Repository navigation
fix(agent-crdt): filter non-value widget writes in the minter [backport cloud/1.55] - #19889
Conversation
|
Important Review skippedIgnore keyword(s) in the title. ⛔ Ignored keywords (2)
Please check the settings in the CodeRabbit UI or the ⚙️ Run configuration
You can disable this status message by setting the Use the checkbox below for a quick retry:
Comment |
🎭 Playwright: ✅ 2147 passed, 0 failed · 1 flaky📊 Browser Reports
🎨 Storybook: ✅ Built — View Storybook🌾 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. No new findings in the changed files. Reproduce locally# the base the gate compared against
npx fallow audit --root . --quiet --format json \
--changed-since 5dfb4e1461efa669ee6ebc62f52f665922a7aa5e
Full docs: ADR-DEVEX-LINT-0015 🧩 Custom Nodes Ecosystem Matrix: ✅ PassedExecutes the frontend JS of every registry pack that ships any, against this change, in both renderers. The eight
On a pin-bump PR, a red gate means the ecosystem moved — not that your diff broke something. That is what the bump PR is for. Outside a pin bump or a cold rebuild of the unpinned tail, a red means the diff. Note the corpus size and the executed count are different numbers, and the verdict is against the executed count: roughly 3,100 of ~5,100 packs ship no extension-shaped JS and are 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 📦 Bundle Size
⚡ Performance Report
Absolute values
Raw data{
"timestamp": "2026-10-03T02:04:28.381Z",
"gitSha": "5e3c43cb83a8f83d84b8658037c629777271727c",
"branch": "op284/fe3161-c2-cloud-1.55",
"measurements": [
{
"name": "canvas-idle",
"durationMs": 2000.3410000000486,
"styleRecalcs": 9,
"styleRecalcDurationMs": 7.384000000000002,
"layouts": 0,
"layoutDurationMs": 0,
"taskDurationMs": 441.4359999999999,
"heapDeltaBytes": -10568812,
"heapUsedBytes": 58527356,
"domNodes": -262,
"jsHeapTotalBytes": 5496832,
"scriptDurationMs": 7.888000000000003,
"eventListeners": -174,
"totalBlockingTimeMs": 0,
"frameDurationMs": 16.666666666666668,
"p95FrameDurationMs": 16.799999999999272
},
{
"name": "canvas-idle",
"durationMs": 1999.3540000000394,
"styleRecalcs": 10,
"styleRecalcDurationMs": 7.281,
"layouts": 0,
"layoutDurationMs": 0,
"taskDurationMs": 453.29600000000005,
"heapDeltaBytes": 8119992,
"heapUsedBytes": 76731872,
"domNodes": -262,
"jsHeapTotalBytes": 4972544,
"scriptDurationMs": 8.486999999999998,
"eventListeners": -152,
"totalBlockingTimeMs": 0,
"frameDurationMs": 16.666666666666668,
"p95FrameDurationMs": 16.699999999999818
},
{
"name": "canvas-mouse-sweep",
"durationMs": 1793.1219999999257,
"styleRecalcs": 45,
"styleRecalcDurationMs": 15.962999999999997,
"layouts": 0,
"layoutDurationMs": 0,
"taskDurationMs": 833.8319999999999,
"heapDeltaBytes": -10339640,
"heapUsedBytes": 58525836,
"domNodes": -260,
"jsHeapTotalBytes": 6283264,
"scriptDurationMs": 162.039,
"eventListeners": -174,
"totalBlockingTimeMs": 0,
"frameDurationMs": 16.666666666666668,
"p95FrameDurationMs": 16.700000000000728
},
{
"name": "canvas-mouse-sweep",
"durationMs": 1759.9609999999757,
"styleRecalcs": 46,
"styleRecalcDurationMs": 14.776000000000002,
"layouts": 0,
"layoutDurationMs": 0,
"taskDurationMs": 781.201,
"heapDeltaBytes": -7411924,
"heapUsedBytes": 61571052,
"domNodes": -259,
"jsHeapTotalBytes": 4972544,
"scriptDurationMs": 157.279,
"eventListeners": -174,
"totalBlockingTimeMs": 0,
"frameDurationMs": 16.670000000000012,
"p95FrameDurationMs": 16.800000000000182
},
{
"name": "canvas-zoom-sweep",
"durationMs": 1722.7799999999434,
"styleRecalcs": 32,
"styleRecalcDurationMs": 16.205,
"layouts": 6,
"layoutDurationMs": 1.1700000000000002,
"taskDurationMs": 385.815,
"heapDeltaBytes": 2989352,
"heapUsedBytes": 71676668,
"domNodes": 78,
"jsHeapTotalBytes": 4980736,
"scriptDurationMs": 14.272999999999998,
"eventListeners": 15,
"totalBlockingTimeMs": 0,
"frameDurationMs": 16.666666666666668,
"p95FrameDurationMs": 16.700000000000728
},
{
"name": "canvas-zoom-sweep",
"durationMs": 1703.4270000000333,
"styleRecalcs": 29,
"styleRecalcDurationMs": 14.587,
"layouts": 6,
"layoutDurationMs": 1.009,
"taskDurationMs": 256.93399999999997,
"heapDeltaBytes": 2992848,
"heapUsedBytes": 71262056,
"domNodes": 77,
"jsHeapTotalBytes": 5242880,
"scriptDurationMs": 10.673000000000002,
"eventListeners": 15,
"totalBlockingTimeMs": 0,
"frameDurationMs": 16.666666666666668,
"p95FrameDurationMs": 16.700000000000728
},
{
"name": "dom-widget-clipping",
"durationMs": 631.8499999999858,
"styleRecalcs": 14,
"styleRecalcDurationMs": 9.335999999999999,
"layouts": 0,
"layoutDurationMs": 0,
"taskDurationMs": 451.009,
"heapDeltaBytes": -8012780,
"heapUsedBytes": 61002416,
"domNodes": 20,
"jsHeapTotalBytes": 6553600,
"scriptDurationMs": 169.549,
"eventListeners": 0,
"totalBlockingTimeMs": 0,
"frameDurationMs": 16.666666666666668,
"p95FrameDurationMs": 16.700000000000728
},
{
"name": "dom-widget-clipping",
"durationMs": 631.3379999999142,
"styleRecalcs": 14,
"styleRecalcDurationMs": 8.529,
"layouts": 0,
"layoutDurationMs": 0,
"taskDurationMs": 443.96500000000003,
"heapDeltaBytes": 18146512,
"heapUsedBytes": 87003960,
"domNodes": 20,
"jsHeapTotalBytes": 5505024,
"scriptDurationMs": 164.206,
"eventListeners": 0,
"totalBlockingTimeMs": 0,
"frameDurationMs": 16.66333333333335,
"p95FrameDurationMs": 16.699999999999818
},
{
"name": "dom-widget-order-dense-n200-w200-i120",
"durationMs": 1179.636999999957,
"styleRecalcs": 0,
"styleRecalcDurationMs": 0,
"layouts": 0,
"layoutDurationMs": 0,
"taskDurationMs": 843.2189999999999,
"heapDeltaBytes": 39205404,
"heapUsedBytes": 107269280,
"domNodes": 0,
"jsHeapTotalBytes": 19034112,
"scriptDurationMs": 73.59999999999998,
"eventListeners": 0,
"totalBlockingTimeMs": 569,
"frameDurationMs": 16.666666666666668,
"p95FrameDurationMs": 16.700000000000728
},
{
"name": "dom-widget-order-dense-n200-w200-i120",
"durationMs": 1226.8260000000737,
"styleRecalcs": 0,
"styleRecalcDurationMs": 0,
"layouts": 0,
"layoutDurationMs": 0,
"taskDurationMs": 930.293,
"heapDeltaBytes": -18227560,
"heapUsedBytes": 73129704,
"domNodes": -296,
"jsHeapTotalBytes": 19529728,
"scriptDurationMs": 77.827,
"eventListeners": -217,
"totalBlockingTimeMs": 557,
"frameDurationMs": 16.666666666666668,
"p95FrameDurationMs": 16.700000000000728
},
{
"name": "dom-widget-order-sparse-n1000-w20-i120",
"durationMs": 843.6970000000201,
"styleRecalcs": 0,
"styleRecalcDurationMs": 0,
"layouts": 0,
"layoutDurationMs": 0,
"taskDurationMs": 470.0740000000003,
"heapDeltaBytes": 16160464,
"heapUsedBytes": 92238592,
"domNodes": 0,
"jsHeapTotalBytes": 0,
"scriptDurationMs": 163.997,
"eventListeners": 0,
"totalBlockingTimeMs": 229,
"frameDurationMs": 16.666666666666668,
"p95FrameDurationMs": 16.80000000000109
},
{
"name": "dom-widget-order-sparse-n1000-w20-i120",
"durationMs": 838.5330000000977,
"styleRecalcs": 0,
"styleRecalcDurationMs": 0,
"layouts": 0,
"layoutDurationMs": 0,
"taskDurationMs": 479.11000000000035,
"heapDeltaBytes": 15391176,
"heapUsedBytes": 92184916,
"domNodes": 0,
"jsHeapTotalBytes": 0,
"scriptDurationMs": 164.36700000000005,
"eventListeners": 0,
"totalBlockingTimeMs": 228,
"frameDurationMs": 16.66333333333332,
"p95FrameDurationMs": 16.799999999999272
},
{
"name": "extension-clean-combined-draw",
"durationMs": 10469.298999999977,
"styleRecalcs": 0,
"styleRecalcDurationMs": 0,
"layouts": 0,
"layoutDurationMs": 0,
"taskDurationMs": 9553.384000000002,
"heapDeltaBytes": 13629768,
"heapUsedBytes": 94515080,
"domNodes": -16,
"jsHeapTotalBytes": 4980736,
"scriptDurationMs": 7978.888999999999,
"eventListeners": 0,
"totalBlockingTimeMs": 1230,
"frameDurationMs": 19.996666666666716,
"p95FrameDurationMs": 50
},
{
"name": "extension-clean-combined-draw",
"durationMs": 10217.951999999968,
"styleRecalcs": 0,
"styleRecalcDurationMs": 0,
"layouts": 0,
"layoutDurationMs": 0,
"taskDurationMs": 9290.443,
"heapDeltaBytes": -4448932,
"heapUsedBytes": 71740992,
"domNodes": -14,
"jsHeapTotalBytes": 12320768,
"scriptDurationMs": 8036.190999999999,
"eventListeners": 0,
"totalBlockingTimeMs": 1476,
"frameDurationMs": 20,
"p95FrameDurationMs": 49.899999999999636
},
{
"name": "extension-rgthree-prototype-draw",
"durationMs": 10481.701999999928,
"styleRecalcs": 0,
"styleRecalcDurationMs": 0,
"layouts": 0,
"layoutDurationMs": 0,
"taskDurationMs": 9583.484999999999,
"heapDeltaBytes": 5016016,
"heapUsedBytes": 86141472,
"domNodes": -10,
"jsHeapTotalBytes": 5242880,
"scriptDurationMs": 8040.584000000001,
"eventListeners": 0,
"totalBlockingTimeMs": 1147,
"frameDurationMs": 19.996666666666716,
"p95FrameDurationMs": 50
},
{
"name": "extension-rgthree-prototype-draw",
"durationMs": 11231.040000000008,
"styleRecalcs": 0,
"styleRecalcDurationMs": 0,
"layouts": 0,
"layoutDurationMs": 0,
"taskDurationMs": 10269.483,
"heapDeltaBytes": 17645008,
"heapUsedBytes": 97308212,
"domNodes": -270,
"jsHeapTotalBytes": 8380416,
"scriptDurationMs": 8717.247000000001,
"eventListeners": -174,
"totalBlockingTimeMs": 1725,
"frameDurationMs": 18.890000000000025,
"p95FrameDurationMs": 50
},
{
"name": "large-graph-idle",
"durationMs": 2022.183000000041,
"styleRecalcs": 8,
"styleRecalcDurationMs": 8.767000000000001,
"layouts": 0,
"layoutDurationMs": 0,
"taskDurationMs": 545.3309999999999,
"heapDeltaBytes": -16413192,
"heapUsedBytes": 72959388,
"domNodes": -243,
"jsHeapTotalBytes": -495616,
"scriptDurationMs": 38.482000000000006,
"eventListeners": -152,
"totalBlockingTimeMs": 0,
"frameDurationMs": 16.66333333333332,
"p95FrameDurationMs": 16.799999999999272
},
{
"name": "large-graph-idle",
"durationMs": 2023.5900000000129,
"styleRecalcs": 10,
"styleRecalcDurationMs": 8.727999999999998,
"layouts": 0,
"layoutDurationMs": 0,
"taskDurationMs": 555.677,
"heapDeltaBytes": -24059252,
"heapUsedBytes": 64990880,
"domNodes": -256,
"jsHeapTotalBytes": 4485120,
"scriptDurationMs": 34.13499999999999,
"eventListeners": -174,
"totalBlockingTimeMs": 0,
"frameDurationMs": 16.666666666666668,
"p95FrameDurationMs": 16.700000000000728
},
{
"name": "large-graph-pan",
"durationMs": 2287.427999999977,
"styleRecalcs": 68,
"styleRecalcDurationMs": 13.096,
"layouts": 0,
"layoutDurationMs": 0,
"taskDurationMs": 1640.484,
"heapDeltaBytes": -23523304,
"heapUsedBytes": 66118940,
"domNodes": -248,
"jsHeapTotalBytes": -757760,
"scriptDurationMs": 869.208,
"eventListeners": -152,
"totalBlockingTimeMs": 0,
"frameDurationMs": 16.66333333333332,
"p95FrameDurationMs": 16.699999999999818
},
{
"name": "large-graph-pan",
"durationMs": 2260.02799999992,
"styleRecalcs": 69,
"styleRecalcDurationMs": 17.170000000000005,
"layouts": 0,
"layoutDurationMs": 0,
"taskDurationMs": 1628.5740000000003,
"heapDeltaBytes": -1421296,
"heapUsedBytes": 87660296,
"domNodes": -252,
"jsHeapTotalBytes": 1564672,
"scriptDurationMs": 856.888,
"eventListeners": -152,
"totalBlockingTimeMs": 0,
"frameDurationMs": 16.666666666666668,
"p95FrameDurationMs": 16.800000000000182
},
{
"name": "large-graph-zoom",
"durationMs": 3537.84799999994,
"styleRecalcs": 71,
"styleRecalcDurationMs": 16.493000000000002,
"layouts": 60,
"layoutDurationMs": 13.592,
"taskDurationMs": 1678.3860000000002,
"heapDeltaBytes": 5722168,
"heapUsedBytes": 97399764,
"domNodes": 36,
"jsHeapTotalBytes": 5541888,
"scriptDurationMs": 807.2900000000001,
"eventListeners": 0,
"totalBlockingTimeMs": 0,
"frameDurationMs": 16.66333333333332,
"p95FrameDurationMs": 16.80000000000109
},
{
"name": "large-graph-zoom",
"durationMs": 3539.549999999963,
"styleRecalcs": 71,
"styleRecalcDurationMs": 16.407,
"layouts": 60,
"layoutDurationMs": 13.350000000000001,
"taskDurationMs": 1684.2939999999999,
"heapDeltaBytes": 7369332,
"heapUsedBytes": 98440864,
"domNodes": 36,
"jsHeapTotalBytes": 5804032,
"scriptDurationMs": 815.354,
"eventListeners": 0,
"totalBlockingTimeMs": 0,
"frameDurationMs": 16.666666666666668,
"p95FrameDurationMs": 16.700000000000728
},
{
"name": "legacy-node-drag",
"durationMs": 3113.344999999981,
"styleRecalcs": 43,
"styleRecalcDurationMs": 7.341,
"layouts": 0,
"layoutDurationMs": 0,
"taskDurationMs": 1875.657,
"heapDeltaBytes": 8707324,
"heapUsedBytes": 74325780,
"domNodes": 6,
"jsHeapTotalBytes": 225280,
"scriptDurationMs": 1004.7470000000001,
"eventListeners": 180,
"totalBlockingTimeMs": 0,
"frameDurationMs": 16.66333333333332,
"p95FrameDurationMs": 16.80000000000109
},
{
"name": "legacy-node-drag",
"durationMs": 3060.4849999999715,
"styleRecalcs": 47,
"styleRecalcDurationMs": 10.097999999999999,
"layouts": 0,
"layoutDurationMs": 0,
"taskDurationMs": 1771.17,
"heapDeltaBytes": -18187620,
"heapUsedBytes": 74894676,
"domNodes": 14,
"jsHeapTotalBytes": 6590464,
"scriptDurationMs": 1025.1399999999999,
"eventListeners": 180,
"totalBlockingTimeMs": 0,
"frameDurationMs": 16.66333333333332,
"p95FrameDurationMs": 16.700000000000728
},
{
"name": "minimap-idle",
"durationMs": 2029.9019999999928,
"styleRecalcs": 6,
"styleRecalcDurationMs": 6.508,
"layouts": 0,
"layoutDurationMs": 0,
"taskDurationMs": 589.7,
"heapDeltaBytes": -17283764,
"heapUsedBytes": 73535992,
"domNodes": -250,
"jsHeapTotalBytes": -495616,
"scriptDurationMs": 35.181999999999995,
"eventListeners": -152,
"totalBlockingTimeMs": 0,
"frameDurationMs": 16.666666666666668,
"p95FrameDurationMs": 16.800000000000182
},
{
"name": "minimap-idle",
"durationMs": 2007.9060000000482,
"styleRecalcs": 6,
"styleRecalcDurationMs": 5.865000000000002,
"layouts": 0,
"layoutDurationMs": 0,
"taskDurationMs": 489.65999999999985,
"heapDeltaBytes": 10191192,
"heapUsedBytes": 74187560,
"domNodes": -247,
"jsHeapTotalBytes": 0,
"scriptDurationMs": 35.486999999999995,
"eventListeners": -152,
"totalBlockingTimeMs": 0,
"frameDurationMs": 16.666666666666668,
"p95FrameDurationMs": 16.700000000000728
},
{
"name": "subgraph-dom-widget-clipping",
"durationMs": 589.3800000000056,
"styleRecalcs": 47,
"styleRecalcDurationMs": 11.264999999999997,
"layouts": 0,
"layoutDurationMs": 0,
"taskDurationMs": 417.1769999999999,
"heapDeltaBytes": 13145128,
"heapUsedBytes": 83846660,
"domNodes": 20,
"jsHeapTotalBytes": 5242880,
"scriptDurationMs": 148.101,
"eventListeners": 6,
"totalBlockingTimeMs": 0,
"frameDurationMs": 16.666666666666668,
"p95FrameDurationMs": 16.700000000000728
},
{
"name": "subgraph-dom-widget-clipping",
"durationMs": 591.1849999999959,
"styleRecalcs": 47,
"styleRecalcDurationMs": 11.706,
"layouts": 0,
"layoutDurationMs": 0,
"taskDurationMs": 404.112,
"heapDeltaBytes": 12812292,
"heapUsedBytes": 83775528,
"domNodes": 20,
"jsHeapTotalBytes": 4980736,
"scriptDurationMs": 144.85999999999999,
"eventListeners": 6,
"totalBlockingTimeMs": 0,
"frameDurationMs": 16.666666666666668,
"p95FrameDurationMs": 16.700000000000728
},
{
"name": "subgraph-idle",
"durationMs": 1990.3929999999264,
"styleRecalcs": 11,
"styleRecalcDurationMs": 8.660999999999998,
"layouts": 0,
"layoutDurationMs": 0,
"taskDurationMs": 357.86300000000006,
"heapDeltaBytes": 10090368,
"heapUsedBytes": 80843580,
"domNodes": -275,
"jsHeapTotalBytes": 4972544,
"scriptDurationMs": 8.031999999999998,
"eventListeners": -198,
"totalBlockingTimeMs": 0,
"frameDurationMs": 16.666666666666668,
"p95FrameDurationMs": 16.700000000000728
},
{
"name": "subgraph-idle",
"durationMs": 2001.0969999999588,
"styleRecalcs": 10,
"styleRecalcDurationMs": 7.3210000000000015,
"layouts": 0,
"layoutDurationMs": 0,
"taskDurationMs": 449.12299999999993,
"heapDeltaBytes": -5562712,
"heapUsedBytes": 58639672,
"domNodes": -276,
"jsHeapTotalBytes": 3923968,
"scriptDurationMs": 7.712,
"eventListeners": -220,
"totalBlockingTimeMs": 0,
"frameDurationMs": 16.66333333333335,
"p95FrameDurationMs": 16.800000000000182
},
{
"name": "subgraph-mouse-sweep",
"durationMs": 1708.056000000056,
"styleRecalcs": 46,
"styleRecalcDurationMs": 16.083,
"layouts": 0,
"layoutDurationMs": 0,
"taskDurationMs": 733.109,
"heapDeltaBytes": 3392096,
"heapUsedBytes": 74040232,
"domNodes": -276,
"jsHeapTotalBytes": 6021120,
"scriptDurationMs": 103.95500000000001,
"eventListeners": -198,
"totalBlockingTimeMs": 0,
"frameDurationMs": 16.666666666666668,
"p95FrameDurationMs": 16.800000000000182
},
{
"name": "subgraph-mouse-sweep",
"durationMs": 1687.1419999999944,
"styleRecalcs": 46,
"styleRecalcDurationMs": 14.683,
"layouts": 0,
"layoutDurationMs": 0,
"taskDurationMs": 642.043,
"heapDeltaBytes": 9927792,
"heapUsedBytes": 80373352,
"domNodes": -274,
"jsHeapTotalBytes": 4972544,
"scriptDurationMs": 106.08999999999999,
"eventListeners": -198,
"totalBlockingTimeMs": 0,
"frameDurationMs": 16.666666666666668,
"p95FrameDurationMs": 16.800000000000182
},
{
"name": "subgraph-transition-enter",
"durationMs": 1401.685999999927,
"styleRecalcs": 17,
"styleRecalcDurationMs": 30.15199999999999,
"layouts": 5,
"layoutDurationMs": 13.387,
"taskDurationMs": 970.3139999999999,
"heapDeltaBytes": -8290988,
"heapUsedBytes": 85321124,
"domNodes": 13667,
"jsHeapTotalBytes": 12582912,
"scriptDurationMs": 42.288999999999994,
"eventListeners": 2383,
"totalBlockingTimeMs": 149,
"frameDurationMs": 16.670000000000012,
"p95FrameDurationMs": 16.700000000000728
},
{
"name": "viewport-pan-sweep",
"durationMs": 8376.29800000002,
"styleRecalcs": 251,
"styleRecalcDurationMs": 38.599,
"layouts": 0,
"layoutDurationMs": 0,
"taskDurationMs": 5812.714,
"heapDeltaBytes": -420600,
"heapUsedBytes": 88758112,
"domNodes": -256,
"jsHeapTotalBytes": 4972544,
"scriptDurationMs": 2890.643,
"eventListeners": -174,
"totalBlockingTimeMs": 0,
"frameDurationMs": 16.66333333333338,
"p95FrameDurationMs": 16.700000000000728
},
{
"name": "viewport-pan-sweep",
"durationMs": 8310.443999999961,
"styleRecalcs": 250,
"styleRecalcDurationMs": 38.949,
"layouts": 0,
"layoutDurationMs": 0,
"taskDurationMs": 5603.428,
"heapDeltaBytes": -211584,
"heapUsedBytes": 89290300,
"domNodes": -256,
"jsHeapTotalBytes": 5234688,
"scriptDurationMs": 2777.6369999999997,
"eventListeners": -174,
"totalBlockingTimeMs": 0,
"frameDurationMs": 16.666666666666668,
"p95FrameDurationMs": 16.80000000000109
},
{
"name": "vue-large-graph-idle",
"durationMs": 17276.042999999958,
"styleRecalcs": 0,
"styleRecalcDurationMs": 0,
"layouts": 0,
"layoutDurationMs": 0,
"taskDurationMs": 16822.674,
"heapDeltaBytes": -62658528,
"heapUsedBytes": 171396152,
"domNodes": -8061,
"jsHeapTotalBytes": -11022336,
"scriptDurationMs": 273.256,
"eventListeners": -14928,
"totalBlockingTimeMs": 0,
"frameDurationMs": 17.776666666666642,
"p95FrameDurationMs": 16.799999999999272
},
{
"name": "vue-large-graph-idle",
"durationMs": 16946.48100000006,
"styleRecalcs": 0,
"styleRecalcDurationMs": 0,
"layouts": 0,
"layoutDurationMs": 0,
"taskDurationMs": 16545.027000000002,
"heapDeltaBytes": -68929852,
"heapUsedBytes": 165433036,
"domNodes": -8061,
"jsHeapTotalBytes": -10760192,
"scriptDurationMs": 259.87100000000004,
"eventListeners": -14928,
"totalBlockingTimeMs": 0,
"frameDurationMs": 17.780000000000047,
"p95FrameDurationMs": 16.700000000000728
},
{
"name": "vue-large-graph-pan",
"durationMs": 20719.459000000028,
"styleRecalcs": 147,
"styleRecalcDurationMs": 19.40799999999998,
"layouts": 0,
"layoutDurationMs": 0,
"taskDurationMs": 20587.707000000002,
"heapDeltaBytes": -34266172,
"heapUsedBytes": 193412404,
"domNodes": -8061,
"jsHeapTotalBytes": -21241856,
"scriptDurationMs": 972.505,
"eventListeners": -14928,
"totalBlockingTimeMs": 60,
"frameDurationMs": 18.886666666666617,
"p95FrameDurationMs": 16.799999999999272
},
{
"name": "vue-large-graph-pan",
"durationMs": 20162.31200000004,
"styleRecalcs": 157,
"styleRecalcDurationMs": 16.679000000000027,
"layouts": 0,
"layoutDurationMs": 0,
"taskDurationMs": 19885.662,
"heapDeltaBytes": -58669004,
"heapUsedBytes": 181363164,
"domNodes": -8061,
"jsHeapTotalBytes": -17018880,
"scriptDurationMs": 905.7760000000001,
"eventListeners": -14928,
"totalBlockingTimeMs": 11,
"frameDurationMs": 17.776666666666642,
"p95FrameDurationMs": 16.80000000000291
},
{
"name": "vue-node-resize-workload",
"durationMs": 6199.474000000009,
"styleRecalcs": 20,
"styleRecalcDurationMs": 83.238,
"layouts": 20,
"layoutDurationMs": 27.027999999999995,
"taskDurationMs": 6136.38,
"heapDeltaBytes": -38040288,
"heapUsedBytes": 183258032,
"domNodes": -8066,
"jsHeapTotalBytes": 12120064,
"scriptDurationMs": 112.471,
"eventListeners": -14928,
"totalBlockingTimeMs": 478,
"frameDurationMs": 17.776666666666642,
"p95FrameDurationMs": 16.80000000000291
},
{
"name": "vue-node-resize-workload",
"durationMs": 6204.146000000037,
"styleRecalcs": 20,
"styleRecalcDurationMs": 78.98300000000003,
"layouts": 20,
"layoutDurationMs": 24.617,
"taskDurationMs": 6128.956,
"heapDeltaBytes": -38519884,
"heapUsedBytes": 196605368,
"domNodes": -8061,
"jsHeapTotalBytes": 16130048,
"scriptDurationMs": 110.45600000000002,
"eventListeners": -14928,
"totalBlockingTimeMs": 470,
"frameDurationMs": 17.776666666666642,
"p95FrameDurationMs": 16.700000000000728
},
{
"name": "vue-zoom-out-idle",
"durationMs": 8451.863000000003,
"styleRecalcs": 0,
"styleRecalcDurationMs": 0,
"layouts": 0,
"layoutDurationMs": 0,
"taskDurationMs": 8319.403999999999,
"heapDeltaBytes": -720168,
"heapUsedBytes": 182395832,
"domNodes": 0,
"jsHeapTotalBytes": -14155776,
"scriptDurationMs": 152.21300000000005,
"eventListeners": 0,
"totalBlockingTimeMs": 0,
"frameDurationMs": 17.780000000000047,
"p95FrameDurationMs": 16.80000000000291
},
{
"name": "vue-zoom-out-idle",
"durationMs": 8345.555000000104,
"styleRecalcs": 0,
"styleRecalcDurationMs": 0,
"layouts": 0,
"layoutDurationMs": 0,
"taskDurationMs": 8080.8690000000015,
"heapDeltaBytes": 7036164,
"heapUsedBytes": 191859092,
"domNodes": 0,
"jsHeapTotalBytes": -14155776,
"scriptDurationMs": 119.66799999999999,
"eventListeners": 0,
"totalBlockingTimeMs": 0,
"frameDurationMs": 17.773333333333238,
"p95FrameDurationMs": 16.80000000000291
},
{
"name": "workflow-execution",
"durationMs": 120.67600000000311,
"styleRecalcs": 7,
"styleRecalcDurationMs": 15.456,
"layouts": 2,
"layoutDurationMs": 1.5270000000000001,
"taskDurationMs": 88.17,
"heapDeltaBytes": 3091784,
"heapUsedBytes": 69713572,
"domNodes": 125,
"jsHeapTotalBytes": 262144,
"scriptDurationMs": 9.525,
"eventListeners": 17,
"totalBlockingTimeMs": 0,
"frameDurationMs": 16.666666666666668,
"p95FrameDurationMs": 16.700000000000728
},
{
"name": "workflow-execution",
"durationMs": 471.79800000003524,
"styleRecalcs": 20,
"styleRecalcDurationMs": 27.773999999999997,
"layouts": 4,
"layoutDurationMs": 1.8120000000000003,
"taskDurationMs": 133.32099999999997,
"heapDeltaBytes": 5102852,
"heapUsedBytes": 71803212,
"domNodes": 146,
"jsHeapTotalBytes": 262144,
"scriptDurationMs": 11.069,
"eventListeners": 81,
"totalBlockingTimeMs": 0,
"frameDurationMs": 16.666666666666668,
"p95FrameDurationMs": 16.800000000000182
}
]
}🔬 E2E Coverage
Top 10 files by uncovered lines
|
This comment has been minimized.
This comment has been minimized.
Playwright videos for new testsDownload the recorded videos and HTML report Recorded projects: cloud Recorded test titles:
|
Codecov Report❌ Patch coverage is
@@ Coverage Diff @@
## cloud/1.55 #19889 +/- ##
=============================================
Coverage ? 84.94%
=============================================
Files ? 2327
Lines ? 133215
Branches ? 34336
=============================================
Hits ? 113153
Misses ? 19700
Partials ? 362
Flags with carried forward coverage won't be shown. Click here to find out more.
🚀 New features to boost your workflow:
|
There was a problem hiding this comment.
🔍 Cursor Review — Consolidated panel
Triggered by @christian-byrne.
Found 9 finding(s).
| Severity | Count |
|---|---|
| 🟡 Medium | 4 |
| 🟢 Low | 5 |
Panel: 7/8 reviewers contributed findings.
Reviewers that did not contribute: gpt-5.6-sol-max:adversarial (error)
Release-gate pass at
|
| # | Where | Severity | Disposition |
|---|---|---|---|
| 1 | docOpMinter.ts widgetOwnerOf |
🟡 ×5 reviewers | Route to #19738. The strongest of the nine — the doc comment and the body disagree, and the comment contradicts owningGraphIdOf's stated invariant in the same file. Either fix is a production change and belongs upstream. |
| 2 | store key in isValueWidgetWrite |
🟡 ×3 | Route to #19738, premise narrowed. The "subgraph-shaped intent" half does not reach production: widgetValueStore.observeValue is bound with the id from registerWidget, and widgets register under resolveNodeRootGraphId = node.graph?.rootGraph.id, so a live set_widget intent always carries a root graph id. graphId: subgraph.id is a fixture shape. The moved-canvas half is real and is what 345f604f was reviewed for. |
| 3 | virtual-node value widgets | 🟡 ×1 | Already dispositioned on #19738 (resolved Low: "wireNodeSnapshot bypasses valueWidgetsOnly entirely when node.isVirtualNode"). It describes an incompleteness, not a regression: a Note's text write mints today and still mints after this change. Not a reason to hold the carrier. |
| 5 6 9 | docOpMinter.ts |
🟢 | Route to #19738. Production changes; cannot land here first. |
| 7 | docOpMinter.test.ts:848 |
🟢 | Route to #19738. A real vacuity claim about fails closed for stale writes from a different rendered graph; the fixture is identical on both branches. |
| 4 8 | the two black-box specs | 🟡 / 🟢 | Route to #19738. The spec files are byte-identical across the two carriers. [8]'s settlement-fence point is the same class the toEqual(['applied']) barrier in 260083ca was added for; [4] is a frame-shape realism gap worth taking. |
Two of eight panel jobs (gpt-5.6-sol-max, adversarial and edge-case) were cancelled at the 30-minute step timeout, not failed on findings. The other six completed. cursor-review is not a required context here.
Gate state at this head
- Required contexts on
cloud/1.55are exactlylint-and-format,test,e2e-status.lint-and-format✅,e2e-status✅,test❌ — from the branch, not this PR; see test(agent-crdt): unblock cloud/1.55 — restore the seal-reentry expectations #19890. MERGEABLE, no conflict withcloud/1.55.- Ordering unchanged: this must not merge before fix(agent-crdt): skip non-value widget graph intents #19738.
DrJKL
left a comment
There was a problem hiding this comment.
Verdict: faithful backport of #19738 with no blocking defects; hold until #19738 merges.
I reviewed 4a515780. It merges cloud/1.55 (through 1d864fd5) into 260083ca and leaves the PR's diff against the base unchanged: the five files are byte-identical between the two heads.
I compared this branch with #19738 at 56089f4e. The added and removed lines in docOpMinter.ts, docOpMinter.test.ts and both specs are identical (git patch-id --stable matches). The production file differs from #19738's only by the pre-existing telemetry surface line. The fixture conflict resolution ports ExecutionHostFrame, sendExecution and sendExecutionBinary with text identical to main's fixture, and both specs use them. At 4a515780, docOpMinter.test.ts, opSender.test.ts and widgetValueStore.test.ts pass 183/183, and pnpm typecheck and pnpm typecheck:browser are clean. At 260083ca, with only docOpMinter.ts reverted to cloud/1.55, the unit file fails 9/42, matching the description. eslint and oxfmt are clean on the five files, and CI ran all four new Playwright tests green at 260083ca.
Every inline comment is P3. All of them apply verbatim to #19738, so they can land there first and be re-picked here, which keeps the two branches identical.
Pre-merge blockers: none in the code. Merge order still applies: this should not land before #19738. The required test context was red at 260083ca on two opSender.test.ts cases this PR does not touch; the merge brings in #19907, and that file now passes locally (CI for 4a515780 was still running when I posted).
Findings without an anchorable line (2)
- P3, PR description, "Branch health". The paragraph still says #19890 must land before this PR can go green. #19890 is open, but #19907 ("correct cloud 1.55 sender seal expectations",
aba3d1c085) is oncloud/1.55, and your merge in4a515780brought it in;opSender.test.tspasses 87/87 at this head. I'd update the paragraph so nobody waits on #19890. - P3 Nit, PR description length.
AGENTS.mdasks for an extremely concise description that follows the template. The body carries a correction to an earlier revision, a telemetry sample, and verification narratives repeated from #19738. The parity proof, the fixture conflict, the merge order, and the red/green results are what a reviewer needs. The closing line names FE-3161 rather than aFixes #nissue, which is fine if no GitHub issue exists.
Existing cursor-review findings, confirmed or contested (9)
widgetOwnerOfdocblock vs body (line 174): confirmed for the docblock, contested for the runtime claim. The body ignoresevent.name, so the docblock is wrong. The collision it describes does not occur:LGraph.collectReservedNodeIdsreserves node ids across a root graph and its subgraphs, and a scratch root/interior collision case resolved correctly. My inline suggestion at lines 159-168 replaces the docblock.- Store key from the resolved owner vs the intent's graph id (line 209): contested. The only
set_widgetemitter iswidgetValueStore.observeValue, which emits the graph id the widget registered under. All threeregisterWidgetcallers (BaseWidget.setNodeId,SubgraphNode,promotionUtils) passrootGraph.id, anddeps.getGraph()isapp.rootGraph. Whenevent.graphId === graph.idthe owner'srootGraph.idequalsevent.graphId. Otherwise the owner is null and the code usesevent.graphId. The key is the registration key in both production cases. The mismatch needs a subgraph-shaped intent, which only the tests emit (see my inline comment at lines 182-186). - Virtual-node value widgets still mint (line 216): contested as a finding against this PR. True, and reproduced with a scratch
Notetext write, but it mints the same oncloud/1.55. It is the C1 opaque-widget half the description scopes out, not a regression here. - Progress spec uses only the legacy
progress_textlayout (line 107): confirmed, P3. The spec never negotiatessupports_progress_text_metadataor sendsexecution_start, so the prompt-id branch ofapi.tsis not exercised. Both branches end in the same widget setter and the same minter gate, so the fix itself is still proven. A frame-parser regression on the metadata branch would not show up here. - Silent drop with no diagnostic (line 407): contested.
observeValueemits only aftergetWidget(widgetId) === state, so every production intent has a store entry, and the!widget && !storedbranch is reachable only from synthetic intents such as thenodeId: 999test. The docblock that describes this branch contradicts itself; I left a suggestion at lines 201-203. 'serialize' in widgetwith an explicitundefined(line 152): confirmed, P3, not a regression. Reproduced with a scratch test. My inline suggestion at lines 151-152 switches to??.- Stale-graph test passes only because of
stored!.serialize = false(test line 848): confirmed for the name, contested for the fix. Deleting line 848 makes the test fail with a minted root op. A stale value write minting is the base behavior, and the bound doc is that graph, so dropping the line would leave a test that asserts the opposite of current behavior. I suggested a rename inline at line 841. - Toast and value assertions are not fenced on the page's receipt of
doc_ops_result(progress spec line 248, and the matching lines in both specs): confirmed, P3.judgeHumanOpspushes tohumanOutcomesbeforethis.send(result), so the host-side poll can finish first. The outcome assertion is still a sound oracle for "nothing was minted"; only the two user-visible assertions after it can pass early. The comment at progress spec 251-253 claims otherwise; my inline comment there covers it. findSubgraphByUuidreturns registered definitions with no live node (line 184): confirmed for the docblock. Behavior does not change: such a write reachesroutedWidgetOperation, finds no path, and is logged and dropped as on the base. The docblock suggestion is inline at line 177.
Checked and cleared (12)
- Narrowing
ExecutionHostFramein the fixture to the oneexecutedframe this branch sends: cleared. The ported text is identical tomain's, and narrowing it would make the next backport touching this fixture conflict again. isValueWidgetWritereturning the samefalsefor "non-value" and "unknown": cleared for the same reason as item 5 above. Production intents always have a store entry.- Deleting the boot-helper docblocks (preview spec 166-169, progress spec 136-139): cleared. They explain the positional
'apply'argument. - Deleting the
sendTextOutputdocblock (preview spec 215-219): cleared. It names theonExecuted→updateTextPreviewWidgetspath, which the body does not show. - Deleting the
progressTextFramedocblock (progress spec 107) and thestreamProgressdocblock (182): cleared. One documents the binary layout; the other says the frame reaches the widget's real setter. - Deleting progress spec comments at 223-224, 269-270, 303-304: cleared. They name
$$node-text-preview, explain why the first frame is sent, and explain why the projection is asserted. - Deleting the virtual-node button test (test line 792) as a duplicate of the generic button case: cleared. It pins the regression 4083772 removed (
isVirtualNode === true && widget.type === 'button'minted). The generic button test runs on a non-virtual node and would not catch that carve-out coming back. - Fallow audit against
origin/cloud/1.55: no issues in the five changed files. The 16 dev-dependency findings it reports are inherited and excluded by its gate.
Coverage
- Depth standard. Lanes: correctness, maintainability, idiomaticity, tests (including the Playwright spec rulebook and helper inventory), documentation. No lane skipped.
- Checks at
4a515780:docOpMinter.test.ts,opSender.test.ts,widgetValueStore.test.ts183/183;pnpm typecheckandpnpm typecheck:browserexit 0. - Checks at
260083ca(same five files):docOpMinter.test.ts42/42; with the minter reverted tocloud/1.55, 9 failed; with mydocOpMinter.tsand test-file suggestions applied, 42/42 andvue-tscexit 0; eslint and oxfmt clean;playwright test --listshows the 4 new tests;opSender.test.ts2 failed there and 87/87 with the base tip's file. - Mutation runs: removing the button check fails 6 tests; removing the incremental store fallback fails 1; removing line 848 fails the stale-graph test.
- Not run here: the Playwright specs against a backend. I relied on CI (all four passed) and did not reproduce the description's black-box red run with the minter reverted.
| function isValueWidget( | ||
| widget: IBaseWidget | undefined, | ||
| stored?: { type: string; serialize?: boolean } | ||
| ): boolean { | ||
| if (!widget && !stored) return false | ||
| if (widgetType(widget, stored) === 'button') return false | ||
| return widgetSerialize(widget, stored) !== false | ||
| } | ||
|
|
||
| function widgetType( | ||
| widget: IBaseWidget | undefined, | ||
| stored?: { type: string } | ||
| ): string | undefined { | ||
| return widget ? widget.type : stored?.type | ||
| } |
There was a problem hiding this comment.
P3 Nit: widgetType is a one-use alias for a ternary, the shape docs/guidance/engineering.md rule 11 asks us not to add. Inlining it keeps the whole predicate in one place. widget?.type ?? stored?.type matches the current ternary because a live widget always has a string type. Ran the file with this and the widgetSerialize suggestion below applied: 42/42, vue-tsc clean.
| function isValueWidget( | |
| widget: IBaseWidget | undefined, | |
| stored?: { type: string; serialize?: boolean } | |
| ): boolean { | |
| if (!widget && !stored) return false | |
| if (widgetType(widget, stored) === 'button') return false | |
| return widgetSerialize(widget, stored) !== false | |
| } | |
| function widgetType( | |
| widget: IBaseWidget | undefined, | |
| stored?: { type: string } | |
| ): string | undefined { | |
| return widget ? widget.type : stored?.type | |
| } | |
| function isValueWidget( | |
| widget: IBaseWidget | undefined, | |
| stored?: { type: string; serialize?: boolean } | |
| ): boolean { | |
| if (!widget && !stored) return false | |
| if ((widget?.type ?? stored?.type) === 'button') return false | |
| return widgetSerialize(widget, stored) !== false | |
| } |
There was a problem hiding this comment.
Addressed in 191cc2a: widgetType and widgetSerialize are gone and isValueWidget computes type and serialize inline.
| if (!widget) return stored?.serialize | ||
| return 'serialize' in widget ? widget.serialize : stored?.serialize |
There was a problem hiding this comment.
P3: 'serialize' in widget treats an own serialize: undefined as authoritative, so a stored serialize: false is ignored and the write mints. I reproduced it with a scratch test: stored flag false, live serialize = undefined, one set_widget for steps minted. BaseWidget copies options with Object.assign, so an explicit undefined key can reach this point. The base had the same live-only behavior, so this is not a regression, but the store fallback this PR adds is meant to cover exactly this gap. ?? treats absent and undefined the same and removes the !widget branch. This is the cursor finding on this line, with a concrete fix.
| if (!widget) return stored?.serialize | |
| return 'serialize' in widget ? widget.serialize : stored?.serialize | |
| return widget?.serialize ?? stored?.serialize |
| /** | ||
| * The live node a widget intent names, or null when this canvas holds none. | ||
| * | ||
| * The widget store keys by ROOT graph id (`resolveNodeRootGraphId`), so an | ||
| * interior node's intent carries the root graph's id and the node id alone is | ||
| * ambiguous: ids are unique per graph, not per hierarchy, so a root node and a | ||
| * subgraph-interior node can share one. Prefer whichever node actually carries | ||
| * the named widget, and fall back to plain hierarchy order when none does — | ||
| * that fallback is the resolution the routing below has always used. | ||
| */ |
There was a problem hiding this comment.
P3: this docblock says the function prefers the node that carries event.name, but the body never reads event.name; it returns the first findNodeInHierarchy match. It also says ids are unique per graph, not per hierarchy, while owningGraphIdOf further down says they are unique across the hierarchy. The second claim is the true one: LGraph.collectReservedNodeIds reserves ids across a root and its subgraphs, so the collision this docblock guards against does not occur. A one-line description of what the function does avoids both problems.
| /** | |
| * The live node a widget intent names, or null when this canvas holds none. | |
| * | |
| * The widget store keys by ROOT graph id (`resolveNodeRootGraphId`), so an | |
| * interior node's intent carries the root graph's id and the node id alone is | |
| * ambiguous: ids are unique per graph, not per hierarchy, so a root node and a | |
| * subgraph-interior node can share one. Prefer whichever node actually carries | |
| * the named widget, and fall back to plain hierarchy order when none does — | |
| * that fallback is the resolution the routing below has always used. | |
| */ | |
| /** The live node a widget intent names, or null when this canvas holds none. */ |
There was a problem hiding this comment.
Addressed in 191cc2a: widgetOwnerOf and its docblock are gone; mintSetWidget calls findNodeInHierarchy directly, and owningGraphIdOf with its conflicting id claim was removed too.
| return eventGraph ? findNodeInHierarchy(eventGraph, event.nodeId) : null | ||
| } | ||
|
|
||
| /** The graph an intent names, when this canvas still renders it. */ |
There was a problem hiding this comment.
P3: findSubgraphByUuid reads the root graph's subgraphs registry, which also holds definitions with no live SubgraphNode, so this can return a graph the canvas does not render. Nothing breaks, because routedWidgetOperation then fails to find a path and logs and drops as the base did, but the docblock promises a stronger condition than the code checks.
| /** The graph an intent names, when this canvas still renders it. */ | |
| /** | |
| * The graph an intent names: the root graph, a registered subgraph | |
| * definition, or one reached through a live subgraph-node path. | |
| */ |
There was a problem hiding this comment.
Addressed in 191cc2a: the "still renders it" docblock on reachableIntentGraph was removed.
| ) | ||
| } | ||
|
|
||
| /** The rendered `seed` input on the node under test. */ |
There was a problem hiding this comment.
P3 Nit: this comment restates the line below it. The name already says it.
| /** The rendered `seed` input on the node under test. */ |
There was a problem hiding this comment.
Addressed in 191cc2a: seedField moved into the rig without the comment.
| // And the human was never told that an edit they did not make was | ||
| // refused and undone. The round trip above has settled, so a rejection | ||
| // toast would already have been rendered by now. |
There was a problem hiding this comment.
P3: the second sentence claims the round trip has settled on the page, but the barrier above is host-side. AgentFollowerHostSocket.judgeHumanOps pushes to humanOutcomes before this.send(result), so the poll can finish before the page has received or handled doc_ops_result. The toHaveCount(0) below can then pass before a rejection toast could exist. This is the cursor settlement finding at line 248. The comment at 235-241 describes the host-side ordering correctly. I'd drop this one (the first sentence restates the assertion) and, separately, fence the toast and value assertions on something the page observes.
| // And the human was never told that an edit they did not make was | |
| // refused and undone. The round trip above has settled, so a rejection | |
| // toast would already have been rendered by now. |
There was a problem hiding this comment.
Addressed in 191cc2a: the comment claiming the page round trip had settled is gone. The toast assertion is still fenced only host-side; I carried that as a P3 in the review.
| }) | ||
| await expect(rejectionToast).toHaveCount(0) | ||
|
|
||
| // The hand edit is still the value on screen, not reverted as collateral. |
There was a problem hiding this comment.
P3 Nit: this comment restates the line below it. The name already says it.
| // The hand edit is still the value on screen, not reverted as collateral. |
| } | ||
| ) | ||
|
|
||
| // The applier saw exactly the hand edit, and took it. |
There was a problem hiding this comment.
P3 Nit: this comment restates the line below it. The name already says it.
| // The applier saw exactly the hand edit, and took it. |
| // And the value the user typed is still the one on screen — not | ||
| // reverted as collateral of an op they never made. |
There was a problem hiding this comment.
P3 Nit: this comment restates the line below it. The name already says it.
| // And the value the user typed is still the one on screen — not | |
| // reverted as collateral of an op they never made. |
Carrier for **[FE-3161](https://linear.app/comfyorg/issue/FE-3161/agent-crdt-locally-minted-set-widget-ops-are-refused-by-the-doc-host) sub-cause C2**, the independent half of a production defect on `cloud-frontend-prod` `1.55.16`: a `set_widget` the user never made is minted from their own canvas, the doc host refuses it, and `revertRejectedOps` reverts it behind a "your edit was rejected" toast. `$$node-text-preview` is injected at run time by `useNodeProgressText` with `serialize: false` and `read_only: true`, and is registered in `widgetValueStore` — so every `progress_text` tick fired a `local` `set_widget` intent. The snapshot mint path already dropped it (`valueWidgetsOnly`); `mintSetWidget` applied no filter. Both paths now share one `isValueWidget` rule: a widget is a value widget unless it is a `button` or `serialize: false`. C1 (subgraph blueprint hosts, `opaque_widgets`) is **not** in scope here — it is gated on FE-3036 item 2. `applyOps` is abort-remainder, the minter enqueues a tick's intents together, and `revertRejectedOps` reverts every op in the batch that did not apply. An unsatisfiable ephemeral write minted in the same tick as a genuine hand edit therefore **took the hand edit down with it**. FE-3161 records that as a code path rather than a production observation; it is now reproduced end to end (see below). 1. **Every subgraph-interior widget edit was silently dropped.** The graph-scope validation resolved the intent with the root graph's own `getNodeById`, but the widget store keys by ROOT graph id, so an interior write arrives naming the root and that lookup misses it. This is the CodeRabbit finding on [`docOpMinter.ts:198`](#19738 (comment)), confirmed by measurement. The branch's own test for the behaviour had been inverted to assert `minted` is empty, and its replacement ("ID-colliding") does not collide — the id allocator is hierarchy-unique, so the fixture built two distinct ids and the case passed vacuously. Resolution now goes through the hierarchy, which is what `owningGraphIdOf` and the routing have always done. The original assertion is restored. 2. **`node.serialize_widgets` was gating the mint.** `SubgraphNode` never sets it, so that gate suppressed every blueprint-host widget edit — "fixing" C1 by trading a visible rejection for an invisible divergence between canvas and document, and erasing the Sentry signal FE-3036's fix has to be verified against. Removed; only the widget's own flags decide. Unit — `src/workbench/extensions/agent/crdt/docOpMinter.test.ts` (42 pass; **9 fail against `origin/main`'s `docOpMinter.ts`**), including a new case that runs the real applier over the minted batch: > `does not let an ephemeral widget write roll a hand edit back with it` — a `serialize: false` preview write and a `steps` edit in one tick; only the edit is minted, and `applyMinted` returns `['applied']`. Black-box — `browser_tests/tests/agent/agentProgressTextWidgetNotMinted.spec.ts` (2 pass; **both fail against `origin/main`**). Real `progress_text` binary frames, and `AgentFollowerHostSocket` in `apply` mode judges each batch with the real comfy-multi-player applier, so rejection codes are the library's: - streaming progress mints nothing and raises no rejection toast — on `main` the toast appears; - a hand edit minted in the same tick as a progress write reaches the document and stays on canvas — on `main` the outcomes are `['rejected','rejected','rejected']` where one `['applied']` is expected. `sendExecutionBinary` was the harness gap this needed: the fixture could emit JSON execution frames but not the binary ones `api.ts` decodes from an `ArrayBuffer`. - `pnpm test:unit src/workbench/extensions/agent src/stores/widgetValueStore.test.ts src/utils/graphTraversalUtil.test.ts` — 135 files, 2710 passed - `pnpm exec playwright test browser_tests/tests/agent/agentProgressTextWidgetNotMinted.spec.ts` — 2 passed - `pnpm typecheck`, `oxfmt --check`, `oxlint --type-aware`, `eslint` — clean on the four changed files
4a51578 to
191cc2a
Compare
DrJKL
left a comment
There was a problem hiding this comment.
Verdict: ready to approve; only P3 items remain, and the required test and e2e-status checks were still running when I reviewed 191cc2a67f.
#19738 merged as 9fdb803, and this head is now a single cherry-pick of it onto cloud/1.55. I compared the two commits file by file. The added and removed lines in docOpMinter.ts, docOpMinter.test.ts, both specs and the new agentNonValueWidgetRig.ts match 9fdb803 exactly. The specs and the rig are byte-identical files. The only difference is agentFollowerHostSocket.ts, where this branch still needs the ExecutionHostFrame type and sendExecution that main already had. That port no longer matches main's text; I left two inline nits. With only docOpMinter.ts reverted to cloud/1.55, the unit file fails 11 of 45 at this head, all in the new cases. With this head's minter it passes.
Twenty-four of my round-one threads are addressed by this head, and I replied in each. The six inline comments here are two carried items and four new ones, all P3.
Pre-merge blockers: none.
Still open without a new inline comment (4)
- P3, subgraph branches of
reachableIntentGraph(round 1 thread). Unchanged. Productionset_widgetintents carry a root graph id, so thefindSubgraphByUuidandfindSubgraphNodePathByIdbranches only run for the subgraph-shaped intents the tests emit. I'm fine leaving it, since it now matchesmain. - P3, legacy
progress_textlayout only (cursor finding at progress spec 107). Unchanged. The spec never negotiatessupports_progress_text_metadataor sendsexecution_start, so it covers the legacy parser branch only. The minter gate is the same either way. - P3, toast and value assertions fenced only host-side (cursor finding at progress spec 248). The misleading comment is gone, but
judgeHumanOpsstill pushes tohumanOutcomesbeforethis.send(result). So thetoHaveCount(0)in theno rejection toaststeps of both specs can pass before the page has handleddoc_ops_result. The['applied']outcome assertion is still a sound check that nothing else was minted. - P3, PR description is stale for this head. It still describes a "cherry-pick of the 15-commit stack" and says this "must not merge before #19738", but #19738 has merged and this is a single cherry-pick of 9fdb803. It also cites 4 Playwright tests and 9 of 42 unit failures; at this head there are 3 tests and 11 of 45. The "Branch health" paragraph still points at #19890, though #19907 already fixed those
opSender.test.tscases on the base. The length and template nit from round 1 also stands.
Checked and cleared this round
routedWidgetOperationnow usesnode?.graph?.id ?? event.graphIdinstead of the removedowningGraphIdOf. It's equivalent: whenowneris null,findNodeInHierarchyreturned null in the old helper too, so both fall back to the intent's graph id.- The rig drops the explicit
'Comfy.VueNodes.Enabled': true.mockAgentBootinagentPanelFixture.tsturns it on from the@vue-nodestag, which both specs carry. - Dropping the preview spec's same-tick test leaves that case covered by the progress spec's same-tick test and the unit test
does not let an ephemeral widget write roll a hand edit back with it. The remaining preview test still covers theexecuted→onExecutedpath. - The virtual-node button row stays, as the regression row for 4083772.
Review record
- P3:
isValueWidgetexplicitserialize: undefinedoverrides the store flag. Round 1 thread: still open; carried inline atdocOpMinter.ts:139-140. - P3: subgraph branches of
reachableIntentGraphserve only fixture-shaped intents. Round 1 thread: still open; carried in the body. - P3 Nit:
widgetTypeone-line alias. Round 1 thread: addressed in 191cc2a (inlined); reply. - P3:
widgetOwnerOfdocblock promised name-based disambiguation. Round 1 thread: addressed in 191cc2a (function and docblock removed); reply. - P3:
reachableIntentGraphdocblock overstated "still renders". Round 1 thread: addressed in 191cc2a (docblock removed); reply. - P3:
isValueWidgetWritedocblock contradicted its fail-closed branch. Round 1 thread: addressed in 191cc2a (docblock rewritten); reply. - P3: table the non-value unit cases. Round 1 thread: addressed in 191cc2a (
it.fortable); reply. - P3:
stored!aftertoBeDefined()(test line 748). Round 1 thread: addressed in 191cc2a (assert.exists); reply. - P3:
stored!aftertoBeDefined()(test line 771). Round 1 thread: addressed in 191cc2a (assert.exists); reply. - P3: stale-graph test name overclaimed. Round 1 thread: addressed in 191cc2a (renamed); reply.
- P3:
stored!aftertoBeDefined()(test line 847). Round 1 thread: addressed in 191cc2a (assert.exists); reply. - P3: duplicated rig and helpers across the two specs. Round 1 thread: addressed in 191cc2a (
AgentNonValueWidgetRig); reply. - P3 Nit: redundant comment, preview spec 151. Round 1 thread: addressed in 191cc2a (moved to the rig); reply.
- P3 Nit: redundant comment, preview spec 235. Round 1 thread: addressed in 191cc2a (moved to the rig); reply.
- P3 Nit: redundant comment, preview spec 248. Round 1 thread: addressed in 191cc2a (moved to the rig); reply.
- P3 Nit: comment restating
sendTextOutput, preview spec 266-267. Round 1 thread: partly addressed; the remaining one-line comment is carried inline at preview spec 152. - P3 Nit: redundant comment, preview spec 291-292. Round 1 thread: addressed in 191cc2a (removed); reply.
- P3 Nit: redundant comment, preview spec 298. Round 1 thread: addressed in 191cc2a (removed); reply.
- P3 Nit: redundant comment, preview spec 346. Round 1 thread: addressed in 191cc2a (test dropped); reply.
- P3 Nit: redundant comment, preview spec 358. Round 1 thread: addressed in 191cc2a (test dropped); reply.
- P3 Nit: redundant comment, progress spec 125. Round 1 thread: addressed in 191cc2a (moved to the rig); reply.
- P3 Nit: redundant comment, progress spec 192. Round 1 thread: addressed in 191cc2a (moved to the rig); reply.
- P3 Nit: redundant comment, progress spec 205. Round 1 thread: addressed in 191cc2a (moved to the rig); reply.
- P3: comment claimed the page round trip had settled, progress spec 251-253. Round 1 thread: addressed in 191cc2a (removed); reply.
- P3 Nit: redundant comment, progress spec 259. Round 1 thread: addressed in 191cc2a (removed); reply.
- P3 Nit: redundant comment, progress spec 298. Round 1 thread: addressed in 191cc2a (removed); reply.
- P3 Nit: redundant comment, progress spec 310-311. Round 1 thread: addressed in 191cc2a (removed); reply.
- P3: PR description "Branch health" paragraph was stale (round 1): still open; the description has drifted further, see the body.
- P3 Nit: PR description length and template (round 1): still open.
- Cursor findings, dispositions in round 1:
widgetOwnerOfdocblock, stale-graph test name andfindSubgraphByUuiddocblock are resolved by 191cc2a; store-key mismatch, virtual-node value widgets and silent drop stay contested (production intents always carry a registered root id);'serialize' in widgetis the open item above; legacyprogress_textframe and the host-side-only toast fence stay open at P3. - New this round: fixture port diverged from
main(docblock and a blank line), a redundant unit test duplicating a table row, and a stray blank-line deletion; all inline.
Checks at 191cc2a67f: vitest run src/workbench/extensions/agent/crdt/ src/stores/widgetValueStore.test.ts 58 files, 847 passed; docOpMinter.test.ts with the minter reverted to cloud/1.55, 11 of 45 failed; with the ?? suggestion applied, 45/45; pnpm typecheck and pnpm typecheck:browser exit 0; eslint and oxfmt clean on the six changed files; playwright test --list shows 3 tests. CI: lint-and-format passed; test and the cloud Playwright job were in progress. I did not run the Playwright specs against a backend.
| const serialize = | ||
| widget && 'serialize' in widget ? widget.serialize : stored?.serialize |
There was a problem hiding this comment.
P3, carried from my earlier thread: 'serialize' in widget still lets an own serialize: undefined override a stored serialize: false. The new test lets an explicit live serialize flag override stale store metadata pins live precedence for an explicit true, and ?? keeps that: I ran the file with this line, 45/45, oxfmt clean.
| const serialize = | |
| widget && 'serialize' in widget ? widget.serialize : stored?.serialize | |
| const serialize = widget?.serialize ?? stored?.serialize |
There was a problem hiding this comment.
Addressed in 54a76e2: isValueWidget now uses widget?.serialize ?? stored?.serialize, and the new leaves it undefined test fails with the old 'serialize' in widget check (I ran it: 1 failed, 44 passed).
| withGraphIntentSource('load', () => { | ||
| graph.add(host) | ||
| subgraph.add(interior) | ||
| }) |
There was a problem hiding this comment.
P3 Nit: this hunk only deletes the blank line before interior.widgets![0].value = 3 in a test the PR doesn't otherwise change. It came in with the earlier comment that was later removed; restoring it keeps the diff to what the fix needs.
| }) | |
| }) | |
There was a problem hiding this comment.
Addressed in 54a76e2: the blank line is restored, so this test is out of the diff.
| it('uses the live serialize flag for direct store-path intents', async () => { | ||
| const { source } = seedGraph(graph) | ||
| source.widgets![0].serialize = false | ||
|
|
||
| emitGraphIntent({ | ||
| type: 'set_widget', | ||
| graphId: graph.id, | ||
| nodeId: source.id, | ||
| name: 'steps', | ||
| value: 21, | ||
| previous: 20 | ||
| }) | ||
| await afterFlush() | ||
|
|
||
| expect(minted).toEqual([]) | ||
| }) | ||
|
|
There was a problem hiding this comment.
P3: this test and the serialize false (live setter) table row exercise the same thing. The live setter goes through widgetValueStore.observeValue, which emits the same root-graph set_widget intent this test emits by hand, and both reach isValueWidget with a resolved owner whose widget has serialize: false. I can't name an implementation that fails one and passes the other, so the table row is enough.
| it('uses the live serialize flag for direct store-path intents', async () => { | |
| const { source } = seedGraph(graph) | |
| source.widgets![0].serialize = false | |
| emitGraphIntent({ | |
| type: 'set_widget', | |
| graphId: graph.id, | |
| nodeId: source.id, | |
| name: 'steps', | |
| value: 21, | |
| previous: 20 | |
| }) | |
| await afterFlush() | |
| expect(minted).toEqual([]) | |
| }) |
There was a problem hiding this comment.
Addressed in 54a76e2: the direct store-path duplicate is gone; the live serialize: false case is covered by the table row.
| const { hostSocket } = rig | ||
|
|
||
| await test.step('stream a text output into the preview widget', async () => { | ||
| // Through the node's own `onExecuted` hook. |
There was a problem hiding this comment.
P3 Nit, carried from my earlier thread: the file docblock at lines 11-15 already says preview_text arrives through onExecuted.
| // Through the node's own `onExecuted` hook. |
|
|
||
| const SUBSCRIBE_TIMEOUT = 15_000 | ||
|
|
||
| /** ComfyUI execution frames the fake host can emit alongside Agent frames. */ |
There was a problem hiding this comment.
P3 Nit: this port no longer matches main's text. The docblock was shortened here and main keeps a blank line after sendExecutionBinary (see line 161). Last round the port was verbatim, which is why I cleared narrowing it; keeping it identical avoids a conflict in this fixture on the next backport.
| /** ComfyUI execution frames the fake host can emit alongside Agent frames. */ | |
| /** | |
| * ComfyUI execution frames the fake host can emit alongside doc and agent | |
| * frames, typed against the app's own ws message shapes so a fixture cannot | |
| * drift from what `executionStore` actually parses. | |
| */ |
There was a problem hiding this comment.
Addressed in 54a76e2: the ExecutionHostFrame docblock matches main again.
DrJKL
left a comment
There was a problem hiding this comment.
Verdict: ready to approve
54a76e279f addresses all six of my round-two comments. It also closes the two items I had left in the body: the host-side-only toast fence and the stale PR description. I replied in each thread. Nothing I found blocks this PR.
One P3 remains, outside this PR. isValueWidget now reads widget?.serialize ?? stored?.serialize here, but main (docOpMinter.ts:140) still has the 'serialize' in widget check that 9fdb803 shipped. So cloud/1.55 and main now classify a live serialize: undefined differently. A small follow-up on main carrying this line and the leaves it undefined test would bring the branches back in line and keep the next backport of this file clean. The legacy-frame-only progress spec also stays open as a P3 test-scope limit, as before.
Verification at 54a76e279f:
vitest run src/workbench/extensions/agent/crdt/ src/stores/widgetValueStore.test.ts: 58 files, 847 passed.docOpMinter.test.tswith the old'serialize' in widgetline put back: onlyuses the store serialize flag when a projected widget leaves it undefinedfails (1 failed, 44 passed). With the whole minter reverted tocloud/1.55: 11 of 45 fail.pnpm typecheckandpnpm typecheck:browserexit 0. eslint and oxfmt are clean on the six changed files.playwright test --listshows 3 tests.- The fixture now differs from
mainonly in the subscribe-refusal members this branch doesn't have. - CI:
lint-and-formatpassed;testwas still in progress when I posted, and there were no failed check runs. I did not run the Playwright specs against a backend.
This backport was made by hand, so I'm posting this as a comment rather than an approval. It is ready once test and e2e-status are green.
Review record
Round 1 (review, at 4a515780)
- P3: explicit
serialize: undefinedoverrode a storedserialize: false(round 2, carried from round 1). Addressed in 54a76e2 (widget?.serialize ?? stored?.serializeplus theleaves it undefinedtest); reply. - P3:
reachableIntentGraphsubgraph branches only reached by fixture-shaped intents (round 1). Left as is, matchingmain; I said in round 2 that I'm fine with that. - P3 Nit:
widgetTypeone-line alias (round 1). Addressed in 191cc2a (inlined intoisValueWidget); reply. - P3:
widgetOwnerOfdocblock promised name-based disambiguation (round 1). Addressed in 191cc2a (function and docblock removed); reply. - P3:
reachableIntentGraphdocblock overstated "still renders" (round 1). Addressed in 191cc2a (docblock removed); reply. - P3:
isValueWidgetWritedocblock contradicted its fail-closed branch (round 1). Addressed in 191cc2a (docblock rewritten); reply. - P3: table the non-value unit cases (round 1). Addressed in 191cc2a (
it.fortable); reply. - P3:
stored!aftertoBeDefined(), first test (round 1). Addressed in 191cc2a (assert.exists); reply. - P3:
stored!aftertoBeDefined(), second test (round 1). Addressed in 191cc2a (assert.exists); reply. - P3: stale-graph test name overclaimed (round 1). Addressed in 191cc2a (renamed); reply.
- P3:
stored!aftertoBeDefined(), third test (round 1). Addressed in 191cc2a (assert.exists); reply. - P3: duplicated rig and helpers across the two specs (round 1). Addressed in 191cc2a (
AgentNonValueWidgetRig); reply. - P3 Nit: comment restating
sendTextOutputin the preview spec (round 2, carried from round 1). Addressed in 54a76e2 (comment removed); reply. - P3 Nit: redundant comment, preview spec 151 (round 1). Addressed in 191cc2a (moved to the rig); reply.
- P3 Nit: redundant comment, preview spec 235 (round 1). Addressed in 191cc2a (moved to the rig); reply.
- P3 Nit: redundant comment, preview spec 248 (round 1). Addressed in 191cc2a (moved to the rig); reply.
- P3 Nit: redundant comment, preview spec 291-292 (round 1). Addressed in 191cc2a (removed); reply.
- P3 Nit: redundant comment, preview spec 298 (round 1). Addressed in 191cc2a (removed); reply.
- P3 Nit: redundant comment, preview spec 346 (round 1). Addressed in 191cc2a (test dropped); reply.
- P3 Nit: redundant comment, preview spec 358 (round 1). Addressed in 191cc2a (test dropped); reply.
- P3 Nit: redundant comment, progress spec 125 (round 1). Addressed in 191cc2a (moved to the rig); reply.
- P3 Nit: redundant comment, progress spec 192 (round 1). Addressed in 191cc2a (moved to the rig); reply.
- P3 Nit: redundant comment, progress spec 205 (round 1). Addressed in 191cc2a (moved to the rig); reply.
- P3 Nit: redundant comment, progress spec 259 (round 1). Addressed in 191cc2a (removed); reply.
- P3 Nit: redundant comment, progress spec 298 (round 1). Addressed in 191cc2a (removed); reply.
- P3 Nit: redundant comment, progress spec 310-311 (round 1). Addressed in 191cc2a (removed); reply.
- P3: comment claimed the page round trip had settled (progress spec 251-253) (round 1). Addressed in 191cc2a (comment removed); reply.
- P3: PR description "Branch health" paragraph was stale (round 1). Addressed by the description rewrite at 2026-10-03T01:18Z: no dependencies, #19907 named.
- P3 Nit: PR description length and template (round 1). Addressed by the same rewrite; it now follows the template in a few lines.
- Cursor-review findings, dispositions in round 1:
widgetOwnerOfdocblock vs body,findSubgraphByUuiddocblock: resolved in 191cc2a (docblocks and helper removed).- Stale-graph test vacuous under its name: resolved in 191cc2a (renamed).
'serialize' in widgetwith an explicitundefined: resolved in 54a76e2.- Toast and value assertions fenced only host-side: resolved in 54a76e2. Both specs now send a later execution frame on the same
/wsafter the outcome poll and wait for the widget to show it. The host had already sentdoc_ops_resulton that socket, so the page has handled it before the toast and value assertions run. - Store-key mismatch, virtual-node value widgets, silent drop: still contested. Production intents always carry a registered root graph id, and virtual-node value writes mint the same on the base.
- Legacy
progress_textframe layout only: still open as a P3 test-scope limit. Not blocking; the minter gate is the same on both parser branches.
Round 2 (review, at 191cc2a67f)
- P3 Nit: stray blank-line deletion in an unrelated test (round 2). Addressed in 54a76e2 (blank line restored); reply.
- P3: direct store-path test duplicated a table row (round 2). Addressed in 54a76e2 (test removed); reply.
- P3 Nit: fixture
ExecutionHostFramedocblock diverged frommain(round 2). Addressed in 54a76e2 (main's docblock restored); reply. - P3 Nit: missing blank line after
sendExecutionBinary(round 2). Addressed in 54a76e2 (restored); reply. - P3: PR description stale for the cherry-picked head (round 2 body). Addressed by the 01:18Z rewrite.
Round 3 (this review, at 54a76e279f)
- P3:
mainstill has the'serialize' in widgetcheck. Open, body only, not blocking this PR.
christian-byrne
left a comment
There was a problem hiding this comment.
Reviewed exact head 54a76e2 under the standing Agent 1.54/1.55 release-backport grant. The serialize-undefined fallback and page-settlement/non-vacuity findings are repaired; style-only diff hygiene was addressed without semantic expansion; focused unit, type, lint/format, and 3/3 black-box proof pass; both correctness mutations fail at the intended assertions; hosted test and e2e-status are terminal green; all review threads are resolved. This is an honest COMMENT review, not a manufactured self-approval.
…ze when the live widget leaves it undefined (#20027) (#20134) Backport of #20027. Auto-backport failed on conflicts in `docOpMinter.ts` and `docOpMinter.test.ts` — resolved manually. ## Conflict resolution `cloud/1.55` already carried a partial version of this fix for `serialize` only (from the earlier hand-backport of #19738, #19889), but never got the `type` precedence fix. #20027's incoming side unifies both fields into one destructuring-default construct that is a strict superset of what was already on this branch, so the resolution takes #20027's version of `isValueWidget` entirely: ```ts const { type = stored?.type, serialize = stored?.serialize } = widget ?? {} ``` The test file conflict was a pure addition (the new `uses the store type when the live widget leaves it undefined` test) with nothing to reconcile against on this branch. ## Verification - `pnpm vitest run src/workbench/extensions/agent/crdt/docOpMinter.test.ts` — 46/46 pass - `pnpm typecheck` — clean
Summary
Backports the merged FE-3161 C2 fix from #19738 to
cloud/1.55: runtime-only non-value widget writes no longer become outboundset_widgetoperations that the doc host rejects.Changes
9fdb803a, its unit coverage, and three black-box Agent cases; exact-head review repair also falls back to stored metadata when a live widget hasserialize: undefinedand fences negative UI assertions on a later page-observed WebSocket frame.opSender.test.tsrelease-branch failures.Review Focus
The production change is limited to non-value widget classification in
docOpMinter. The release fixture ports the execution-frame sender required by the source specs; no unrelatedmainchanges are carried.Exact-head proof: 45/45 focused unit tests, 3/3 cloud Playwright cases,
pnpm typecheck,pnpm typecheck:browser, ESLint, and oxfmt pass. Replacing theserializefallback with the prior property-presence check kills the new undefined case; bypassing the non-value gate kills all three browser cases.Fixes the C2 half of FE-3161.
Justification: The standing Agent 1.54/1.55 release-backport grant authorizes an honest COMMENT review and exact-head admin merge without manufacturing self-approval; DrJKL reviewed the prior substantive head as ready to approve, this repair resolves the remaining exact-head correctness findings, all review threads are resolved, and required checks are terminal green at
54a76e279fe25f1f90ab33dd4a008e86e2b9301a.