| R1-15 | P3 | Narrative test comment in layoutStore.test | r4240113857 | addressed: b8ebca9 (r2). |
| R1-16 | P3 | Only production change in layoutStore is a redundant comment | r4240113861 | addressed: 4e6c496 (r2); file out of diff at head. |
| R1-17 | P3 | Redundant refusal comment before membership guard | r4240113864 | addressed: 8944966 (r2). |
| R1-18 | P3 | One-use dialog error message | r4240113865 | addressed: 03c31a2 (r2). |
| R1-19 | P3 | assert docblock conflates privacy and grouping | r4240113867 | addressed: 6df6c72 (r2); head docblock checked. |
Body-only (r1 body):
| # | Sev | Finding | Source | Final disposition and evidence |
|---|---|---|---|---|
| R1-B1 | P2 | Title/description describe an obsolete diff (main.ts, LGraph.add, reportAssertionFailure, absent asserts, "Breaking: none") | review 5481738317 §1 | addressed: carried R2-B1; rewritten at 1f02468, verified r3 body §1; follow-up wording R3-B1..B4, R4-B2..B5, R5-B1..B3. |
| R1-B2 | P2 | Wrapper identity uses parallel nested WeakMaps, snapshot, three invalidations; resolve identity from the list | review 5481738317 lead | addressed: snapshot removed, installed lookup from list (r2 ledger M5), maps flattened (R2-5 → r3), helper inlined (R3-8 → r4). |
| R1-B3 | P3 q | Multi-live-link recovery reachability unverified | review 5481738317 §3 | addressed: converted to R2-13 deletion finding; recovery removed, verified r3. |
| R1-B4 | P3 q | Filtered link lookup local to connect vs global getLinks contract | review 5481738317 §3.1 | addressed: kept local (author issuecomment 6115082010); verified r3 body §1. |
| R1-B5 | P3 q | Rename report-only assertion? | review 5481738317 §3.2 | contested→accepted: r2 accepted author explanation (r4242618928); no rename required. |
R2 : review 5485392232 at 6df6c72 (now DISMISSED)
Inline (13 roots, all with author replies; dispositions verified in r3 body §1):
| # | Sev | Finding | Source | Final disposition and evidence |
|---|---|---|---|---|
| R2-1 | P2 | Wrapper/source type mismatch misread as source change; wrapper-local hidden state lost | r4243604565 | addressed: recorded source type, probe passes (r3). |
| R2-2 | P2 | Replacement by assignment skips removal lifecycle; input _widget detached |
r4243604567 | addressed for normal callbacks (r3); exception/reentrancy follow-ups R3-1/R3-2 → addressed r4. |
| R2-3 | P2 | Bind loop iterates pre-dedupe array; store holds wrong type | r4243604568 | addressed: final list bound (r3). |
| R2-4 | P2 | Reorder regression test passes on old buggy code | r4243604572 | addressed after R3-4: splice/readd test fails pre-fix (r4 body §1). |
| R2-5 | P3 | Reverse-map node dimension and remember helper survived | r4243604578 | addressed: flattened (r3); helper residual R3-8 → addressed r4. |
| R2-6 | P3 | Redundant candidate === custom_widget arm |
r4243604581 | addressed (r3). |
| R2-7 | P3 | "cached" name for installed lookup | r4243604586 | addressed (r3). |
| R2-8 | P3 | Slot helper repeats bounds branch | r4243604593 | addressed: 27 boundary inputs agree (r3). |
| R2-9 | P3 | Dedupe comment stale | r4243604598 | addressed: deleted (r3). |
| R2-10 | P3 | toBeDefined plus if/throw duplicate precondition |
r4243604600 | addressed: assert.exists (r3). |
| R2-11 | P3 | Fixture copied six times in widgetsView tests | r4243604605 | addressed (r3); unused override residual R3-14 → addressed r4. |
| R2-12 | P3 | Bounds test checks only length; guard removal stays green | r4243604607 | addressed: upper-bound mutation fails (r3); cross-product R3-11 → addressed r4. |
| R2-13 | P3 | Multi-live-link recovery unreachable; remove it | r4243604608 | addressed: removed 1f02468 (r3); collector residual R3-9 → addressed r4. |
Body-only (r2 body):
| # | Sev | Finding | Source | Final disposition and evidence |
|---|---|---|---|---|
| R2-B1 | P2 | Description still stale (carry of R1-B1), plus useApp reporter removal and digit-string policy | review 5485392232 §1 | addressed: r3 body §1. |
| R2-B2 | P3 | Test boundary: copied tests, not.toThrow, casts rose to eleven (carry of R1-11b) |
review 5485392232 §2 | addressed: Reflect.apply table, casts gone (r3). |
| R2-B3 | P3 q | Raw refresh position depends on type change (push→end, unshift→front) | review 5485392232 §4 | addressed: refresh in installed slot (r3 body §1). |
| R2-B4 | q | Filtered-link lookup still unanswered (carry of R1-B4) | review 5485392232 §4 | addressed: as R1-B4. |
R3 : review 5485541105 at 1f02468 (now DISMISSED)
Inline (16 roots, all with author replies; dispositions verified in r4 body §1):
| # | Sev | Finding | Source | Final disposition and evidence |
|---|---|---|---|---|
| R3-1 | P2 | Throwing extension onRemove aborts refresh halfway | r4243732902 | addressed: no callbacks during refresh (r4). |
| R3-2 | P2 | Reentrant onRemove mutates list during index-dependent reconciliation | r4243732903 | addressed: same boundary (r4). |
| R3-3 | P3 q | Which resources does refresh dispose (source-owned onRemove copied)? | r4243732907 | addressed: 0 calls at refresh, 1 at removal (r4); wrapper-local case → R4-1. |
| R3-4 | P2 | Splice/reinsert suggestion applied to the wrong test | r4243732910 | addressed: fails pre-fix, passes head (r4). |
| R3-5 | P3 | Restore in-place reorder in neighboring test | r4243732922 | addressed: dedupe mutation fails (r4). |
| R3-6 | P3 | acceptedRawSources duplicates installedBySource | r4243732927 | addressed (r4). |
| R3-7 | P3 | Stale predicate duplicates toConcreteWidget identity | r4243732930 | addressed (r4). |
| R3-8 | P3 | rememberConcreteWidget one-caller wrapper | r4243732935 | addressed: inlined (r4). |
| R3-9 | P3 | Full link collector when only first link used | r4243732939 | addressed: findExistingLink (r4; head checked). |
| R3-10 | P3 | originalWidgets[index] misnamed source |
r4243732943 | addressed (r4). |
| R3-11 | P3 | 2×9 invalid-slot cross-product | r4243732946 | addressed: each retained class has a killing mutation (r4). |
| R3-12 | P3 | Redundant call-count assertion | r4243732950 | addressed (r4). |
| R3-13 | P3 | Redundant widget inequality | r4243732952 | addressed (r4). |
| R3-14 | P3 | Unused fixture override parameter | r4243732956 | addressed (r4). |
| R3-15 | P3 | Store release not protected by tests | r4243732960 | addressed: removing release fails (r4). |
| R3-16a | P3 | Narrow node.widgets once with assert.exists; drop non-null assertions |
r4243732961 | addressed (r4). |
| R3-16b | P3 | Explicit assert/vi imports for consistency |
r4243732961 | addressed (r4; author r4243761948). |
Body-only (r3 body):
| # | Sev | Finding | Source | Final disposition and evidence |
|---|---|---|---|---|
| R3-B1 | P3 | "No UI change; no browser or Agent behavior changed" is false | review 5485541105 §2 | addressed: honest no-E2E wording (r4 body §1). |
| R3-B2 | P3 | Bound "direct array mutation" claim; slot-replace/splice/whole-list still lose state (pre-existing) | review 5485541105 §2 | addressed: exclusions documented (r4). |
| R3-B3 | P3 | Compat notes: invalid output indices become no-ops; digit strings reach callbacks as numbers | review 5485541105 §2 | addressed: r4 partial, completed r5 (r5 doc lane D-r4-2). |
| R3-B4 | P3 | Summary exceeds template's 1-2 sentences | review 5485541105 §2 | addressed (r4). |
| R3-B5 | P3 | Optional split: assert signature and accessors separate from graph/widget work | review 5485541105 §2 | open, non-blocking: carried R4, R5-B4; r6 maintainability finding 6 (303 prod lines). No owner reply on record. |
| R3-B6 | P3 q | Transient _widget unbinding contract undefined |
review 5485541105 §3 | addressed: direct rebind, transient step removed (r4). |
R4 : review 5485620268 at c64509c (now DISMISSED)
Inline (5 roots, all with author replies; dispositions verified in r5 body §1):
| # | Sev | Finding | Source | Final disposition and evidence |
|---|---|---|---|---|
| R4-1 | P2 | Cleanup attached to returned wrapper lost on refresh | r4243811745 | addressed for direct and known chains (r5); opaque residual R5-1 → addressed r6 (fast path). |
| R4-2a | P3 | Six refresh tests prove two entry paths; collapse to table | r4243811750 | addressed: two-row table (r5). |
| R4-2b | P3 | Use const refreshed instead of let plus not.toThrow |
r4243811750 | addressed in r5 rewrite (r5 tests lane). |
| R4-2c | P3 | Narrow widgets once instead of optional no-op calls | r4243811750 | addressed in r5 rewrite. |
| R4-2d | P3 | Inspect installed slot, not find-BooleanWidget-then-assert | r4243811750 | addressed in r5 rewrite. |
| R4-3 | P3 | $0 titles make 2/'2'/''/' ' indistinguishable |
r4243811751 | addressed: %j (r5). |
| R4-4 | P3 | Operations selected by [1]/[0] | r4243811757 | addressed: named destructuring (r5). |
| R4-5 | P3 | Array representation of at most one link | r4243811760 | addressed: scalar comparison (r5); new helper → R5-5 → addressed r6. |
Body-only (r4 body):
| # | Sev | Finding | Source | Final disposition and evidence |
|---|---|---|---|---|
| R4-B1 | P3 q | Renamed+retyped source takes sibling's name | review 5485620268 §2 | addressed: first-in-array policy documented (r5); trigger clarified after R5-B3 (r6). |
| R4-B2 | P3 | Name the removeOutput behavior change explicitly | review 5485620268 §2 | addressed (r5 doc lane). |
| R4-B3 | P3 | Restore "in production" qualifier for accessor TypeError | review 5485620268 §2 | addressed (r5 doc lane). |
| R4-B4 | P3 | Delete or scope "covered at shared LiteGraph boundary" | review 5485620268 §2 | addressed: scoped to widget/graph (r5 doc lane). |
| R4-B5 | P3 | Do not claim preservation of all wrapper-local state | review 5485620268 §2 | addressed: phrase removed (r5 doc lane). |
| R4-B6 | P3 | Optional split (carry of R3-B5) | review 5485620268 §1 | open, non-blocking: as R3-B5. |
R5 : review 5485704684 at 47c07f3 (now DISMISSED)
Inline (11 roots; 8 have author replies "Addressed in 527e7cc19c"; dispositions from r6 lanes):
| # | Sev | Finding | Source | Final disposition and evidence |
|---|---|---|---|---|
| R5-1 | P2 | Unchanged-source opaque hand chain runs source teardown twice; removal can abort | r4243893941 | addressed: previousOriginal === nextOriginal fast path; r5 probes pass, K1 control fails 3 tests (r6 correctness, tests). Related new r6 candidate R6-C1 (changed source that extends old) is parent-owned. |
| R5-2 | P3 | Unconstrained T: accepts non-callables, promises mock attrs, metadata erased/cast |
r4243893946 | addressed for the posted claims (r6 idiomaticity R-1..R-4); residual return-type over-promise is new r6 I6-1/I6-2, parent-owned. |
| R5-3 | P3 | Recursive callback/found bag rebuilds per level | r4243893948 | addressed: iterative walk (r6 maintainability); residual duplicate exit (maint finding 2) and function expression (I6-6) are new r6 candidates. |
| R5-4a | P3 | Test-only rebaseChainCallback re-export; move tests beside owner |
r4243893950 | addressed: re-export deleted, tests in utils/chainCallback.test.ts (r6 maintainability, idiomaticity R-7). |
| R5-4b | P3 | One-line useChainCallback proxy leaves two names; migrate callers if removing | r4243893950 | retained-intentional: owner keeps it as compatibility surface (r4243915214); the comment's ask was conditional; r6 maintainability notes signature drift only. |
| R5-4c | P3 | Rename opaque test's vague "replacement" label | r4243893950 | addressed: renamed with "local" (r6 tests T3 row). |
| R5-5 | P3 | Two-line repair helper with one caller; inline | r4243893951 | addressed: inlined, Fallow clean (r6 maintainability); method/local name echo is maint finding 5 (below bar). |
| R5-6 | P3 | chainedCallback value vs ChainedCallback type |
r4243893954 | addressed: renamed chained; stem echo moved to type (r6 maint finding 4, new). |
| R5-7 | P3 | Moved chainCallback docblock repeats names | r4243893957 | addressed: deleted (r6 maintainability). |
| R5-8 | P3 | Rebase docblock over-promises for opaque closures | r4243893960 | addressed: deleted; body distinguishes metadata-backed vs opaque (r6 maintainability, documentation). |
| R5-9 | P3 | Harmful-callback body: use bare vi.fn, drop refreshComplete/throw/splice |
r4243893965 | addressed at 527e7cc: spy only records receiver (r6 tests T1 row). Reviewer acknowledged the fix. The new subscribed-flag test cleanup is non-blocking feedback in this review. |
| R5-10 | P3 | Title names entry path but not attachment form | r4243893967 | open: title still through $entry (head widgetsView.test.ts :275-276); no author reply (r6 tests). |
| R5-11 | P3 | Explicit type arguments inferable; suggestion chainCallback(widget.onRemove, cleanup) |
r4243893971 | open: now <IBaseWidget, NonNullable<IBaseWidget['onRemove']>>; plain call compiles (r6 idiomaticity I6-4, type-proofs Q1); no author reply. |
Body-only (r5 body):
| # | Sev | Finding | Source | Final disposition and evidence |
|---|---|---|---|---|
| R5-B1 | P3 | State limits of callback rebasing: only metadata-backed chains; changed-source opaque closure may call stale callback | review 5485704684 §2 | addressed: body N2a/N2b (r6 documentation); behavior unchanged and disclosed (r6 correctness R5-C2 settled). r6 doc lane has a wording candidate for cleared callbacks; parent-owned. |
| R5-B2 | P3 q | Receiver/ownership contract for transferred teardown | review 5485704684 §2 | addressed: body N3a/N3b; test asserts receiver is replacement (r6 correctness, documentation). |
| R5-B3 | P3 | Collision occurs only with rename plus type refresh; describe array position | review 5485704684 §2 | addressed: body N4, probe rows match (r6 documentation). |
| R5-B4 | P3 | Optional split, ~301 prod lines (carry of R3-B5) | review 5485704684 §2 | open, non-blocking: 303 lines at head; split remains optional. |
Round 5 harmful-callback finding 4243893965 is addressed at this head; reviewer acknowledgment.
Round 6 dispositions
-
[P3] Nit: Include undefined in the returned call signature. Open, non-blocking; inline in this review at
src/lib/litegraph/src/utils/chainCallback.ts:16. -
[P3] Nit: The new Callable constraint lets these conditional helpers use the standard utilities. Open, non-blocking; inline in this review at
src/lib/litegraph/src/utils/chainCallback.ts:50. -
[P3] Nit: chainCallback is now a one-call wrapper around chainCallbacks, whose name differs by one letter. Open, non-blocking; inline in this review at
src/lib/litegraph/src/utils/chainCallback.ts:50. -
[P3] Nit: The undefined-link case can take the same fallback as an unknown chain. Open, non-blocking; inline in this review at
src/lib/litegraph/src/utils/chainCallback.ts:82. -
[P3] Nit: The cast that required a function expression is gone. Open, non-blocking; inline in this review at
src/lib/litegraph/src/utils/chainCallback.ts:34. -
[P3] Nit: The method and its result are both named mirrorNeedsRepair. Open, non-blocking; inline in this review at
src/lib/litegraph/src/subgraph/SubgraphOutput.ts:77. -
[P3] Nit: Carried from round 5: the type parameters still infer under the new Callable constraint. Open, non-blocking; inline in this review at
src/lib/litegraph/src/node/widgetsView.test.ts:267. -
[P3] Nit: Carried from round 5: these rows vary cleanup attachment as well as entry path, but the title names only $entry. Open, non-blocking; inline in this review at
src/lib/litegraph/src/node/widgetsView.test.ts:275. -
[P3] Nit: Once identity holds, invoking this function and counting its calls tests the closure constructed in the test. Open, non-blocking; inline in this review at
src/lib/litegraph/src/utils/chainCallback.test.ts:54. -
[P3] Nit: The first refresh table already proves both entry paths reach the shared transfer. Open, non-blocking; inline in this review at
src/lib/litegraph/src/node/widgetsView.test.ts:340. -
[P3] Nit: Vitest already records receivers in localCleanup.mock.contexts. Open, non-blocking; inline in this review at
src/lib/litegraph/src/node/widgetsView.test.ts:290. -
P3 cleared-handler compatibility note: open, non-blocking; described above.
-
P3 optional split: open, non-blocking; described above.
-
Changed-source opaque double-run candidate: retained documented limitation; no unsafe closure-deduplication fix requested.
-
Truthy nonfunction input: rejected, excluded by the callable type and traced consumers.
-
Rename ChainedCallback after the requested result rename: rejected, the requested naming change is complete.
-
Delete deferral/removal/store assertions: rejected; preserve observable boundary outcomes.
-
Generic overload boundary: merged into the absent-original return-type finding; specialization to void remains an alternative.
-
Mutable result/function expression findings from two lanes: merged into one inline suggestion.
-
Throwing cleanup spies: merged into the low-level identity and integration-test simplifications.
Checks at this head: six focused suites, 126/126 passed; three opaque regressions red against the previous head; scoped type proofs and mutation tests run; Fallow no new findings. No full typecheck, full build or browser E2E.
dante01yoon
left a comment
There was a problem hiding this comment.
Request changes for a stale input locator after a source is renamed and retyped during wrapper refresh. This extends the earlier input-binding discussion: the direct reference is repaired, but name-based consumers and serialization still use the old locator.
Own verification at 840dae8, isolated checkout with its own frozen install:
- PASS: 288 existing tests across 14 changed/adjacent suites; app and browser typechecks; standalone callback type assertions; OSS production build; changed-file formatting, type-aware lint and ESLint. Seven healthy browser cases passed against the production build and a local core backend, including both refresh paths, toggle interaction/serialization/removal, three extension dialogs, and two subgraph operations.
- Regression proof: reverting the four widget/slot/subgraph implementation files to the PR base made 25/105 current tests fail and both wrapper browser cases fail. Restoring the head passed 111 focused tests and the healthy browser cases. Seven targeted mutations independently failed their intended regression assertions; all were reverted.
- FAIL: independent renamed-refresh probes through both entry paths leave promoted-value adoption at false instead of true (2 failures; 2 unchanged-name controls pass). Browser interaction works, but both name lookup and the serialized input locator remain stale. A disposable locator-sync control passes; restoring the head reproduces the failures.
The browser extension fixtures are controlled local data; no live Cloud account or model execution is claimed. The inherited opaque-closure limitation and cleared-handler policy were considered as documented, rather than raised again. No API response shapes or frontend-derived server facts were added by this diff. Website validation and performance benchmarking are not relevant to the changed runtime paths.
Please cover the renamed input locator and promoted-value adoption in the checked-in refresh regression. The bot author cannot receive a normal user assignee; the existing human owners will be preserved when handing this back.
| transferConcreteWidgetOnRemove(outgoing, replacement) | ||
| if (outgoing instanceof BaseWidget) outgoing.releaseRegisteredState() | ||
| for (const input of node.inputs) { | ||
| if (input._widget === outgoing) input._widget = replacement |
There was a problem hiding this comment.
issue: Keep the input widget locator consistent when refresh also renames the source. This fixes _widget, but leaves input.widget.name unchanged. For a non-extensible source changed from wrapped/number to renamed/toggle, both addCustomWidget(source) and raw widgets.unshift(source) install the replacement while getWidgetFromSlot(input) and getSlotFromWidget(replacement) can no longer resolve the binding. inputAsSerialisable also saves the old widget name.
This is a further case beyond the earlier binding fix: the new _widget reference is correct, but adoptPromotedWidgetValue still looks up by name and silently skips copying a promoted user value when unpacking. My independent probes leave the target at false instead of the host's true through both refresh paths; unchanged-name controls pass. Both production-browser cases reproduce the stale locator after a successful toggle interaction. Synchronizing the locator in a disposable control passes, and restoring this head fails again. Could we preserve the complete input locator during transfer and add those behavioral assertions to the refresh regression?
There was a problem hiding this comment.
Fixed in 668baea. transferReplacedWidgetState now carries the whole locator rather than only _widget: it points input.widget.name (and type, where the locator carries one) at the replacement, and it matches inputs bound through the locator as well as by direct reference, reusing the inputBindsWidget predicate the rest of the file already resolves bindings with.
That restores all four name-based consumers you listed. adoptPromotedWidgetValue was the one with a silent cost to the user: it looks the target widget up by name, so it found nothing and returned without copying, leaving the promoted value behind on unpack.
The refresh regression now asserts those behaviors across both entry paths: getWidgetFromSlot, getSlotFromWidget, the serialized locator, and promoted-value adoption. I added a browser case for the persisted half too, because the stale name was being written into the saved workflow, so a reload could not rebind the input. Reverting the fix turns both red and reproduces exactly what you saw.
Could you take another look?
A refresh that re-installs a mutated widget source repaired the input's direct `_widget` reference but left the `widget` locator on the outgoing name. Every name-based consumer resolves through that locator, so `getWidgetFromSlot`, `getSlotFromWidget`, `inputAsSerialisable` and promoted-value adoption all silently unbound while the widget itself kept working on screen — including the serializer, so the saved workflow persisted a binding the next load cannot resolve. Transfer both halves of the binding, and match inputs bound through the locator as well as by direct reference, using the `inputBindsWidget` predicate the rest of the file already resolves bindings with. Covered in the checked-in refresh regression across both entry paths (`addCustomWidget` and `widgets.unshift`) with assertions on the four name-based consumers, plus a browser case asserting the serialized workflow. Both are red on the parent and green here. Addresses #14233 (comment)
This branch changes `assert()` from `asserts condition` to `void`, because outside DEV it returns normally and the narrowing was a lie. Every call site that relied on narrowing therefore needs its own guard, which this branch already did for the sites that existed when it was opened. `main` has since added six more accessors that use `assert()` for narrowing, so merging produced eight type errors in files this branch does not otherwise touch. Apply the same conversion this branch already used: throw a `TypeError` with the identical message, then drop the now-unused import. These errors are not new to this commit. The same eight reproduce on the pre-merge head merged with main, so the branch was already unlandable and CI had not re-run since 2026-10-12 to show it.
There was a problem hiding this comment.
Verdict: request changes
The ordinary rename is fixed, but a rename collision can still direct promoted values to a sibling widget. Widget lifecycle/binding is escalated to the owner after six rounds with new material findings. Separately, the new guards repeat the rejected TypeError workaround, and an overlapping PR still depends on assertion narrowing.
1. Widget binding: escalated to the owner
- C1, P1, reproduced: an earlier sibling keeps
renamed; the refreshed source becomesrenamed#1, but its locator remainsrenamed. Both refresh paths resolve and serialize the sibling. Real promoted-value adoption changes the sibling to true and leaves the intended replacement false. Both ordinary-rename controls pass; both collision cases fail. A test-only final-locator repair restores correct adoption in both cases. - T1, P2, coverage: all new refresh cases supply
_widget. They do not distinguish locator-only matching or protect the optional locator-type update.
These rows go to owner escalation rather than another inline widget-fix cycle. The retread spans r1 C1/M1, r2 C1/C2/M1, r3 C1/C2, r4 C1, r5 C1, and r8 C1/T1. Earlier outcomes are recorded in the previous hand-back.
2. P2: establish handled failure outcomes for the TypeError guards
Confirmed: all eight guards require the broader error-handling contract, regardless of which constructor the lint matcher accepts. This includes app/dialog access, auth/current-user/session/workspace access, workflow runtime access, and dynamic COMBO construction. Five leaf accessors throw before registration without invoking the assertion reporter; the production assertion control reports and returns. The workflow accessor is also called during unload, and dynamic-widget construction already has a no-widget refusal path.
Use typed refusal with handling/reporting at the owning caller, or establish the documented invariant exception with a guaranteed handler. Retain assert(): void; changing the exception constructor or restoring unsound narrowing is insufficient. This remains the owner concern tracked in FE-3388. I have not reproduced missing registration during normal app startup and am not claiming a new authentication race or absence of global error reporting.
3. P2: coordinate the assertion contract with PR #21717
PR #21717, inspected at 61fc997d6c, constructs optional changeTrackerClass immediately after assert(changeTrackerClass, ...) in its workflow runtime factory. A bounded strict TypeScript proof importing this PR's actual assertion fails with TS18048: the constructor is possibly undefined. This was not a full combined-tree build. Coordinate that consumer before both changes merge; preserve the truthful assertion signature and handle the missing constructor explicitly.
Existing findings and author response
Addressed at 668baeaa65: Dante's exact non-colliding rename reproduction in the input-locator thread and change request. Independent controls confirm the author's fix reply for that case. The collision row above is a distinct remaining case, not a claim that the unchanged original reproduction still fails.
The earlier callback/type/test cleanup remains addressed in 840dae874e. Its hand-back review records all earlier findings and replies. No additional unanswered own thread requires an acknowledgment at this head.
The still-unresolved assert return-type discussion remains answered: runtime behavior is unchanged, and void removes an unsound production narrowing guarantee. That conclusion does not approve the replacement exception contracts.
Verification and CI
Reviewed head 31bc0c563d. Hosted unit, E2E, ecosystem, lint, typecheck and Fallow checks passed. The snapshot contains 98 successful checks, 23 skipped checks and one failed check.
The failure is coverage packaging. LCOV reports duplicate inner_clicked function names at different start lines in LGraphCanvas.ts. That source and the coverage workflow are unchanged from the effective base. This is a coverage-tooling failure, not a failing browser assertion; coverage packaging remains red. I did not rerun CI.
The focused widgets-view suite passed all 19 cases. Playwright discovery found both new Chromium cases; I did not launch a backend or execute them locally. Those cases inspect graph serialization in memory, not a save/reload cycle. I retained the browser regressions under the repository's fix-coverage policy rather than requesting their deletion as duplicate unit tests. The inline findings cover their placement and setup, the test assertion primitive, and three implementation comments.
| /** | ||
| * Repoints an input at a widget that replaced the one it was bound to. | ||
| * | ||
| * Both halves of the binding have to move. `_widget` is the direct reference, | ||
| * but `widget` is the locator every name-based consumer resolves through — | ||
| * {@link LGraphNode.getWidgetFromSlot}, {@link LGraphNode.getSlotFromWidget}, | ||
| * `inputAsSerialisable`, and promoted-value adoption on unpack. Leaving the | ||
| * locator on the outgoing name silently unbinds all of them whenever a refresh | ||
| * also renames the source. | ||
| */ |
There was a problem hiding this comment.
[P3] Remove the private helper's change-history docblock. The helper name states its job, and the regression and commit preserve the rename rationale; the consumer list creates another place to maintain it.
| /** | |
| * Repoints an input at a widget that replaced the one it was bound to. | |
| * | |
| * Both halves of the binding have to move. `_widget` is the direct reference, | |
| * but `widget` is the locator every name-based consumer resolves through — | |
| * {@link LGraphNode.getWidgetFromSlot}, {@link LGraphNode.getSlotFromWidget}, | |
| * `inputAsSerialisable`, and promoted-value adoption on unpack. Leaving the | |
| * locator on the outgoing name silently unbinds all of them whenever a refresh | |
| * also renames the source. | |
| */ |
| // Link should not be present, but just in case, disconnect it | ||
| const existingLink = this.getLinks().at(0) | ||
| if (existingLink != null) { | ||
| // Re-read after the extension hook because it may mutate topology. |
There was a problem hiding this comment.
[P3] Remove this implementation narration. The post-hook lookup and the topology-mutation regression already express this ordering.
| // Re-read after the extension hook because it may mutate topology. |
| // Serialized slots may retain duplicate, dangling, or unrelated ids. | ||
| // Repair the mirror only after the hook approves and inside the change. |
There was a problem hiding this comment.
[P3] Keep the repair rationale in the change description and regression cases. repairMirror and the surrounding transaction already express the implementation.
| // Serialized slots may retain duplicate, dangling, or unrelated ids. | |
| // Repair the mirror only after the hook approves and inside the change. |
| @@ -0,0 +1,102 @@ | |||
| import { expect } from '@playwright/test' | |||
There was a problem hiding this comment.
[P3] Move this new spec under browser_tests/tests/nodes/. The browser-test placement convention forbids adding new top-level specs.
| test(`keeps the saved input locator resolvable through ${entry}`, async ({ | ||
| comfyPage | ||
| }) => { | ||
| await comfyPage.nodeOps.clearGraph() |
There was a problem hiding this comment.
[P3] Put the graph reset in this describe's beforeEach; clearing the workflow is setup, not the refresh being tested. Also keep the two-row entry-path data inside this describe instead of the free-standing ENTRY_PATHS constant. The existing nodeOps.clearGraph() helper is the right operation; no new helper is needed.
| if (!widgets) | ||
| throw new Error(`Test node ${node.title} did not register widgets`) |
There was a problem hiding this comment.
[P3] Use Vitest's narrowing assertion here instead of hand-writing its failure branch. Add assert to the existing Vitest import too.
| if (!widgets) | |
| throw new Error(`Test node ${node.title} did not register widgets`) | |
| assert.exists(widgets, `Test node ${node.title} did not register widgets`) |
|
Merged current main 79bffa3487 into this branch. Resolved five import conflicts by keeping main’s graph-model package paths and preserving the PR’s widget logic and tests; also migrated the browser regression’s stale widget-type import. Verification: 115 unit tests passed, both browser cases discovered, focused lint had no errors (existing architecture warnings), and pre-push checks completed. Earlier review findings remain open; no behavior fixes or new review in this conflict-only round. |
dante01yoon
left a comment
There was a problem hiding this comment.
Re-review at 3a787cd, covering 668baea (locator fix), 31bc0c5 (assert call sites from main) and the 3a787cd merge.
My earlier finding is fixed. transferReplacedWidgetState now moves input.widget.name/type along with _widget, and the refresh regression checks getWidgetFromSlot, getSlotFromWidget, the serialized locator and promoted-value adoption on both entry paths. I checked that the test fails without the fix: with the old _widget-only rebind put back in this worktree, both refresh cases fail (expected undefined to be BooleanWidget, 2/19). With the head code, widgetsView.test.ts + LGraphNode.test.ts pass (86/86). Gate 1 is met for this commit.
Blocking: the head does not typecheck. The 3a787cd merge removed import { assert } from '@/base/assert' from comfyWorkflow.ts, but main's reportWorkflowStateAnomaly still calls assert(false, message, {...}) (line 184). CI lint-pr / typecheck fails with TS2554 at 184:26, and typecheck-aux fails with TS2304: Cannot find name 'assert'. Lint, unit, and E2E were skipped behind that failure, so none of them ran on this head.
Remaining binding case (independently reproduced, same as the open bot finding C1): if a sibling already holds the new name, the refreshed widget gets renamed#1, but the locator is set to renamed. On both entry paths, getWidgetFromSlot(input) then returns the sibling, and inputAsSerialisable saves { name: 'renamed' }. Before 668baea the locator dangled. Now it resolves to the wrong widget, so saving and promoted-value adoption target the sibling without any error.
Gates 2/3: no API types or client-derived server facts in this round. The TypeError-guard and #21717 assertion-contract threads are still open with the owner, so I'm not repeating them here.
| import { markRaw } from 'vue' | ||
|
|
||
| import { assert } from '@/base/assert' | ||
| import type { ExecutedWsMessage } from '@/platform/remote/comfyui/execution/types' |
There was a problem hiding this comment.
issue: The merge dropped the assert import here, but reportWorkflowStateAnomaly (line 184, added on main) still calls assert(false, message, {...}). CI typecheck fails (TS2554 at 184:26, TS2304: Cannot find name 'assert' in typecheck-aux), and at runtime the anomaly reporter would throw a ReferenceError instead of reporting. Since that call only reports and doesn't narrow, restoring the import should be enough.
| ): void { | ||
| input._widget = replacement | ||
| if (!input.widget) return | ||
| input.widget.name = replacement.name |
There was a problem hiding this comment.
issue: This copies replacement.name before the duplicate-name suffix is applied. Repro (both addCustomWidget and widgets.unshift): node has a toggle sibling renamed, the sealed source wrapped is renamed to renamed and retyped to toggle. Result: the widgets are ['renamed', 'renamed#1'], the locator is renamed, and getWidgetFromSlot(input) returns the sibling. inputAsSerialisable saves { name: 'renamed' }, so promoted-value adoption and reload bind to the wrong widget. Could the locator be synced from the final installed name (after naming is settled), with a collision case added to the refresh regression?
| /** | ||
| * Repoints an input at a widget that replaced the one it was bound to. | ||
| * | ||
| * Both halves of the binding have to move. `_widget` is the direct reference, | ||
| * but `widget` is the locator every name-based consumer resolves through — | ||
| * {@link LGraphNode.getWidgetFromSlot}, {@link LGraphNode.getSlotFromWidget}, | ||
| * `inputAsSerialisable`, and promoted-value adoption on unpack. Leaving the | ||
| * locator on the outgoing name silently unbinds all of them whenever a refresh | ||
| * also renames the source. | ||
| */ |
There was a problem hiding this comment.
issue: Repo rule: no added code comments other than a 1–2 line JSDoc on exported API. rebindInputToReplacement isn't exported, and this is a 9-line block. The same applies to the new explanatory comments in widgetsView.test.ts (around lines 306–309 and 317–318) and the prose comments in widgetRefreshLocator.spec.ts. Please drop them; the commit message already explains the reasoning.
Summary
Preserves concrete widget identity and framework state across direct array mutation and
addCustomWidgetrefreshes. Refresh transfers registration and input bindings without executing teardown, rebases metadata-backedchainCallback/useChainCallbackcleanup onto the current source callback, and keeps the final widget in its installed slot.Compatibility notes
onRemovestays disabled when the source callback is unchanged. If the source callback changes, refresh installs the new source callback.onRemove, with the replacement asthis. Cleanup that needs old-wrapper-only state must retain that state in its closure.removeOutputindices now do nothing, including typed calls. Runtime digit-only string input/output indices remain accepted and callbacks receive numeric indices; public signatures remain numeric.assertnow returnsvoidbecause production assertions do not narrow. In production, unregistereduseAppanduseExtensionDialogServiceaccess now throwsTypeErrorinstead of reporting and returning undefined.Verification
6df6c72reproduced wrapper-local state loss and direct-refresh lifecycle, binding, ordering, and store failures.1f02468reproduced callback-throw and callback-mutation failures through both refresh entry paths.c64509creproduced lost wrapper-local cleanup for both direct attachment anduseChainCallbackchaining.47c07f3reproduced duplicate opaque source teardown and incomplete removal through both refresh entry paths.typecheck:app.