Repository navigation
Conversation
A linked file read that failed was reported by the workflow load queue with the original error, whose message names the file, before the link loader reported its own fixed message. The loader now tells the workflow service it reports the failure itself, so the queue skips its report for that open. Other callers keep the queue's report. The ownership check a link passes to openWorkflow stopped at the graph load. A link replaced while an extension hook such as Load3D's beforeConfigureGraph was still running would still configure and activate its workflow under the newer link's URL. openWorkflow now passes the check to loadGraphData, which gives the load up before configuring the graph and again before activating the workflow; the service then repaints the retained workflow and drops the fetched file. The graph-load call moves into loadWorkflowGraph so the queued open does not grow, and loadGraphDataNow loses its comments per AGENTS.md. While a navigation guard holds a newer link, only the final sweep made no replace. The Settings loader, and every other loader after the workflow-id link, still replaced the committed query, cancelling the newer navigation. Those steps now do not run while a newer link is in flight. Steps before the workflow-id link still run, since at boot the stash can hold a link the loader has not reached yet.
🎭 Playwright: ✅ 3277 passed, 0 failed📊 Browser Reports
🎨 Storybook: ✅ Built — View Storybook🧩 Custom Nodes Ecosystem Matrix: 🟡 Withheld — could not measure
Findings and check detailsExecutes the frontend JS of every registry pack that ships any, against this change, in both renderers. The eight The matrix runs on same-repository version-bump PRs with relevant changes, main-branch pushes, and manual dispatches. This candidate does not qualify, so every expensive stage was skipped. Full docs: scripts/registry-census/README.md 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
🌾 Fallow: ✅ PassedFindings and check detailsFallow 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. Full docs: ADR-DEVEX-LINT-0015 Reproduce locally# the base the gate compared against
npx fallow audit --root . --quiet --format json \
--changed-since 62580009f52f2f06f5534b1e11225ab054294224
🏛️ Architecture: ✅ PassedFindings and check detailsWarnings (436)
Full docs: Frontend domain catalog Reproduce locallypnpm architecture:check |
|
Closing as a duplicate of #23269, which fixes the same three things. |
Codecov Report✅ All modified and coverable lines are covered by tests. @@ Coverage Diff @@
## main #23271 +/- ##
==========================================
- Coverage 87.65% 87.55% -0.11%
==========================================
Files 3094 2987 -107
Lines 165998 162701 -3297
Branches 42313 41188 -1125
==========================================
- Hits 145508 142448 -3060
+ Misses 19539 19460 -79
+ Partials 951 793 -158
Flags with carried forward coverage won't be shown. Click here to find out more.
... and 116 files with indirect coverage changes 🚀 New features to boost your workflow:
|
Summary
TL;DR: Three fixes to
?workflowId=links, following #23051 and #23245.Why / Before / After
workflows/Customer Jane Doe.jsonand its file read returns 500. The workflow load queue reportsFailed to load file 'workflows/Customer Jane Doe.json': 500 Internal Server Errorto Sentry and Datadog. The link loader's fixed message comes after it. After: the loader passesreportLoadFailure: falsetoopenWorkflow, and the queue skips its report for that open. The only report is "Could not open workflow link". Other callers ofopenWorkflowreport as before, sincesrc/AGENTS.mdallows file names in error messages.beforeConfigureGraphhook waits for its imports, you go to link B. A still lands on the canvas and becomes the active tab, with B in the address bar. After:openWorkflowpasses itsisCurrentcheck toloadGraphData. The load stops beforerootGraph.configureand again before the workflow is activated. The tab you had is repainted and stays active. The fetched copy of A is dropped, as it already was when the check failed before the graph load.?workflowId=A&settings=plan-credits. While A's URL cleanup waits, you go to link B, and a sign-in guard holds B. The Settings loader then replaces the URL with A's query minussettings, and that replace cancels B. After: while a newer link waits, no URL loader after the workflow-id loader runs, the final sweep included. The loaders before it (deployment, workspace) still run: at boot after sign-in, the stash can hold a link the loader has not started yet.The ADR is updated to match.
loadGraphDataNowandbuildUrlActionStepslose their existing comments, per AGENTS.md "Delete the existing comments in any function you modify". The graph-load call inopenWorkflowmoves intoloadWorkflowGraph, so the queued open stays below its old complexity.Seen in production by: not seen; found by reading the code path from the link loader through the workflow service,
app.loadGraphDataand the URL loaders.Rule bent: none.
Verification
Each new test fails on main:
useWorkflowLinkUrlLoader.fileRead.test.tsopens a link through the real workflow service with a 500 file read and expects exactly one report, the fixed one. On main the queue reports the error that names the file first.app.test.ts:isCurrentturns false whilebeforeConfigureGraphwaits; the load resolves false without configuring or activating. The same whileafterConfigureGraphwaits, without activating. On main the option is ignored.workflowService.test.ts: a graph load that gives up makesopenWorkflowreturn false, repaint the current workflow and drop the fetched file. On mainisCurrentnever reaches the graph load.useWorkflowLinkUrlLoader.test.tsnow runs the real Settings loader (only the Settings dialog is mocked) on a memory router with guards; B lands. On main the Settings replace cancels B and the URL ends empty.useUrlActionLoaders.test.ts: the in-flight case now expects the top-up loader not to run, and the workspace loader still to run.Not run: unit tests, typecheck, lint and e2e, which CI runs.
pnpm test:browser --project=cloud browser_tests/tests/workflow/cloudWorkflowLink.spec.tscovers open, refusal and sign-in, not these races.