Repository navigation
test: add failing repros for promoted widget external-link rendering - #18085
claude[bot] wants to merge 2 commits into
Conversation
Wiring a parent-graph node into a subgraph host's promoted input renders wrong in three ways: the linked width row collapses to a bare dot with no label (also on plain nodes), the promoted textarea stays visible because the DOM host widget's connectionSuppressed flag never reaches the widget value store, and after undo/redo the wire end lands 38px below the width slot dot. The Playwright cases are marked test.fail and the Vitest cases it.fails so they stay red on purpose until the fixes land.
|
Important Review skippedBot user detected. To trigger a single review, invoke the ⚙️ Run configurationConfiguration used: Repository: Comfy-Org/ComfyUI_frontend/.coderabbit.yaml Review profile: ASSERTIVE Plan: Advanced Run ID: You can disable this status message by setting the Use the checkbox below for a quick retry:
Comment |
🎨 Storybook: ✅ Built — View Storybook🎭 Playwright: ✅ 2044 passed, 0 failed · 1 flaky📊 Browser Reports
📦 Bundle: 9.92 MB gzip 🔴 +3 BDetailsSummary
Category Glance App Entry Points — 3.77 kB (baseline 3.77 kB) • ⚪ 0 BMain entry bundles and manifests
Status: 1 added / 1 removed Graph Workspace — 1.45 MB (baseline 1.45 MB) • ⚪ 0 BGraph editor runtime, canvas, workflow orchestration
Status: 2 added / 2 removed / 1 unchanged Views & Navigation — 132 kB (baseline 132 kB) • ⚪ 0 BTop-level views, pages, and routed surfaces
Status: 14 added / 14 removed / 4 unchanged Panels & Settings — 637 kB (baseline 637 kB) • ⚪ 0 BConfiguration panels, inspectors, and settings screens
Status: 10 added / 10 removed / 17 unchanged User & Accounts — 37.9 kB (baseline 37.9 kB) • ⚪ 0 BAuthentication, profile, and account management bundles
Status: 6 added / 6 removed / 5 unchanged Editors & Dialogs — 126 kB (baseline 126 kB) • ⚪ 0 BModals, dialogs, drawers, and in-app editors
Status: 7 added / 7 removed / 1 unchanged UI Components — 124 kB (baseline 124 kB) • ⚪ 0 BReusable component library chunks
Status: 7 added / 7 removed / 9 unchanged Data & Services — 3.86 MB (baseline 3.86 MB) • ⚪ 0 BStores, services, APIs, and repositories
Status: 13 added / 13 removed / 3 unchanged Utilities & Hooks — 565 kB (baseline 565 kB) • ⚪ 0 BHelpers, composables, and utility bundles
Status: 19 added / 19 removed / 19 unchanged Vendor & Third-Party — 17.4 MB (baseline 17.4 MB) • ⚪ 0 BExternal libraries and shared vendor chunks Status: 18 unchanged Other — 17.7 MB (baseline 17.7 MB) • 🟢 -22 BBundles that do not match a named category
Status: 73 added / 73 removed / 224 unchanged ⚡ Performance Report
Show regressions
All metrics
Historical variance (last 15 runs)
Trend (last 15 commits on main)
Raw data{
"timestamp": "2026-09-19T03:43:33.822Z",
"gitSha": "a9d86989ed108ef719c7e060f83f67416e16005f",
"branch": "test/promoted-widget-external-link-repro",
"measurements": [
{
"name": "canvas-idle",
"durationMs": 1993.1380000000445,
"styleRecalcs": 10,
"styleRecalcDurationMs": 6.258999999999999,
"layouts": 0,
"layoutDurationMs": 0,
"taskDurationMs": 270.109,
"heapDeltaBytes": 11369968,
"heapUsedBytes": 73277536,
"domNodes": -258,
"jsHeapTotalBytes": 4972544,
"scriptDurationMs": 6.379999999999998,
"eventListeners": -152,
"totalBlockingTimeMs": 0,
"frameDurationMs": 16.666666666666668,
"p95FrameDurationMs": 16.800000000000182
},
{
"name": "canvas-idle",
"durationMs": 1991.6150000000243,
"styleRecalcs": 9,
"styleRecalcDurationMs": 4.903999999999999,
"layouts": 0,
"layoutDurationMs": 0,
"taskDurationMs": 284.36000000000007,
"heapDeltaBytes": -5230176,
"heapUsedBytes": 56866660,
"domNodes": -258,
"jsHeapTotalBytes": 4972544,
"scriptDurationMs": 7.170000000000001,
"eventListeners": -174,
"totalBlockingTimeMs": 0,
"frameDurationMs": 16.66333333333332,
"p95FrameDurationMs": 16.800000000000182
},
{
"name": "canvas-mouse-sweep",
"durationMs": 1811.7389999999887,
"styleRecalcs": 71,
"styleRecalcDurationMs": 75.967,
"layouts": 14,
"layoutDurationMs": 4.859000000000001,
"taskDurationMs": 660.061,
"heapDeltaBytes": -1067504,
"heapUsedBytes": 61036072,
"domNodes": -259,
"jsHeapTotalBytes": 4448256,
"scriptDurationMs": 108.559,
"eventListeners": -174,
"totalBlockingTimeMs": 6,
"frameDurationMs": 16.666666666666668,
"p95FrameDurationMs": 16.800000000000182
},
{
"name": "canvas-mouse-sweep",
"durationMs": 1725.6110000000717,
"styleRecalcs": 73,
"styleRecalcDurationMs": 21.614,
"layouts": 14,
"layoutDurationMs": 4.548,
"taskDurationMs": 540.745,
"heapDeltaBytes": -906200,
"heapUsedBytes": 61292776,
"domNodes": -259,
"jsHeapTotalBytes": 4448256,
"scriptDurationMs": 105.13600000000001,
"eventListeners": -174,
"totalBlockingTimeMs": 0,
"frameDurationMs": 16.666666666666668,
"p95FrameDurationMs": 16.800000000000182
},
{
"name": "canvas-zoom-sweep",
"durationMs": 1708.8449999999966,
"styleRecalcs": 30,
"styleRecalcDurationMs": 11.817,
"layouts": 6,
"layoutDurationMs": 1.138,
"taskDurationMs": 208.571,
"heapDeltaBytes": 2164328,
"heapUsedBytes": 64213116,
"domNodes": 75,
"jsHeapTotalBytes": 4980736,
"scriptDurationMs": 9.272,
"eventListeners": 15,
"totalBlockingTimeMs": 0,
"frameDurationMs": 16.666666666666668,
"p95FrameDurationMs": 16.700000000000728
},
{
"name": "canvas-zoom-sweep",
"durationMs": 1719.208999999978,
"styleRecalcs": 31,
"styleRecalcDurationMs": 11.052000000000001,
"layouts": 6,
"layoutDurationMs": 1.035,
"taskDurationMs": 205.28800000000004,
"heapDeltaBytes": 2290744,
"heapUsedBytes": 63143272,
"domNodes": 77,
"jsHeapTotalBytes": 4980736,
"scriptDurationMs": 9.408,
"eventListeners": 15,
"totalBlockingTimeMs": 0,
"frameDurationMs": 16.66333333333332,
"p95FrameDurationMs": 16.700000000000728
},
{
"name": "dom-widget-clipping",
"durationMs": 503.4059999999272,
"styleRecalcs": 12,
"styleRecalcDurationMs": 7.300999999999998,
"layouts": 0,
"layoutDurationMs": 0,
"taskDurationMs": 301.116,
"heapDeltaBytes": 17594540,
"heapUsedBytes": 79463028,
"domNodes": 20,
"jsHeapTotalBytes": 5767168,
"scriptDurationMs": 108.342,
"eventListeners": 0,
"totalBlockingTimeMs": 0,
"frameDurationMs": 16.666666666666682,
"p95FrameDurationMs": 16.800000000000182
},
{
"name": "dom-widget-clipping",
"durationMs": 415.42600000002494,
"styleRecalcs": 13,
"styleRecalcDurationMs": 8.279000000000002,
"layouts": 0,
"layoutDurationMs": 0,
"taskDurationMs": 289.46500000000003,
"heapDeltaBytes": 17611016,
"heapUsedBytes": 79633844,
"domNodes": 21,
"jsHeapTotalBytes": 5767168,
"scriptDurationMs": 103.30300000000001,
"eventListeners": 0,
"totalBlockingTimeMs": 0,
"frameDurationMs": 16.666666666666668,
"p95FrameDurationMs": 16.700000000000273
},
{
"name": "large-graph-idle",
"durationMs": 2002.4529999999459,
"styleRecalcs": 6,
"styleRecalcDurationMs": 4.165000000000002,
"layouts": 0,
"layoutDurationMs": 0,
"taskDurationMs": 329.488,
"heapDeltaBytes": 19742124,
"heapUsedBytes": 83290784,
"domNodes": -244,
"jsHeapTotalBytes": 0,
"scriptDurationMs": 25.828000000000003,
"eventListeners": -152,
"totalBlockingTimeMs": 0,
"frameDurationMs": 16.666666666666668,
"p95FrameDurationMs": 16.700000000000728
},
{
"name": "large-graph-idle",
"durationMs": 1993.7100000000783,
"styleRecalcs": 8,
"styleRecalcDurationMs": 5.061000000000001,
"layouts": 0,
"layoutDurationMs": 0,
"taskDurationMs": 352.18499999999995,
"heapDeltaBytes": -195264,
"heapUsedBytes": 77722276,
"domNodes": -248,
"jsHeapTotalBytes": -1056768,
"scriptDurationMs": 23.361,
"eventListeners": -174,
"totalBlockingTimeMs": 0,
"frameDurationMs": 16.666666666666668,
"p95FrameDurationMs": 16.700000000000728
},
{
"name": "large-graph-pan",
"durationMs": 2401.8319999998994,
"styleRecalcs": 68,
"styleRecalcDurationMs": 13.994000000000002,
"layouts": 0,
"layoutDurationMs": 0,
"taskDurationMs": 1182.871,
"heapDeltaBytes": -1134648,
"heapUsedBytes": 79703384,
"domNodes": -243,
"jsHeapTotalBytes": 217088,
"scriptDurationMs": 596.237,
"eventListeners": -174,
"totalBlockingTimeMs": 0,
"frameDurationMs": 16.666666666666668,
"p95FrameDurationMs": 16.800000000000182
},
{
"name": "large-graph-pan",
"durationMs": 2135.5310000000145,
"styleRecalcs": 66,
"styleRecalcDurationMs": 11.66,
"layouts": 0,
"layoutDurationMs": 0,
"taskDurationMs": 1053.833,
"heapDeltaBytes": 20099320,
"heapUsedBytes": 84222708,
"domNodes": 10,
"jsHeapTotalBytes": 225280,
"scriptDurationMs": 567.469,
"eventListeners": 0,
"totalBlockingTimeMs": 0,
"frameDurationMs": 16.666666666666668,
"p95FrameDurationMs": 16.800000000000182
},
{
"name": "large-graph-zoom",
"durationMs": 3208.5519999999406,
"styleRecalcs": 63,
"styleRecalcDurationMs": 13.534999999999998,
"layouts": 60,
"layoutDurationMs": 14.921999999999999,
"taskDurationMs": 1283.025,
"heapDeltaBytes": 9811492,
"heapUsedBytes": 76052772,
"domNodes": 8,
"jsHeapTotalBytes": 0,
"scriptDurationMs": 595.6909999999999,
"eventListeners": 0,
"totalBlockingTimeMs": 0,
"frameDurationMs": 16.670000000000012,
"p95FrameDurationMs": 16.799999999999272
},
{
"name": "large-graph-zoom",
"durationMs": 3104.432999999972,
"styleRecalcs": 67,
"styleRecalcDurationMs": 14.983000000000002,
"layouts": 60,
"layoutDurationMs": 13.530999999999999,
"taskDurationMs": 1119.8260000000002,
"heapDeltaBytes": -11608780,
"heapUsedBytes": 68458076,
"domNodes": -245,
"jsHeapTotalBytes": -794624,
"scriptDurationMs": 498.79300000000006,
"eventListeners": -174,
"totalBlockingTimeMs": 0,
"frameDurationMs": 16.670000000000012,
"p95FrameDurationMs": 16.800000000000182
},
{
"name": "legacy-node-drag",
"durationMs": 2279.61700000003,
"styleRecalcs": 45,
"styleRecalcDurationMs": 8.628,
"layouts": 0,
"layoutDurationMs": 0,
"taskDurationMs": 1259.8540000000003,
"heapDeltaBytes": 5202008,
"heapUsedBytes": 71286852,
"domNodes": 10,
"jsHeapTotalBytes": 1048576,
"scriptDurationMs": 633.517,
"eventListeners": 180,
"totalBlockingTimeMs": 0,
"frameDurationMs": 16.666666666666668,
"p95FrameDurationMs": 16.800000000000182
},
{
"name": "legacy-node-drag",
"durationMs": 2244.7650000000294,
"styleRecalcs": 45,
"styleRecalcDurationMs": 7.858999999999998,
"layouts": 0,
"layoutDurationMs": 0,
"taskDurationMs": 1224.85,
"heapDeltaBytes": 15293308,
"heapUsedBytes": 80639876,
"domNodes": 10,
"jsHeapTotalBytes": 1728512,
"scriptDurationMs": 628.9630000000001,
"eventListeners": 180,
"totalBlockingTimeMs": 0,
"frameDurationMs": 16.666666666666668,
"p95FrameDurationMs": 16.800000000000182
},
{
"name": "minimap-idle",
"durationMs": 2020.4760000000306,
"styleRecalcs": 9,
"styleRecalcDurationMs": 6.978,
"layouts": 0,
"layoutDurationMs": 0,
"taskDurationMs": 412.72900000000004,
"heapDeltaBytes": -5456716,
"heapUsedBytes": 81013184,
"domNodes": -243,
"jsHeapTotalBytes": 290816,
"scriptDurationMs": 28.314,
"eventListeners": -174,
"totalBlockingTimeMs": 0,
"frameDurationMs": 16.666666666666668,
"p95FrameDurationMs": 16.700000000000728
},
{
"name": "minimap-idle",
"durationMs": 2008.8190000000168,
"styleRecalcs": 7,
"styleRecalcDurationMs": 5.382,
"layouts": 0,
"layoutDurationMs": 0,
"taskDurationMs": 370.28399999999993,
"heapDeltaBytes": -5256012,
"heapUsedBytes": 81908528,
"domNodes": -246,
"jsHeapTotalBytes": -495616,
"scriptDurationMs": 24.798000000000002,
"eventListeners": -174,
"totalBlockingTimeMs": 0,
"frameDurationMs": 16.66333333333335,
"p95FrameDurationMs": 16.799999999999272
},
{
"name": "subgraph-dom-widget-clipping",
"durationMs": 430.49200000007204,
"styleRecalcs": 47,
"styleRecalcDurationMs": 10.097999999999999,
"layouts": 0,
"layoutDurationMs": 0,
"taskDurationMs": 277.22400000000005,
"heapDeltaBytes": 13784924,
"heapUsedBytes": 76276652,
"domNodes": 20,
"jsHeapTotalBytes": 6029312,
"scriptDurationMs": 99.35199999999999,
"eventListeners": 6,
"totalBlockingTimeMs": 0,
"frameDurationMs": 16.663333333333338,
"p95FrameDurationMs": 16.700000000000273
},
{
"name": "subgraph-dom-widget-clipping",
"durationMs": 425.7119999999759,
"styleRecalcs": 48,
"styleRecalcDurationMs": 9.998,
"layouts": 0,
"layoutDurationMs": 0,
"taskDurationMs": 270.637,
"heapDeltaBytes": 13198220,
"heapUsedBytes": 76468716,
"domNodes": 22,
"jsHeapTotalBytes": 4980736,
"scriptDurationMs": 95.821,
"eventListeners": 6,
"totalBlockingTimeMs": 0,
"frameDurationMs": 16.666666666666668,
"p95FrameDurationMs": 16.800000000000182
},
{
"name": "subgraph-idle",
"durationMs": 1988.8290000000097,
"styleRecalcs": 11,
"styleRecalcDurationMs": 6.775999999999999,
"layouts": 0,
"layoutDurationMs": 0,
"taskDurationMs": 284.15100000000007,
"heapDeltaBytes": -922920,
"heapUsedBytes": 62334196,
"domNodes": -272,
"jsHeapTotalBytes": 4710400,
"scriptDurationMs": 5.921000000000001,
"eventListeners": -220,
"totalBlockingTimeMs": 0,
"frameDurationMs": 16.666666666666668,
"p95FrameDurationMs": 16.800000000000182
},
{
"name": "subgraph-idle",
"durationMs": 1993.2179999999562,
"styleRecalcs": 10,
"styleRecalcDurationMs": 7.019000000000001,
"layouts": 0,
"layoutDurationMs": 0,
"taskDurationMs": 298.68300000000005,
"heapDeltaBytes": 17530168,
"heapUsedBytes": 80684204,
"domNodes": -273,
"jsHeapTotalBytes": 4972544,
"scriptDurationMs": 6.969000000000001,
"eventListeners": -198,
"totalBlockingTimeMs": 0,
"frameDurationMs": 16.670000000000012,
"p95FrameDurationMs": 16.700000000000728
},
{
"name": "subgraph-mouse-sweep",
"durationMs": 1726.8909999999096,
"styleRecalcs": 78,
"styleRecalcDurationMs": 26.884999999999998,
"layouts": 18,
"layoutDurationMs": 6.101999999999999,
"taskDurationMs": 593.8050000000001,
"heapDeltaBytes": -2958304,
"heapUsedBytes": 59930508,
"domNodes": -275,
"jsHeapTotalBytes": 4186112,
"scriptDurationMs": 94.037,
"eventListeners": -220,
"totalBlockingTimeMs": 0,
"frameDurationMs": 16.666666666666668,
"p95FrameDurationMs": 16.800000000000182
},
{
"name": "subgraph-mouse-sweep",
"durationMs": 1675.7330000000366,
"styleRecalcs": 77,
"styleRecalcDurationMs": 23.134999999999994,
"layouts": 18,
"layoutDurationMs": 5.464999999999999,
"taskDurationMs": 518.194,
"heapDeltaBytes": 14844328,
"heapUsedBytes": 78124976,
"domNodes": -280,
"jsHeapTotalBytes": 5234688,
"scriptDurationMs": 76.39199999999998,
"eventListeners": -208,
"totalBlockingTimeMs": 0,
"frameDurationMs": 16.666666666666668,
"p95FrameDurationMs": 16.700000000000728
},
{
"name": "subgraph-transition-enter",
"durationMs": 783.9629999999715,
"styleRecalcs": 18,
"styleRecalcDurationMs": 18.280000000000005,
"layouts": 5,
"layoutDurationMs": 7.6579999999999995,
"taskDurationMs": 535.4279999999999,
"heapDeltaBytes": -6429500,
"heapUsedBytes": 80598780,
"domNodes": 13587,
"jsHeapTotalBytes": 10747904,
"scriptDurationMs": 18.318999999999996,
"eventListeners": 2383,
"totalBlockingTimeMs": 65,
"frameDurationMs": 16.666666666666668,
"p95FrameDurationMs": 16.699999999999818
},
{
"name": "viewport-pan-sweep",
"durationMs": 8330.050000000028,
"styleRecalcs": 249,
"styleRecalcDurationMs": 41.660000000000004,
"layouts": 0,
"layoutDurationMs": 0,
"taskDurationMs": 3851.896,
"heapDeltaBytes": -3608064,
"heapUsedBytes": 74857320,
"domNodes": -239,
"jsHeapTotalBytes": -307200,
"scriptDurationMs": 1889.7440000000001,
"eventListeners": -152,
"totalBlockingTimeMs": 0,
"frameDurationMs": 16.666666666666668,
"p95FrameDurationMs": 16.700000000000728
},
{
"name": "viewport-pan-sweep",
"durationMs": 8137.751999999978,
"styleRecalcs": 249,
"styleRecalcDurationMs": 40.237,
"layouts": 0,
"layoutDurationMs": 0,
"taskDurationMs": 3706.042,
"heapDeltaBytes": -12781196,
"heapUsedBytes": 65137540,
"domNodes": -248,
"jsHeapTotalBytes": 1564672,
"scriptDurationMs": 1827.582,
"eventListeners": -174,
"totalBlockingTimeMs": 0,
"frameDurationMs": 16.666666666666668,
"p95FrameDurationMs": 16.799999999999272
},
{
"name": "vue-large-graph-idle",
"durationMs": 11895.141999999963,
"styleRecalcs": 0,
"styleRecalcDurationMs": 0,
"layouts": 0,
"layoutDurationMs": 0,
"taskDurationMs": 11620.986,
"heapDeltaBytes": -34866108,
"heapUsedBytes": 195535644,
"domNodes": -8062,
"jsHeapTotalBytes": -6610944,
"scriptDurationMs": 169.763,
"eventListeners": -14928,
"totalBlockingTimeMs": 0,
"frameDurationMs": 17.219999999999953,
"p95FrameDurationMs": 16.700000000000728
},
{
"name": "vue-large-graph-idle",
"durationMs": 9169.828999999936,
"styleRecalcs": 0,
"styleRecalcDurationMs": 0,
"layouts": 0,
"layoutDurationMs": 0,
"taskDurationMs": 8569.236,
"heapDeltaBytes": -61222768,
"heapUsedBytes": 170465328,
"domNodes": -8062,
"jsHeapTotalBytes": -13688832,
"scriptDurationMs": 130.27300000000002,
"eventListeners": -14928,
"totalBlockingTimeMs": 0,
"frameDurationMs": 16.666666666666668,
"p95FrameDurationMs": 16.700000000000728
},
{
"name": "vue-large-graph-pan",
"durationMs": 14300.15999999989,
"styleRecalcs": 70,
"styleRecalcDurationMs": 15.354999999999979,
"layouts": 0,
"layoutDurationMs": 0,
"taskDurationMs": 14128.934000000001,
"heapDeltaBytes": -36004936,
"heapUsedBytes": 183763756,
"domNodes": -8061,
"jsHeapTotalBytes": -13111296,
"scriptDurationMs": 621.39,
"eventListeners": -14928,
"totalBlockingTimeMs": 4,
"frameDurationMs": 16.666666666666668,
"p95FrameDurationMs": 16.700000000000728
},
{
"name": "vue-large-graph-pan",
"durationMs": 14524.069000000054,
"styleRecalcs": 71,
"styleRecalcDurationMs": 14.17199999999999,
"layouts": 0,
"layoutDurationMs": 0,
"taskDurationMs": 14247.111999999997,
"heapDeltaBytes": -30648200,
"heapUsedBytes": 187227092,
"domNodes": -8061,
"jsHeapTotalBytes": -18337792,
"scriptDurationMs": 649.775,
"eventListeners": -14928,
"totalBlockingTimeMs": 51,
"frameDurationMs": 16.66333333333326,
"p95FrameDurationMs": 16.700000000000728
},
{
"name": "vue-node-resize-workload",
"durationMs": 4334.062999999901,
"styleRecalcs": 20,
"styleRecalcDurationMs": 53.15999999999998,
"layouts": 20,
"layoutDurationMs": 22.508,
"taskDurationMs": 4273.737000000001,
"heapDeltaBytes": -35105148,
"heapUsedBytes": 195500236,
"domNodes": -8061,
"jsHeapTotalBytes": 19865600,
"scriptDurationMs": 69.92300000000002,
"eventListeners": -14928,
"totalBlockingTimeMs": 87,
"frameDurationMs": 16.670000000000012,
"p95FrameDurationMs": 16.80000000000109
},
{
"name": "vue-node-resize-workload",
"durationMs": 4182.226000000014,
"styleRecalcs": 20,
"styleRecalcDurationMs": 52.207,
"layouts": 20,
"layoutDurationMs": 21.722000000000005,
"taskDurationMs": 4138.782000000001,
"heapDeltaBytes": -20139676,
"heapUsedBytes": 197619864,
"domNodes": -8062,
"jsHeapTotalBytes": 21372928,
"scriptDurationMs": 63.33,
"eventListeners": -14928,
"totalBlockingTimeMs": 14,
"frameDurationMs": 17.220000000000013,
"p95FrameDurationMs": 16.80000000000109
},
{
"name": "vue-zoom-out-idle",
"durationMs": 5887.363999999934,
"styleRecalcs": 0,
"styleRecalcDurationMs": 0,
"layouts": 0,
"layoutDurationMs": 0,
"taskDurationMs": 5508.224999999999,
"heapDeltaBytes": -15778204,
"heapUsedBytes": 174143884,
"domNodes": 0,
"jsHeapTotalBytes": -28835840,
"scriptDurationMs": 74.14300000000001,
"eventListeners": 0,
"totalBlockingTimeMs": 0,
"frameDurationMs": 17.219999999999953,
"p95FrameDurationMs": 16.799999999999272
},
{
"name": "vue-zoom-out-idle",
"durationMs": 6036.236000000031,
"styleRecalcs": 0,
"styleRecalcDurationMs": 0,
"layouts": 0,
"layoutDurationMs": 0,
"taskDurationMs": 5822.343,
"heapDeltaBytes": 18716492,
"heapUsedBytes": 197509704,
"domNodes": 0,
"jsHeapTotalBytes": -17563648,
"scriptDurationMs": 81.16099999999999,
"eventListeners": 0,
"totalBlockingTimeMs": 0,
"frameDurationMs": 16.666666666666668,
"p95FrameDurationMs": 16.700000000000728
},
{
"name": "workflow-execution",
"durationMs": 422.4739999999656,
"styleRecalcs": 17,
"styleRecalcDurationMs": 20.237,
"layouts": 4,
"layoutDurationMs": 1.701,
"taskDurationMs": 94.714,
"heapDeltaBytes": 5168540,
"heapUsedBytes": 66844820,
"domNodes": 151,
"jsHeapTotalBytes": 262144,
"scriptDurationMs": 10.372,
"eventListeners": 81,
"totalBlockingTimeMs": 0,
"frameDurationMs": 16.663333333333338,
"p95FrameDurationMs": 16.799999999999727
},
{
"name": "workflow-execution",
"durationMs": 418.3349999999564,
"styleRecalcs": 15,
"styleRecalcDurationMs": 17.61,
"layouts": 4,
"layoutDurationMs": 1.6129999999999998,
"taskDurationMs": 79.667,
"heapDeltaBytes": 4751196,
"heapUsedBytes": 66454736,
"domNodes": 120,
"jsHeapTotalBytes": 262144,
"scriptDurationMs": 7.623000000000001,
"eventListeners": 65,
"totalBlockingTimeMs": 0,
"frameDurationMs": 16.666666666666668,
"p95FrameDurationMs": 16.800000000000182
}
]
}🔬 E2E Coverage
Top 10 files by uncovered lines
|
Playwright videos for new testsDownload the recorded videos and HTML report Recorded projects: chromium Recorded test titles:
|
Codecov Report✅ All modified and coverable lines are covered by tests. @@ Coverage Diff @@
## main #18085 +/- ##
==========================================
+ Coverage 82.77% 83.80% +1.03%
==========================================
Files 2936 2242 -694
Lines 144439 127635 -16804
Branches 39004 32559 -6445
==========================================
- Hits 119562 106967 -12595
+ Misses 23620 20378 -3242
+ Partials 1257 290 -967
Flags with carried forward coverage won't be shown. Click here to find out more. 🚀 New features to boost your workflow:
|
Causal isolation (local, no fixes pushed)Each mechanism below was isolated by instrumenting the failing repro, then finding the smallest change that flips it to passing and confirming it flips back when reverted. 1. Wire drifts 38 px after undo/redo (
2. Promoted textarea stays visible after a String is wired in
3. Width label disappears (
Spec issue in this PR: the drift test clicks Two change-tracker bugs seen along the way (not fixed here): a Vue slot drag-connect is not captured until the next mouseup because Generated by Claude Code |
…rift repro Clicking the same empty canvas pixel before undo and again before redo lands inside the double-click window, which opens node search and swallows Ctrl+Y, so the redo never ran and the drift assertion was never reached. Clicking a different empty pixel lets redo run; with the test.fail() marker stripped the test now fails on the drift assertion (Expected <= 2, Received 38) in 3/3 local runs.
|
CI: Failing step: Why this is not the PR's failure:
Generated by Claude Code |
Undo and redo reload the graph, which drops every DOM-measured slot offset while the Vue node stays mounted with identical rows. Litegraph then falls back to its own widget layout for the link endpoint, so the wire lands on the wrong row: 38px below the width dot on a promoted subgraph host, 10px on a plain EmptyLatentImage. Ports the repro spec from #18085 with the drift case as a real table-driven regression test covering a plain node too, and adds a NodeSlots case for re-measuring after the layout store clears the graph.
|
Closing as superseded: the fixes landed in #18110, #18117 and #18118 (with the WidgetGrid label fix in #17576), and their regression tests replace the expected-failure repros here. Branch left in place. Generated by Claude Code |
…Comfy-Org#18117) <!-- ccr-slack-attribution --> _Requested by **Christian Byrne** · [Slack thread](https://comfy-organization.slack.com/archives/C0BGH348Z0C/p1789592320785689?thread_ts=1789592320.785689&cid=C0BGH348Z0C)_ ## ELI5 Dragging a wire out of a slot on a Vue node and dropping it on another slot connects the two nodes, but the undo history never hears about it. Press Ctrl+Z straight after the drop and nothing happens; you first have to click somewhere on the canvas, and that click is what finally writes the undo entry. This PR writes the entry the moment the drop finishes, the same way the classic canvas already does. ## Motivation Before: connect two Vue node slots by dragging, press Ctrl+Z: the link stays. Only a later canvas click, key press or widget edit records the connection, so undo looks like it needs two steps or does nothing. After: Ctrl+Z right after the drop removes the link. Why: the change tracker records canvas edits from the window `mouseup` event and from `LGraphCanvas.processMouseUp`. The Vue slot drag calls `preventDefault()` on `pointerdown`, which suppresses the compatibility `mouseup`, and the drop is resolved entirely inside the Vue composable, so neither hook runs and `captureCanvasState()` is never called. ## Provenance - **Authored by:** interactive session - **Verified:** Playwright `linkInteraction.spec.ts`, new case `undo right after dropping a link on a slot removes that link`: fails on the test-only commit (`expected 0, received 1` links after undo) and passes with the fix. The rest of the spec has the same outcome with and without the fix (9 cases fail in the local sandbox either way: 8 `toHaveScreenshot` baselines and one minimap overlay intercepting the subgraph button; a control run with the fix stashed reproduced all 9). A local probe confirmed the undo queue holds exactly one entry right after the drop. Root and `browser_tests` typecheck, full `pnpm lint`, `pnpm format:check`, eslint, oxlint and oxfmt on the changed files: clean. - **Deviations:** none ## Reviewer context - **Type:** bug fix, undo/redo for Vue node link drags - **Slots into:** Comfy-Org#18085 (repro and isolation PR). The "undo needs an extra click" symptom surfaced while isolating the link drift there. The spurious undo entry after a redo is a separate tracker bug, fixed in Comfy-Org#18118. - No unit test: the composable's pointer flow has no seam short of the ~200-line module-mock harness in `useSlotLinkInteraction.autoPan.test.ts`, so the e2e is the lowest level that proves the wiring. ## How `useSlotLinkInteraction.finishInteraction()` calls `captureCanvasState()` on the active workflow's change tracker after the drop is resolved and the canvas marked dirty, mirroring the `LGraphCanvas.processMouseUp` wrapper the tracker installs for the legacy canvas. `captureCanvasState()` compares against the last snapshot, so a drop that connects nothing pushes no entry, a second capture from any other hook is a no-op, and the legacy path is untouched. ## Red-green verification | Commit | CI | Purpose | |---|---|---| | `test: cover undo right after a Vue node slot drag-connect` (d8f3af7) | red | the new Playwright case fails: the link survives Ctrl+Z (e2e-status failed on shard 14 and the new-tests video job, everything else green) | | `fix: capture the canvas state when a Vue node slot link drag finishes` (2f9f6df) | green | same case passes, all checks green | ## Test plan - [x] CI red on the test-only commit (e2e-status failed on shard 14 and the new-tests video job) - [x] CI green on the fix commit - [ ] Drag a link between two Vue node slots, press Ctrl+Z immediately: the link is removed - [ ] Drag a link out of a slot and drop it on empty canvas, dismiss the search box: the undo queue does not grow --- _Generated by [Claude Code](https://claude.ai/code/session_01JwD8PkozfznCoUwAG46WAg)_ --------- Co-authored-by: Claude <noreply@anthropic.com>
Comfy-Org#18118) <!-- ccr-slack-attribution --> _Requested by **Christian Byrne** · [Slack thread](https://comfy-organization.slack.com/archives/C0BGH348Z0C/p1789592320785689?thread_ts=1789592320.785689&cid=C0BGH348Z0C)_ ## ELI5 Redo rebuilds the graph from a saved snapshot. While rebuilding, the graph works out afresh the order in which nodes should run and stamps each node with a new number. The undo history then compares the live graph with its snapshot, sees those numbers differ, and concludes you edited something, so the next plain click on the canvas records a phantom "change". Pressing Ctrl+Z after that restores a snapshot that looks identical, so undo appears to do nothing and you have to press it twice. This PR makes the comparison ignore that run-order number, which the user never edits. ## Motivation Before: connect two nodes, undo, redo, click the canvas, press Ctrl+Z: the link stays. The click after the redo pushed an undo entry that differs from the live graph only in node `order`, so the first Ctrl+Z is a visible no-op and the redo queue is wiped. After: the click after a redo records nothing, and Ctrl+Z removes the redone link. Why: `ChangeTracker.graphEqual` deep-compares node objects, including `order`. `order` is the execution index that `LGraph.updateExecutionOrder()` recomputes on every configure, add or remove; a snapshot captured before a reload and the graph rebuilt from it legitimately disagree on it. The auto-queue comparison in the same file already lists `order` among the non-execution node properties it ignores. ## Provenance - **Authored by:** interactive session - **Verified:** Vitest `changeTracker.test.ts`, new case `does not push when only the recomputed node execution order differs`: fails on the test-only commit (`expected [ Array(1) ] to have a length of +0 but got 1`), and the whole file passes with the fix (63/63). Playwright `changeTracker.spec.ts`, new case `Undo after a redo and a canvas click removes the redone link`: fails on the test-only commit at its final assertion (`expected 0, received 1` links) with every earlier step passing, and passes with the fix. The other 6 failing cases in that spec fail identically in the local sandbox with the fix stashed (legacy-menu and subgraph-entry environment issues) and are green in CI. Full `pnpm lint`, `pnpm format:check`, typecheck, eslint, oxlint and oxfmt on the changed files: clean. - **Deviations:** none ## Reviewer context - **Type:** bug fix, undo/redo change detection - **Slots into:** Comfy-Org#18085 (repro and isolation PR). The phantom entry was found while isolating the link drift there. The missing undo entry for a Vue node slot drag-connect is a separate bug, fixed in Comfy-Org#18117. - The e2e clicks two different empty spots on purpose: two same-spot clicks inside the double-click window open node search, which swallows Ctrl+Z and Ctrl+Y. ## How `graphEqual` compares the two node lists with `order` omitted from every node. Nothing else in the comparison changes. ## Red-green verification | Commit | CI | Purpose | |---|---|---| | `test: cover the spurious undo entry pushed by a canvas click after redo` (e7539e0) | red | the unit case pushes an entry for an order-only difference (unit `test` job failed); the Playwright case leaves the link in place after the post-redo Ctrl+Z (e2e-status failed on shard 2 and the new-tests video job) | | `fix: ignore recomputed node execution order when comparing graph snapshots` (c2cffd8) | green | same cases pass, all checks green | ## Test plan - [x] CI red on the test-only commit (unit `test` job and e2e-status failed, shard 2 and the new-tests video job) - [x] CI green on the fix commit - [ ] Connect two nodes, Ctrl+Z, Ctrl+Y, click empty canvas, Ctrl+Z: the link is removed on the first press - [ ] Move a node, click the canvas, Ctrl+Z: the move is still undone (a real change is still detected) --- _Generated by [Claude Code](https://claude.ai/code/session_01JwD8PkozfznCoUwAG46WAg)_ --------- Co-authored-by: Claude <noreply@anthropic.com>
…tay on their slots through undo/redo (Comfy-Org#18110) <!-- ccr-slack-attribution --> _Requested by **Christian Byrne** · [Slack thread](https://comfy-organization.slack.com/archives/C0BGH348Z0C/p1789592320785689?thread_ts=1789592320.785689&cid=C0BGH348Z0C)_ ## ELI5 A Vue node measures where each of its slot dots sits and stores the measurements so the canvas knows where to draw the wire end. Undo and redo rebuild the graph from a snapshot, and that rebuild throws all those measurements away, but the node on screen never changes and never re-measures. With no measurement the canvas falls back to its own guess of where the widget row is, and that guess can be a full row off, so the wire visibly lands on the wrong slot. This PR tells mounted Vue nodes to measure again whenever their stored measurements are dropped. ## Motivation Before: wiring a node into a Vue-rendered widget slot, then pressing undo and redo, could leave the wire end drawn on a different row than the slot dot it is connected to (38 px off on a promoted subgraph host with a multiline text widget above the slot, 10 px off on a plain EmptyLatentImage's third widget). The link stayed correct in the graph, only its rendering was wrong, and it stayed wrong until something else resized the node. After: the wire end stays on the slot dot through undo and redo, for subgraph hosts and plain nodes alike. Why: the graph reload behind undo and redo purges every DOM-measured slot offset from the layout store, and nothing re-measured a Vue node that survived the reload with identical rows and size. ## Provenance - **Authored by:** interactive session - **Verified:** Playwright `subgraphPromotedWidgetExternalLink.spec.ts` drift rows on the test-only commit: 2 failed on the drift assertion (host 38 px, plain batch_size 10 px); with the fix: both pass and the rest of the spec keeps its expected outcomes. Vitest `src/renderer/core/layout` + `src/renderer/extensions/vueNodes/components`: 33 files, 262 passed (the new NodeSlots case fails pre-fix with `expected null to deeply equal { x: 0, y: 20 }`). `vue-tsc --noEmit` and `vue-tsc --project browser_tests/tsconfig.json`: clean. eslint, oxlint, oxfmt on the changed files: clean. `knip --cache`: clean. `changeTracker.spec.ts` fails identically in the local sandbox with and without the fix (legacy-canvas collapse clicks and default-workflow node locators), so those failures are environmental there; CI is the arbiter. - **Deviations:** none ## Reviewer context - **Type:** bug fix, Vue node link rendering after undo/redo - **Slots into:** Comfy-Org#18085 — test: add failing repros for promoted widget external-link rendering. This PR ports that spec and supersedes its `test.fail()` drift case (`keeps the wire on the promoted width slot dot after undo and redo`), which becomes a real table-driven regression test here with a plain-node row added; Comfy-Org#18085 should drop that case once this lands. The label and textarea cases no longer carry `test.fail()`: the width-label cases assert the label stays visible after wiring, and the promoted text case asserts the textarea is enabled before wiring and visible but disabled afterward, with its input slot dot shown. ## How - `layoutStore` gains a reactive `slotOffsetVersion` that is bumped only when measured slot offsets are dropped in bulk (`clearGraph`, `clearViewGeometry`), never on a measurement write, so re-measuring cannot feed back into itself. - `NodeSlots.vue` adds that version to its existing `syncSlotOffsets` watch. Every mounted Vue node renders one `NodeSlots`, and `syncSlotOffsets` measures the whole `[data-node-id]` element, so one trigger per node re-measures every slot (inputs, outputs and widget rows). Purging on clear stays as is; keeping stale offsets would be wrong when the reload does change rows. ## Red-green verification | Commit | CI | Purpose | |---|---|---| | `test: cover the wire drifting off a Vue widget slot after undo and redo` | red | Playwright drift rows (host + plain node) and the NodeSlots re-measure case fail | | `fix: re-measure Vue node slot offsets after the layout store drops them` | green | same tests pass | ## Test plan - [ ] CI red on the test-only commit, green on the fix commit - [ ] Load a workflow with a Vue node whose widget rows differ in height from litegraph's layout (a multiline text widget above a number widget), wire a primitive into the number widget, click the canvas, Ctrl+Z, click elsewhere, Ctrl+Y: the wire end stays on the slot dot - [ ] Enter and leave a subgraph, then check links to widget slots still land on their dots --- _Generated by [Claude Code](https://claude.ai/code/session_01JwD8PkozfznCoUwAG46WAg)_ --------- Co-authored-by: Claude <noreply@anthropic.com> Co-authored-by: bymyself <cbyrne@comfy.org>
…omfy-Org#18123) <!-- ccr-slack-attribution --> _Requested by **Christian Byrne** · [Slack thread](https://comfy-organization.slack.com/archives/C0BGH348Z0C/p1789592320785689?thread_ts=1789592320.785689&cid=C0BGH348Z0C)_ ## ELI5 When a wire takes over a node's input, the input's own control steps aside, but its name tag should stay so you can still tell which input the wire feeds. Vue nodes used to take the name tag away too, leaving a bare dot. Comfy-Org#17576 fixed that on main; this PR adds the browser test that proves it stays fixed, and lets the screenshot workflow regenerate the one baseline that moved. ## Motivation In Vue nodes, wiring a `PrimitiveInt` into the `width` input of a subgraph host or a plain `EmptyLatentImage` collapsed the row to an unlabeled dot, so the node read `text`, then straight to `height`, and the report described the widget as having vanished. The legacy canvas keeps the row and its name for a connection-suppressed widget (`occupiesCanvasRow`), so this was a renderer parity gap on every plain node, not a subgraph bug. The repro landed as `test.fail()` rows in Comfy-Org#18085; this PR converts them into a real regression test. ## What is in this PR after rebasing onto Comfy-Org#17576 This PR originally patched `WidgetGrid.vue` with `:dot-only="!row.standalone"`. Comfy-Org#17576 landed first with `:dot-only="row.showsControl"` plus the forwarded `label`; since `standalone` is `!showsControl && suppressedByConnection`, that already covers the connection-suppressed row (and additionally labels converted-widget rows), and its unit test asserts the label text directly. Both source commits were dropped in the rebase as already upstream. What remains: - `browser_tests/tests/vueNodes/widgets/connectedWidgetRowLabel.spec.ts` + fixture `subgraph-promoted-int-text-with-primitives.json`: wires a `PrimitiveInt` into `width` on a promoted subgraph host and on a plain `EmptyLatentImage`, then asserts the `width` label and the slot dot both stay visible. - `subgraphSlots.spec.ts` "link interactions": tagged `@screenshot` so the `New Browser Test Expectations` workflow can regenerate its `vue-io-highlight.png` baseline. The baseline this branch regenerated was byte-identical to the one Comfy-Org#17576 merged, so that commit dropped too and the test runs against main's image. ## Provenance - **Authored by:** interactive session - **Verified:** `pnpm typecheck` (root plus account-core, billing-web, account-ui, billing-contract) exit 0; `WidgetGrid.test.ts` + `InputSlot.test.ts`: 2 files, 10 passed; `oxfmt --check` and `eslint` on the changed files clean; pre-push `pnpm knip` passed. Playwright coverage runs in CI on this head. - **Deviations:** the `fix:` and unit-test commits were superseded by Comfy-Org#17576 during the rebase; the PR is test-only now. ## Reviewer context - **Type:** test-only, Playwright regression coverage for connection-suppressed widget rows in Vue nodes - **Slots into:** Comfy-Org#17576 (the fix), Comfy-Org#18085 (the original `test.fail()` repros this replaces) - **Not in scope:** the promoted textarea behavior on connect (Comfy-Org#18114, closed), the redo wire drift (separate PR) ## Test plan - [ ] CI green: `connectedWidgetRowLabel.spec.ts` passes on both rows, `vue-io-highlight.png` holds against main's baseline - [ ] In Vue nodes, wire `PrimitiveInt` into `EmptyLatentImage.width`: the `width` label stays next to the dot, the number control is gone, the wire lands on the dot; disconnect restores the control 🤖 Generated with [Claude Code](https://claude.com/claude-code) https://claude.ai/code/session_01JwD8PkozfznCoUwAG46WAg --------- Co-authored-by: Claude <noreply@anthropic.com>
<!-- ccr-slack-attribution --> _Requested by **Jo Zhang, Christian Byrne** · [Slack thread](https://comfy-organization.slack.com/archives/C0BGH348Z0C/p1789707519592879?thread_ts=1789707519.592879&cid=C0BGH348Z0C)_ ## ELI5 A subgraph node shows several "promoted" widgets from the nodes hidden inside it (like width, height, seed). When you rewire the connection behind just one of those widgets, the others could flicker, duplicate, or vanish for a moment, and the wired-in widget itself could disappear. This PR fixes that and adds tests at both levels: unit tests prove the mechanism, and the browser-level repro tests — which this PR's own CI confirmed now pass with the fix — are kept as plain regression tests. - Fixes #14495 ## Motivation This started as a REPRO-ONLY PR for a reported bug: wiring a new node into a subgraph's promoted "prompt" widget makes the subgraph's *other* promoted widgets render duplicated, and a closely related report has the rewired widget itself disappear with redo misplacing the wire. Investigation traced a confirmed mechanism for the disappearing half at the data-model layer (see Summary). Once that mechanism was nailed down precisely enough to fix without guessing, and a second, more contained duplication theory was checked and refuted, it became worth landing the fix and the validating tests in the same PR rather than leaving a fully-diagnosed, small, well-scoped bug unfixed. ## Provenance - **Authored by:** interactive session, Claude Code - **Verified:** `pnpm exec vitest run src/lib/litegraph/src/subgraph/SubgraphWidgetPromotion.test.ts` — 51 passed, 0 failed (was 50 passed / 1 expected fail before the fix; that case now passes for real and was flipped from `it.fails` to `it`). `pnpm exec vitest run src/lib/litegraph/src/subgraph/ src/renderer/extensions/vueNodes/widgets/utils/multilineTextarea.promoted.test.ts src/stores/widgetValueStore.test.ts src/stores/domWidgetStore.test.ts` — 413 passed, 0 failed. `pnpm typecheck`, `pnpm lint`, `pnpm knip` — clean. E2E confirmation came from this PR's own CI (see Summary). Re-verified after the 2026-10-01 rebase onto main: same scoped suites green, typecheck/lint/knip clean. - **Deviations:** none remaining. The previous push deliberately left the e2e `test.fail()` cases unflipped pending CI evidence; that evidence has now arrived (see "E2E confirmation" below) and they are flipped to plain `test()` in this push, exactly per the plan recorded in the earlier Test plan. ## Reviewer context - **Type:** fix + test — a small, contained timing fix plus the tests that prove it, validate/refute an alternate duplication theory, and lock in the browser-level symptom as regression coverage - **Slots into:** a subgraph promoted-widget-promotion bug investigation; no project or epic, ad-hoc follow-up to that investigation - **Not in scope:** `createPromotedMultilineWidget`'s visibility/suppression path — that's PR #18085's territory (a sibling investigation into the same widget family) ## Summary ### The fix (`src/lib/litegraph/src/subgraph/SubgraphNode.ts`) Rewiring the interior link behind a promoted widget (e.g. dragging a new node onto an already-promoted widget's socket) removes the old link before the new one lands. `SubgraphInputNode`'s `'input-disconnected'` handler used to demote the widget **synchronously** — clearing `input.widgetId`/`.widget`/`._widget` immediately — while deferring the matching `widgetValueStore` cleanup to a `queueMicrotask`. For one tick, `hostNode.widgets.length` (synchronous) and the store's tracked id count for that node (deferred) disagreed. The Vue widget grid (`useProcessedWidgets`/`computeProcessedWidgets`) reads its render *order* from the store and its liveness *filter* from `hostNode.widgets` — i.e. from both sides of that exact disagreement. The fix defers the demotion itself by a microtask (mirroring the already-deferred store cleanup), tracked with a per-input cancellation token. The `'input-connected'` listener clears that token on a same-tick reconnect, so a rewire never drops the widget from `hostNode.widgets` at all. A genuine disconnect with no reconnect still fully demotes once the microtask runs — nothing about final-state behavior changes, only the timing of an in-between state that used to be observable. **Deliberate timing-change note** (per this repo's own ecosystem-impact rule for changes touching `node.widgets` access): this changes when a disconnect becomes observable on `hostNode.widgets` — previously synchronous, now one microtask tick later for a genuine disconnect. Two existing tests in `SubgraphWidgetPromotion.test.ts` asserted this synchronously with no `await`; both were updated to `await Promise.resolve()` after the disconnecting call, with a comment explaining why. No public API shape or callback signature changed — only internal timing. ### New tests - `SubgraphWidgetPromotion.test.ts`, "keeps every promoted widget present while rewiring the interior link behind one of them" (previously `it.fails`, now a plain `it`): builds a host with 4 promoted widgets from 4 different interior nodes, rewires the link behind one of them onto a *new* interior node mid-test, and asserts `hostNode.widgets.length` stays at 4 throughout — including immediately after the disconnect step, before the reconnect lands. - `SubgraphWidgetPromotion.test.ts`, "Promoted textarea dual-registration (PM-1328 duplication lead)": tests a second, independently-raised theory for the duplication symptom — that a promoted textarea's DOM-hosted widget (a `DOMWidgetImpl` living in `useDomWidgetStore`, materialized by `createPromotedHostWidget`) and a store-projected row could both end up registered for the same input, which would duplicate a row in the widget grid. Built a minimal `SubgraphNode` subclass wired the same way the app's real subclass is, exercised it across a rewire and a `configure()` reconfigure. **Both pass as-is — the theory is refuted.** `_projectPromotedWidget`'s `if (input._widget) return input._widget` guard is unconditional across every call site and there's no observable window where `input.widgetId` is set while `input._widget` isn't. Kept as regression coverage. ### E2E confirmation (CI run [35419314544](https://github.com/Comfy-Org/ComfyUI_frontend/actions/runs/35419314544)) CI on the previous head answered the open question the `test.fail()` repros were posted to ask — the fix **does** resolve the browser-level symptom end-to-end: - Both `subgraphPromotedWidgetRewireDuplication.spec.ts` repro scenarios (interior rewire and external boundary wire) were reported "Expected to fail, but passed" on every run and retry (video job + shard 12, 4/4 retries). Flipped to plain `test()` regression tests, and retitled to describe the behavior they now guarantee. - `subgraphPromotion.spec.ts` "Promoted STRING widget edit survives a rebind of the interior link" — a *pre-existing* `test.fail()` for #14495 — also passed on all 4 shard-12 attempts. Flipped to plain `test()`; hence "Fixes #14495" above. - The external-boundary undo test went through two CI-informed revisions after the rebase: which snapshot undo targets depends on change-tracker capture timing the test cannot control (captures fire on window mouseup, before the click handlers that apply promotions, so the entry pushed by the slot drop — #18117 — can lag the built baseline by one action), and an attempt to pin the snapshot with an extra canvas click timed out in CI. The test now asserts the invariant both reported symptoms violate instead of a fixed widget set: whatever state undo restores, the host's model holds no duplicate widgets and the grid renders exactly one row per model widget, none otherwise. ## Test plan - [x] Watch this PR's `playwright-video-new-tests` CI check: the `test.fail()` cases were reported as unexpectedly passed (run 35419314544) - [x] Flip those cases (and the pre-existing #14495 one) to plain `test()` - [x] Make the external-boundary undo test deterministic (it now asserts render/model consistency after undo instead of assuming which snapshot undo targets) - [ ] CI green on the rebased head (the "Check for AI agent co-author trailers" check was removed from main in #19393; the first commit still carries trailers, and rewording it remains a maintainer call since it rewrites shared history) 🤖 Generated with [Claude Code](https://claude.com/claude-code) https://claude.ai/code/session_01U6aDANuoAYDtxd9WiCk44Q --------- Co-authored-by: Claude <noreply@anthropic.com>
<!-- ccr-slack-attribution --> _Requested by **Jo Zhang, Christian Byrne** · [Slack thread](https://comfy-organization.slack.com/archives/C0BGH348Z0C/p1789707519592879?thread_ts=1789707519.592879&cid=C0BGH348Z0C)_ A subgraph node shows several "promoted" widgets from the nodes hidden inside it (like width, height, seed). When you rewire the connection behind just one of those widgets, the others could flicker, duplicate, or vanish for a moment, and the wired-in widget itself could disappear. This PR fixes that and adds tests at both levels: unit tests prove the mechanism, and the browser-level repro tests — which this PR's own CI confirmed now pass with the fix — are kept as plain regression tests. - Fixes #14495 This started as a REPRO-ONLY PR for a reported bug: wiring a new node into a subgraph's promoted "prompt" widget makes the subgraph's *other* promoted widgets render duplicated, and a closely related report has the rewired widget itself disappear with redo misplacing the wire. Investigation traced a confirmed mechanism for the disappearing half at the data-model layer (see Summary). Once that mechanism was nailed down precisely enough to fix without guessing, and a second, more contained duplication theory was checked and refuted, it became worth landing the fix and the validating tests in the same PR rather than leaving a fully-diagnosed, small, well-scoped bug unfixed. - **Authored by:** interactive session, Claude Code - **Verified:** `pnpm exec vitest run src/lib/litegraph/src/subgraph/SubgraphWidgetPromotion.test.ts` — 51 passed, 0 failed (was 50 passed / 1 expected fail before the fix; that case now passes for real and was flipped from `it.fails` to `it`). `pnpm exec vitest run src/lib/litegraph/src/subgraph/ src/renderer/extensions/vueNodes/widgets/utils/multilineTextarea.promoted.test.ts src/stores/widgetValueStore.test.ts src/stores/domWidgetStore.test.ts` — 413 passed, 0 failed. `pnpm typecheck`, `pnpm lint`, `pnpm knip` — clean. E2E confirmation came from this PR's own CI (see Summary). Re-verified after the 2026-10-01 rebase onto main: same scoped suites green, typecheck/lint/knip clean. - **Deviations:** none remaining. The previous push deliberately left the e2e `test.fail()` cases unflipped pending CI evidence; that evidence has now arrived (see "E2E confirmation" below) and they are flipped to plain `test()` in this push, exactly per the plan recorded in the earlier Test plan. - **Type:** fix + test — a small, contained timing fix plus the tests that prove it, validate/refute an alternate duplication theory, and lock in the browser-level symptom as regression coverage - **Slots into:** a subgraph promoted-widget-promotion bug investigation; no project or epic, ad-hoc follow-up to that investigation - **Not in scope:** `createPromotedMultilineWidget`'s visibility/suppression path — that's PR #18085's territory (a sibling investigation into the same widget family) Rewiring the interior link behind a promoted widget (e.g. dragging a new node onto an already-promoted widget's socket) removes the old link before the new one lands. `SubgraphInputNode`'s `'input-disconnected'` handler used to demote the widget **synchronously** — clearing `input.widgetId`/`.widget`/`._widget` immediately — while deferring the matching `widgetValueStore` cleanup to a `queueMicrotask`. For one tick, `hostNode.widgets.length` (synchronous) and the store's tracked id count for that node (deferred) disagreed. The Vue widget grid (`useProcessedWidgets`/`computeProcessedWidgets`) reads its render *order* from the store and its liveness *filter* from `hostNode.widgets` — i.e. from both sides of that exact disagreement. The fix defers the demotion itself by a microtask (mirroring the already-deferred store cleanup), tracked with a per-input cancellation token. The `'input-connected'` listener clears that token on a same-tick reconnect, so a rewire never drops the widget from `hostNode.widgets` at all. A genuine disconnect with no reconnect still fully demotes once the microtask runs — nothing about final-state behavior changes, only the timing of an in-between state that used to be observable. **Deliberate timing-change note** (per this repo's own ecosystem-impact rule for changes touching `node.widgets` access): this changes when a disconnect becomes observable on `hostNode.widgets` — previously synchronous, now one microtask tick later for a genuine disconnect. Two existing tests in `SubgraphWidgetPromotion.test.ts` asserted this synchronously with no `await`; both were updated to `await Promise.resolve()` after the disconnecting call, with a comment explaining why. No public API shape or callback signature changed — only internal timing. - `SubgraphWidgetPromotion.test.ts`, "keeps every promoted widget present while rewiring the interior link behind one of them" (previously `it.fails`, now a plain `it`): builds a host with 4 promoted widgets from 4 different interior nodes, rewires the link behind one of them onto a *new* interior node mid-test, and asserts `hostNode.widgets.length` stays at 4 throughout — including immediately after the disconnect step, before the reconnect lands. - `SubgraphWidgetPromotion.test.ts`, "Promoted textarea dual-registration (PM-1328 duplication lead)": tests a second, independently-raised theory for the duplication symptom — that a promoted textarea's DOM-hosted widget (a `DOMWidgetImpl` living in `useDomWidgetStore`, materialized by `createPromotedHostWidget`) and a store-projected row could both end up registered for the same input, which would duplicate a row in the widget grid. Built a minimal `SubgraphNode` subclass wired the same way the app's real subclass is, exercised it across a rewire and a `configure()` reconfigure. **Both pass as-is — the theory is refuted.** `_projectPromotedWidget`'s `if (input._widget) return input._widget` guard is unconditional across every call site and there's no observable window where `input.widgetId` is set while `input._widget` isn't. Kept as regression coverage. [35419314544](https://github.com/Comfy-Org/ComfyUI_frontend/actions/runs/35419314544)) CI on the previous head answered the open question the `test.fail()` repros were posted to ask — the fix **does** resolve the browser-level symptom end-to-end: - Both `subgraphPromotedWidgetRewireDuplication.spec.ts` repro scenarios (interior rewire and external boundary wire) were reported "Expected to fail, but passed" on every run and retry (video job + shard 12, 4/4 retries). Flipped to plain `test()` regression tests, and retitled to describe the behavior they now guarantee. - `subgraphPromotion.spec.ts` "Promoted STRING widget edit survives a rebind of the interior link" — a *pre-existing* `test.fail()` for #14495 — also passed on all 4 shard-12 attempts. Flipped to plain `test()`; hence "Fixes #14495" above. - The external-boundary undo test went through two CI-informed revisions after the rebase: which snapshot undo targets depends on change-tracker capture timing the test cannot control (captures fire on window mouseup, before the click handlers that apply promotions, so the entry pushed by the slot drop — #18117 — can lag the built baseline by one action), and an attempt to pin the snapshot with an extra canvas click timed out in CI. The test now asserts the invariant both reported symptoms violate instead of a fixed widget set: whatever state undo restores, the host's model holds no duplicate widgets and the grid renders exactly one row per model widget, none otherwise. - [x] Watch this PR's `playwright-video-new-tests` CI check: the `test.fail()` cases were reported as unexpectedly passed (run 35419314544) - [x] Flip those cases (and the pre-existing #14495 one) to plain `test()` - [x] Make the external-boundary undo test deterministic (it now asserts render/model consistency after undo instead of assuming which snapshot undo targets) - [ ] CI green on the rebased head (the "Check for AI agent co-author trailers" check was removed from main in #19393; the first commit still carries trailers, and rewording it remains a maintainer call since it rewrites shared history) 🤖 Generated with [Claude Code](https://claude.com/claude-code) https://claude.ai/code/session_01U6aDANuoAYDtxd9WiCk44Q --------- (cherry picked from commit 9ee67cb) (The two conflicting test files are byte-identical between core/1.54 and cloud/1.54 at their merge base, so this is the same resolution as the core/1.54 carrier. `SubgraphNode.ts` differs slightly between the two branches but the fix applied cleanly to both.)
<!-- ccr-slack-attribution --> _Requested by **Jo Zhang, Christian Byrne** · [Slack thread](https://comfy-organization.slack.com/archives/C0BGH348Z0C/p1789707519592879?thread_ts=1789707519.592879&cid=C0BGH348Z0C)_ A subgraph node shows several "promoted" widgets from the nodes hidden inside it (like width, height, seed). When you rewire the connection behind just one of those widgets, the others could flicker, duplicate, or vanish for a moment, and the wired-in widget itself could disappear. This PR fixes that and adds tests at both levels: unit tests prove the mechanism, and the browser-level repro tests — which this PR's own CI confirmed now pass with the fix — are kept as plain regression tests. - Fixes #14495 This started as a REPRO-ONLY PR for a reported bug: wiring a new node into a subgraph's promoted "prompt" widget makes the subgraph's *other* promoted widgets render duplicated, and a closely related report has the rewired widget itself disappear with redo misplacing the wire. Investigation traced a confirmed mechanism for the disappearing half at the data-model layer (see Summary). Once that mechanism was nailed down precisely enough to fix without guessing, and a second, more contained duplication theory was checked and refuted, it became worth landing the fix and the validating tests in the same PR rather than leaving a fully-diagnosed, small, well-scoped bug unfixed. - **Authored by:** interactive session, Claude Code - **Verified:** `pnpm exec vitest run src/lib/litegraph/src/subgraph/SubgraphWidgetPromotion.test.ts` — 51 passed, 0 failed (was 50 passed / 1 expected fail before the fix; that case now passes for real and was flipped from `it.fails` to `it`). `pnpm exec vitest run src/lib/litegraph/src/subgraph/ src/renderer/extensions/vueNodes/widgets/utils/multilineTextarea.promoted.test.ts src/stores/widgetValueStore.test.ts src/stores/domWidgetStore.test.ts` — 413 passed, 0 failed. `pnpm typecheck`, `pnpm lint`, `pnpm knip` — clean. E2E confirmation came from this PR's own CI (see Summary). Re-verified after the 2026-10-01 rebase onto main: same scoped suites green, typecheck/lint/knip clean. - **Deviations:** none remaining. The previous push deliberately left the e2e `test.fail()` cases unflipped pending CI evidence; that evidence has now arrived (see "E2E confirmation" below) and they are flipped to plain `test()` in this push, exactly per the plan recorded in the earlier Test plan. - **Type:** fix + test — a small, contained timing fix plus the tests that prove it, validate/refute an alternate duplication theory, and lock in the browser-level symptom as regression coverage - **Slots into:** a subgraph promoted-widget-promotion bug investigation; no project or epic, ad-hoc follow-up to that investigation - **Not in scope:** `createPromotedMultilineWidget`'s visibility/suppression path — that's PR #18085's territory (a sibling investigation into the same widget family) Rewiring the interior link behind a promoted widget (e.g. dragging a new node onto an already-promoted widget's socket) removes the old link before the new one lands. `SubgraphInputNode`'s `'input-disconnected'` handler used to demote the widget **synchronously** — clearing `input.widgetId`/`.widget`/`._widget` immediately — while deferring the matching `widgetValueStore` cleanup to a `queueMicrotask`. For one tick, `hostNode.widgets.length` (synchronous) and the store's tracked id count for that node (deferred) disagreed. The Vue widget grid (`useProcessedWidgets`/`computeProcessedWidgets`) reads its render *order* from the store and its liveness *filter* from `hostNode.widgets` — i.e. from both sides of that exact disagreement. The fix defers the demotion itself by a microtask (mirroring the already-deferred store cleanup), tracked with a per-input cancellation token. The `'input-connected'` listener clears that token on a same-tick reconnect, so a rewire never drops the widget from `hostNode.widgets` at all. A genuine disconnect with no reconnect still fully demotes once the microtask runs — nothing about final-state behavior changes, only the timing of an in-between state that used to be observable. **Deliberate timing-change note** (per this repo's own ecosystem-impact rule for changes touching `node.widgets` access): this changes when a disconnect becomes observable on `hostNode.widgets` — previously synchronous, now one microtask tick later for a genuine disconnect. Two existing tests in `SubgraphWidgetPromotion.test.ts` asserted this synchronously with no `await`; both were updated to `await Promise.resolve()` after the disconnecting call, with a comment explaining why. No public API shape or callback signature changed — only internal timing. - `SubgraphWidgetPromotion.test.ts`, "keeps every promoted widget present while rewiring the interior link behind one of them" (previously `it.fails`, now a plain `it`): builds a host with 4 promoted widgets from 4 different interior nodes, rewires the link behind one of them onto a *new* interior node mid-test, and asserts `hostNode.widgets.length` stays at 4 throughout — including immediately after the disconnect step, before the reconnect lands. - `SubgraphWidgetPromotion.test.ts`, "Promoted textarea dual-registration (PM-1328 duplication lead)": tests a second, independently-raised theory for the duplication symptom — that a promoted textarea's DOM-hosted widget (a `DOMWidgetImpl` living in `useDomWidgetStore`, materialized by `createPromotedHostWidget`) and a store-projected row could both end up registered for the same input, which would duplicate a row in the widget grid. Built a minimal `SubgraphNode` subclass wired the same way the app's real subclass is, exercised it across a rewire and a `configure()` reconfigure. **Both pass as-is — the theory is refuted.** `_projectPromotedWidget`'s `if (input._widget) return input._widget` guard is unconditional across every call site and there's no observable window where `input.widgetId` is set while `input._widget` isn't. Kept as regression coverage. [35419314544](https://github.com/Comfy-Org/ComfyUI_frontend/actions/runs/35419314544)) CI on the previous head answered the open question the `test.fail()` repros were posted to ask — the fix **does** resolve the browser-level symptom end-to-end: - Both `subgraphPromotedWidgetRewireDuplication.spec.ts` repro scenarios (interior rewire and external boundary wire) were reported "Expected to fail, but passed" on every run and retry (video job + shard 12, 4/4 retries). Flipped to plain `test()` regression tests, and retitled to describe the behavior they now guarantee. - `subgraphPromotion.spec.ts` "Promoted STRING widget edit survives a rebind of the interior link" — a *pre-existing* `test.fail()` for #14495 — also passed on all 4 shard-12 attempts. Flipped to plain `test()`; hence "Fixes #14495" above. - The external-boundary undo test went through two CI-informed revisions after the rebase: which snapshot undo targets depends on change-tracker capture timing the test cannot control (captures fire on window mouseup, before the click handlers that apply promotions, so the entry pushed by the slot drop — #18117 — can lag the built baseline by one action), and an attempt to pin the snapshot with an extra canvas click timed out in CI. The test now asserts the invariant both reported symptoms violate instead of a fixed widget set: whatever state undo restores, the host's model holds no duplicate widgets and the grid renders exactly one row per model widget, none otherwise. - [x] Watch this PR's `playwright-video-new-tests` CI check: the `test.fail()` cases were reported as unexpectedly passed (run 35419314544) - [x] Flip those cases (and the pre-existing #14495 one) to plain `test()` - [x] Make the external-boundary undo test deterministic (it now asserts render/model consistency after undo instead of assuming which snapshot undo targets) - [ ] CI green on the rebased head (the "Check for AI agent co-author trailers" check was removed from main in #19393; the first commit still carries trailers, and rewording it remains a maintainer call since it rewrites shared history) 🤖 Generated with [Claude Code](https://claude.com/claude-code) https://claude.ai/code/session_01U6aDANuoAYDtxd9WiCk44Q --------- (cherry picked from commit 9ee67cb) ## Backport adaptation (core/1.54) The fix itself (`SubgraphNode.ts`, `interfaces.ts`) cherry-picked clean and is byte-identical to upstream. Only two test files conflicted, both because the cherry-pick's context included neighbouring code that is not on this branch: - `useErrorClearingHooks.test.ts` — the conflict hunk also carried an unrelated test, "retires an absorbed promoted media error when its host is deleted", which main gained from a different, unbackported PR. Dropped that test and applied only this PR's change to the test that does exist here: make "keeps promoted missing media until the last fanout consumer is deleted" async and await the deferred demotion. - `SubgraphWidgetPromotion.test.ts` — import-only conflict. Accepted the three new imports (`WidgetValue`, `createPromotedMultilineWidget`, `useDomWidgetStore`) required by the new PM-1328 describe block; verified all three modules and symbols exist on this branch. The new block also uses `LegacyWidget`, which main imports from the litegraph barrel in a neighbouring non-conflicting import hunk, so this branch's version of that hunk won and left it out; the test failed with `ReferenceError: LegacyWidget is not defined` until it was added to the barrel import. `LegacyWidget` already exists and is exported from the barrel on this branch. No backport-only compatibility shim was needed and no public API shape or callback signature changed. The deliberate timing change this PR documents (a genuine disconnect becomes observable on `hostNode.widgets` one microtask later) is carried as-is.
Requested by Christian Byrne · Slack thread
Summary
Intentionally failing repro tests (Playwright
test.fail(), Vitestit.fails) for the "promoted subgraph widget vanishes / wire drifts after redo" report when a parent-graph node is wired into a subgraph host's already-promoted input. No fix in this PR; every red test here is the reproduction, verified locally to fail for the reason stated in its title.Before / After
Before (today, Vue nodes): wire a
PrimitiveIntinto the host's promotedwidthand thewidthrow collapses to an unlabeled 24px dot row, so the node readstext, then straight toheight. Wire aPrimitiveStringinto the promotedtextand the textarea stays fully visible and editable while the wire lands. Undo, then redo the width connection and litegraph draws the wire end 38px below thewidthslot dot, visually on theheightrow.After (expected, what the tests assert): a connection-suppressed widget row keeps its label next to the slot dot (legacy-canvas parity,
occupiesCanvasRow), the promoted textarea hides like a plainCLIPTextEncodetextarea does, and the wire end stays on thewidthslot dot through undo/redo.Changes
browser_tests/tests/subgraph/subgraphPromotedWidgetExternalLink.spec.ts(+ assetsubgraph-promoted-int-text-with-primitives.json): 4test.fail()e2e repros plus 1 passing contrast test;src/lib/litegraph/src/subgraph/SubgraphPromotedWidgetExternalLink.test.ts: 7 passing litegraph-level tests pinning the mechanism;src/renderer/extensions/vueNodes/composables/useProcessedWidgets.promotedExternalLink.regression.test.ts: 2it.failsrender-model repros plus 2 passing contrasts.What each test targets and what we learned
Two root-cause hypotheses were on the table. Each test names the one it exercises.
Hypothesis A:
connectionSuppressedwritten to a visibility component the Vue renderer never reads. Confirmed for the promoted textarea, refuted for the "stuck flag" part.LGraphNode.updateComputedDisabled()setswidget.connectionSuppressed = connectedon every_addLink/_removeLink. For a store-projection promoted widget (promotedWidgetStoreProjection.ts) the setter writes straight intouseWidgetValueStore()visibility, so the INT row suppresses correctly. For the promoted textarea,createPromotedMultilineWidgetbuilds aDOMWidgetImplthat is never passed throughBaseWidget.setNodeId(), so its_visibilityis a private component disconnected from the store entrySubgraphNode._setWidgetregistered. The store'ssuppression.byConnectionstaysfalse,computeProcessedWidgetskeepsvisible: true, and the textarea never hides.it.failsx2 in the regression test file reproduce exactly this (store flagfalse, render modelvisible: true) against the realSubgraphNode+ realcreatePromotedMultilineWidget; the e2ehides the promoted text control after a String is wired into the hostshows it end to end, with the plainCLIPTextEncodecontrast test passing.registerWidgetre-registration only copiesbyExtension" part does not hold:LGraph.configure(what undo/redo goes through) clears the widget store for the root graph and re-runsupdateComputedDisabledon every node. Two passing tests (lifts the connection suppression when an undo-style reload…,keeps the connection suppression when a redo-style reload…) pin that.Hypothesis B:
SubgraphInputinput-disconnecteddemotes the promoted widget with aqueueMicrotaskstore cleanup. Does not fire for this trigger.does not demote the promoted widget when the external link landscapturesinput-disconnectedandwidget-demotedduring an external parent-graph connect and asserts zero events; the widget id and store entry survive. That mechanism is about rewiring the interior link inside the subgraph, not about wiring into the host's boundary slot from outside.Finding not covered by either hypothesis (the "width vanishes" half): the collapsed row is the Vue
WidgetGridstandalone-slot rendering, not a subgraph bug.WidgetGrid.vuerenders a connection-suppressed widget as anInputSlotwithdot-only, so the label disappears on plain nodes too. The table-driven e2ekeeps the width label visible after an Int is wired into the {promoted subgraph host, plain EmptyLatentImage}fails identically on both rows (locatorgetByText('width')not found after wiring). Legacy canvas keeps the name (occupiesCanvasRowdoc comment), so this is a Vue-nodes parity gap.Redo drift reproduced:
keeps the wire on the promoted width slot dot after undo and redowires the Int, undoes, redoes, then comparesnode.getInputPos(1)(what litegraph draws) to the DOM slot dot. Before undo the gap is 0px; after redo it is 38px (deterministic across three local runs). The attachedafter-redo.pngshows the wire landing on theheightrow. Likely the layout store's slot offsets for the host are cleared by the reload andWidgetGrid'slayoutKeywatcher does not re-sync them, so litegraph falls back to index-based slot positions; not root-caused in this PR.Local run evidence
Unit:
Tests 9 passed | 2 expected fail (11)across the two Vitest files. Running the twoit.failscases as plainitgivesexpected false to be trueonsuppression.byConnectionandvisible: true / suppressedByConnection: falseon the render model.E2E (
--project=chromium, Vue nodes):5 passed= 4 expected failures + 1 pass. With thetest.fail()markers removed the failures are:getByText('width')element(s) not found (host and plain rows), textareatoHaveCount(0)received1, andwire end drifted off the width slot after redo: Expected <= 2, Received 38.Review Focus
widthlabel expectation encodes legacy-canvas parity as the contract. If the bare-dot row is the intended Vue design, drop those two rows and keep the other repros; the report reads it as the widget vanishing.mouseup, so the spec clicks empty canvas before undo; (2) an earlier draft of the drift test clicked the same empty pixel before undo and again before redo, which lands inside the double-click window, opens node search and swallows the redo shortcut, so redo looked like a no-op and the drift assertion was never reached. The spec now clicks a second empty pixel before redo; with the marker stripped the drift assertion fails 3/3 locally.agentConversationFixture) was not used: it requires a recorded cloud thread and the fixture schema rejects hand-writtengraph_ops. The e2e uses a UI slot drag, which reaches the sameLGraph._addLinkpath the agentconnectop does.Generated with Claude Code
https://claude.ai/code/session_01JwD8PkozfznCoUwAG46WAg
Generated by Claude Code