Injected client crashes on route navigation with TypeError (_JsonMap is not a subtype of List<Object?>) — page renders blank, no error surfaces to app (Flutter 3.44.5)

[rocmiller]

Environment

• Flutter 3.44.5 (stable), Dart 3.12.2, Framework f94f4fc76b, Engine 83675ed276
• macOS (Darwin 25.5), Chrome (current stable)
• Reproduced with both flutter run -d web-server (no Dart Debug Extension) and -d chrome via VS Code
• Frontend server runs with the (now mandatory) DDC library bundle format: --dartdevc-canary --dartdevc-module-format=ddc. --no-web-experimental-hot-reload no longer changes this (deprecated no-op), so there is no opt-out.

Symptom

In a debug web build of a large multi-package Flutter app (go_router), navigating from a list screen to a detail route (context.goNamed(...) → route with a plain builder:) causes:

• the new page to render completely blank (URL updates, shell/sidebar still visible),
• all subsequent pointer interaction to go dead until a full browser reload,
• no Flutter exception anywhere — we subscribed to the VM service Extension stream and Flutter.Error never fires; the app's own debugPrints stop at that moment,
• the only console artifact is the injected client crashing at exactly that navigation:

Unhandled error detected in the injected client.js script.

TypeError: Instance of '_JsonMap': type '_JsonMap' is not a subtype of type 'List<Object?>'
    at Rti._generalAsCheckImplementation [as _as] (client.js:2729)
    at Rti._installSpecializedAsCheck (client.js:2664)
    at BuiltJsonSerializers._deserialize$3 (client.js:20953)
    at BuiltJsonSerializers.deserialize$2$specifiedType (client.js:20938)
    at BuiltJsonSerializers.deserialize$1 (client.js:20946)
    at Object._deserializeEvent (client.js:9920)
    at client.js:26905
    at _wrapJsFunctionForAsync_closure.$protected (client.js:3893)
    ... (stream/zone plumbing frames)

So the client dies deserializing an event from the debug service (_deserializeEvent received a JSON object where a List<Object?> was expected — looks like a BatchedEvents-shaped payload mismatch), and after that the route's code never runs/paints.

Evidence that this is tooling, not app code

• ext.flutter.debugDumpRenderTree on the blank page shows the route's subtree stuck NEEDS-LAYOUT / size: MISSING while ancestors are laid out; ext.flutter.debugPaint still repaints the rest of the UI (frame pipeline alive).
• The same commit, same data, same browser in --release (and --profile) renders the detail route perfectly.
• Reproduced across two independent branches of the app; also unaffected by go_router version (17.0.0 vs 17.3.0 both fail in debug).
• Everything worked on Flutter 3.38.x; the regression appeared after upgrading the app to 3.44.5.

What we couldn't do

• Produce a minimal repro so far — in our app it reproduces 100% of the time on this navigation, but it's a large app (multi-package workspace, go_router shell + subroutes, Firebase). Happy to run instrumented builds / patched dwds against it to help narrow this down.
• Find an opt-out: --no-web-experimental-hot-reload is accepted but the frontend server still runs --dartdevc-canary --dartdevc-module-format=ddc.

Possibly related

• [web] Hot reload via reloadSources hangs with non-root <base href> or --start-paused flutter/flutter#189432 (web hot reload hangs via reloadSources on 3.44.5 — same subsystem, different symptom)
• Flutter Web Application UI Components Not Rendering in Chrome. #2424 (older "UI components not rendering" injected-client report)

[Markzipan]

Thanks for the report! What version fo DWDS are you running? We have similar issues that have been resolved by using a version of Flutter with some recent-ish updates (that remove built_value). See: #2806

[nshahan]

It does seem like this is coming from an older version of dwds. The stack trace include member names that are no longer part of client.js.

[rocmiller]

Looks like this is basically a dupe of #2806 — stack traces are identical. The fix landed on Flutter main after da2864a but it hasn't been cherry-picked to the 3.44.x stable branch yet. Just checked: 3.44.8 (latest stable as of Jul 23) and it still pins dwds 26.2.5, so we're stuck waiting. Happy to close this as a dupe of #2806 if that's easier to track.

[Markzipan]

Thanks @rocmiller. The next release is coming up soon, but I'll chat with the team about cherry-picking this fix.