fix(worker,ui): block cancelling a task while its unit merge is draining

TaskStateService.CancelAsync allowed cancelling a WaitingForReview task
even while PlanningMergeOrchestrator was mid-drain on it: ApproveReview
awaits the whole multi-subtask merge synchronously, so a concurrent
CancelReview (UI or MCP) could flip the parent to Cancelled while the
orchestrator kept merging children onto the target branch, then
FinalizeParentDoneAsync would find the parent no longer WaitingForReview
and give up - leaving the merged diffs stranded with no rollback.

CancelAsync now rejects with a clear reason when HasActiveMerge(taskId)
is true. TaskStateService can't take PlanningMergeOrchestrator as a
direct constructor dependency (circular back to ITaskStateService), so
it takes a lazily-resolved Func<IActiveMergeState> instead, mirroring
the existing Func<ITaskStateService> cycle-break already used for
PlanningChainCoordinator.

UI polish: DetailsIslandViewModel.IsMergeDraining gates
CancelReviewCommand's CanExecute (same shape as
WorktreesOverviewModalViewModel.IsMerging), and the command's catch now
raises ErrorReported instead of swallowing the rejection silently.
This commit is contained in:
mika kuns
2026-08-06 13:38:02 +02:00
parent 0d1e3b9a6f
commit 79f90a9a8e
10 changed files with 205 additions and 5 deletions
+26 -1
View File
@@ -1,7 +1,7 @@
# Review, merge & conflict resolution
> **Explore-note — verify before trusting.** Distilled map of a subsystem, not authoritative.
> Last verified against commit `20bce9b` (2026-08-06).
> Last verified against commit `0d1e3b9` (2026-08-06).
> Drift check: `git log --oneline 20bce9b..HEAD -- src/ClaudeDo.Worker/Lifecycle src/ClaudeDo.Worker/State src/ClaudeDo.Worker/Planning src/ClaudeDo.Ui/ViewModels/Conflicts src/ClaudeDo.Worker/External`
> Stable structure only (no line numbers). See docs/explore-notes/README.md.
@@ -56,6 +56,31 @@ like any other `Done` child once the parent is approved. The MCP surface has no
An empty branch is still mergeable by design (some tasks — e.g. an audit — legitimately produce
no diff); this is a visibility fix, not a merge gate.
## Cancel is blocked while a unit merge is draining
`ApproveReview` on a task with children awaits `PlanningMergeOrchestrator.StartAsync` /
`DrainAsync` synchronously — the parent sits in `WaitingForReview` for the whole (potentially
minutes-long, one-child-at-a-time) drain. Without a guard, a concurrent `CancelReview` (UI or
`update_task_status`/`cancel_task` via MCP) could flip the parent to `Cancelled` mid-drain while
the orchestrator kept merging children's worktrees onto the target branch; `FinalizeParentDoneAsync`
then finds the parent no longer `WaitingForReview` and gives up, leaving the merged children's
diffs stranded with no rollback.
`TaskStateService.CancelAsync` now rejects with a `TransitionResult` reason whenever
`PlanningMergeOrchestrator.HasActiveMerge(taskId)` is true for the task being cancelled, before any
DB write. Cycle note: `TaskStateService` can't take a direct constructor dependency on
`PlanningMergeOrchestrator` (which itself depends on `ITaskStateService`), so it takes a lazily-resolved
`Func<IActiveMergeState>` instead — same cycle-breaking shape as the existing `Func<ITaskStateService>`
handed to `PlanningChainCoordinator`. `IActiveMergeState` (`Planning/Interfaces/`) is `PlanningMergeOrchestrator`'s
only public surface `TaskStateService` needs.
The UI mirrors this as polish: `DetailsIslandViewModel.IsMergeDraining` (set on
`PlanningMergeStartedEvent`, cleared on `PlanningMergeAborted`/`PlanningCompleted` for the bound
task) gates `CancelReviewCommand`'s `CanExecute` — same shape as `WorktreesOverviewModalViewModel.IsMerging`
gating `CanMergeAll`. The worker-side guard remains the actual correctness fix; the button gate
just avoids inviting a click the worker would reject. `CancelReviewAsync`'s catch now raises
`ErrorReported` (→ shell `FlashFooterError`) instead of swallowing the rejection silently.
## Post-merge verify gate
A list can set `ListConfigEntity.VerifyCommand` (List Settings modal → Verification).
@@ -35,6 +35,9 @@ public sealed partial class DetailsIslandViewModel : ViewModelBase, IDisposable
private readonly Action<string, string, string, DateTime> _workerTaskFinishedHandler;
private readonly Action<string> _workerWorktreeUpdatedHandler;
private readonly Action<string> _workerTaskUpdatedHandler;
private readonly Action<string, string> _workerPlanningMergeStartedHandler;
private readonly Action<string> _workerPlanningMergeAbortedHandler;
private readonly Action<string> _workerPlanningCompletedHandler;
[ObservableProperty] private bool _isNotesMode;
[ObservableProperty] private bool _isPrepMode;
@@ -351,6 +354,7 @@ public sealed partial class DetailsIslandViewModel : ViewModelBase, IDisposable
ResetAndRetryCommand.NotifyCanExecuteChanged();
ContinueCommand.NotifyCanExecuteChanged();
SendRoadblockReplyCommand.NotifyCanExecuteChanged();
CancelReviewCommand.NotifyCanExecuteChanged();
}
};
_worker.PropertyChanged += _workerPropertyChangedHandler;
@@ -392,6 +396,24 @@ public sealed partial class DetailsIslandViewModel : ViewModelBase, IDisposable
};
_worker.TaskUpdatedEvent += _workerTaskUpdatedHandler;
_workerPlanningMergeStartedHandler = (planningTaskId, _) =>
{
if (Task?.Id == planningTaskId) IsMergeDraining = true;
};
_worker.PlanningMergeStartedEvent += _workerPlanningMergeStartedHandler;
_workerPlanningMergeAbortedHandler = planningTaskId =>
{
if (Task?.Id == planningTaskId) IsMergeDraining = false;
};
_worker.PlanningMergeAbortedEvent += _workerPlanningMergeAbortedHandler;
_workerPlanningCompletedHandler = planningTaskId =>
{
if (Task?.Id == planningTaskId) IsMergeDraining = false;
};
_worker.PlanningCompletedEvent += _workerPlanningCompletedHandler;
ChildOutcomes.CollectionChanged += (_, _) =>
{
Merge.SyncChildOutcomes(HasChildOutcomes, Subtasks.Count);
@@ -409,6 +431,9 @@ public sealed partial class DetailsIslandViewModel : ViewModelBase, IDisposable
_worker.TaskFinishedEvent -= _workerTaskFinishedHandler;
_worker.WorktreeUpdatedEvent -= _workerWorktreeUpdatedHandler;
_worker.TaskUpdatedEvent -= _workerTaskUpdatedHandler;
_worker.PlanningMergeStartedEvent -= _workerPlanningMergeStartedHandler;
_worker.PlanningMergeAbortedEvent -= _workerPlanningMergeAbortedHandler;
_worker.PlanningCompletedEvent -= _workerPlanningCompletedHandler;
AgentSettings.Dispose();
Prep.Dispose();
}
@@ -438,6 +463,7 @@ public sealed partial class DetailsIslandViewModel : ViewModelBase, IDisposable
DequeueCommand.NotifyCanExecuteChanged();
ResetAndRetryCommand.NotifyCanExecuteChanged();
ContinueCommand.NotifyCanExecuteChanged();
CancelReviewCommand.NotifyCanExecuteChanged();
// A state change means a new run/review cycle: the diff must be
// re-inspected before merge can be approved again.
ReviewDiffViewed = false;
@@ -547,6 +573,7 @@ public sealed partial class DetailsIslandViewModel : ViewModelBase, IDisposable
OnPropertyChanged(nameof(TaskIdBadge));
Monitor.Reset();
RoadblockReplyDraft = string.Empty;
IsMergeDraining = false;
Subtasks.Clear();
ChildOutcomes.Clear();
Attachments.Clear();
@@ -1204,14 +1231,28 @@ public sealed partial class DetailsIslandViewModel : ViewModelBase, IDisposable
catch { /* stale review action; broadcast reconciles */ }
}
[RelayCommand]
// True while a unit merge is actively draining this task's subtasks onto the target
// branch — set from PlanningMergeStarted, cleared on PlanningMergeAborted/Completed.
// Mirrors the worker-side guard in TaskStateService.CancelAsync (the real correctness
// fix); this just keeps the button from inviting a click the worker will reject anyway.
[ObservableProperty]
[NotifyCanExecuteChangedFor(nameof(CancelReviewCommand))]
private bool _isMergeDraining;
[RelayCommand(CanExecute = nameof(CanCancelReview))]
private async System.Threading.Tasks.Task CancelReviewAsync()
{
if (Task is null || !_worker.IsConnected) return;
try { await _worker.CancelReviewAsync(Task.Id); }
catch { /* stale review action; broadcast reconciles */ }
catch (Exception ex)
{
ErrorReported?.Invoke(ex.Message);
}
}
private bool CanCancelReview() =>
Task != null && _worker.IsConnected && !IsMergeDraining;
private async System.Threading.Tasks.Task ReloadAttachmentsAsync()
{
if (Task is null) return;
@@ -0,0 +1,10 @@
namespace ClaudeDo.Worker.Planning;
// Lets TaskStateService guard CancelAsync against an in-progress unit merge without a
// hard constructor cycle back to PlanningMergeOrchestrator (which itself depends on
// ITaskStateService) — TaskStateService takes a Func<IActiveMergeState> instead, resolved
// lazily so both singletons can finish constructing before either calls into the other.
public interface IActiveMergeState
{
bool HasActiveMerge(string taskId);
}
@@ -13,7 +13,7 @@ namespace ClaudeDo.Worker.Planning;
/// <summary>A unit-merge conflict currently paused, driven by an MCP session rather than the UI.</summary>
public sealed record ExternalPlanningMergeConflict(string PlanningTaskId, string SubtaskId);
public sealed class PlanningMergeOrchestrator
public sealed class PlanningMergeOrchestrator : IActiveMergeState
{
private readonly IDbContextFactory<ClaudeDoDbContext> _dbFactory;
private readonly TaskMergeService _merge;
+4
View File
@@ -101,12 +101,16 @@ builder.Services.AddSingleton<IQueuePicker, QueuePicker>();
builder.Services.AddSingleton<RunCancellationRegistry>();
builder.Services.AddSingleton<Func<ITaskStateService>>(sp => () => sp.GetRequiredService<ITaskStateService>());
// PlanningMergeOrchestrator itself depends on ITaskStateService, so TaskStateService can only
// reach it lazily (Func<IActiveMergeState>) — same cycle-breaking shape as the Func above.
builder.Services.AddSingleton<Func<IActiveMergeState>>(sp => () => sp.GetRequiredService<PlanningMergeOrchestrator>());
builder.Services.AddSingleton<ITaskStateService>(sp => new TaskStateService(
sp.GetRequiredService<IDbContextFactory<ClaudeDoDbContext>>(),
sp.GetRequiredService<HubBroadcaster>(),
sp.GetRequiredService<IQueueWaker>(),
sp.GetRequiredService<PlanningChainCoordinator>(),
sp.GetRequiredService<RunCancellationRegistry>(),
sp.GetRequiredService<Func<IActiveMergeState>>(),
sp.GetRequiredService<ILogger<TaskStateService>>()));
// Agent file management.
@@ -16,6 +16,7 @@ public sealed class TaskStateService : ITaskStateService
private readonly IQueueWaker _waker;
private readonly PlanningChainCoordinator _chain;
private readonly RunCancellationRegistry _runCancels;
private readonly Func<IActiveMergeState> _mergeState;
private readonly ILogger<TaskStateService> _logger;
public TaskStateService(
@@ -24,6 +25,7 @@ public sealed class TaskStateService : ITaskStateService
IQueueWaker waker,
PlanningChainCoordinator chain,
RunCancellationRegistry runCancels,
Func<IActiveMergeState> mergeState,
ILogger<TaskStateService> logger)
{
_dbFactory = dbFactory;
@@ -31,6 +33,7 @@ public sealed class TaskStateService : ITaskStateService
_waker = waker;
_chain = chain;
_runCancels = runCancels;
_mergeState = mergeState;
_logger = logger;
}
@@ -255,6 +258,14 @@ public sealed class TaskStateService : ITaskStateService
public async Task<TransitionResult> CancelAsync(
string taskId, DateTime finishedAt, CancellationToken ct, bool allowFromIdle = false)
{
// A unit merge drains its children's worktrees onto the target branch while the parent
// sits in WaitingForReview for the whole (potentially minutes-long) drain. Cancelling out
// from under it would flip the parent to Cancelled while the orchestrator keeps merging —
// FinalizeParentDoneAsync then finds the parent no longer WaitingForReview and gives up,
// leaving the merged children's diffs stranded with no way to roll them back.
if (_mergeState().HasActiveMerge(taskId))
return new TransitionResult(false, "A merge is in progress for this task; wait for it to finish before cancelling.");
List<string> cancelledChildIds;
await using (var ctx = await _dbFactory.CreateDbContextAsync(ct))
{
@@ -55,6 +55,7 @@ public abstract class StubWorkerClient : IWorkerClient
public void RaiseHandoffRequested(string taskId, IReadOnlyList<string> survivingTaskIds) => HandoffRequestedEvent?.Invoke(taskId, survivingTaskIds);
public void RaisePlanningMergeConflict(string planningTaskId, string subtaskId, IReadOnlyList<string> files, bool externallyDriven)
=> PlanningMergeConflictEvent?.Invoke(planningTaskId, subtaskId, files, externallyDriven);
public void RaisePlanningMergeStarted(string planningTaskId, string targetBranch) => PlanningMergeStartedEvent?.Invoke(planningTaskId, targetBranch);
public void RaisePlanningMergeAborted(string planningTaskId) => PlanningMergeAbortedEvent?.Invoke(planningTaskId);
public void RaisePlanningCompleted(string planningTaskId) => PlanningCompletedEvent?.Invoke(planningTaskId);
@@ -161,4 +161,69 @@ public class DetailsIslandReviewActionsTests : IDisposable
Assert.Equal(worker.ExceptionMessage, reportedError);
}
[Fact]
public void CancelReview_IsDisabled_WhileUnitMergeIsDraining()
{
var worker = new RecordingWorkerClient();
var vm = BuildVm(worker);
vm.Bind(new TaskRowViewModel { Id = "task-drain-1", Status = TaskStatus.WaitingForReview });
Assert.True(vm.CancelReviewCommand.CanExecute(null));
worker.RaisePlanningMergeStarted("task-drain-1", "main");
Assert.False(vm.CancelReviewCommand.CanExecute(null));
worker.RaisePlanningCompleted("task-drain-1");
Assert.True(vm.CancelReviewCommand.CanExecute(null));
}
[Fact]
public void CancelReview_ReDisables_UntilMergeAborted()
{
var worker = new RecordingWorkerClient();
var vm = BuildVm(worker);
vm.Bind(new TaskRowViewModel { Id = "task-drain-2", Status = TaskStatus.WaitingForReview });
worker.RaisePlanningMergeStarted("task-drain-2", "main");
Assert.False(vm.CancelReviewCommand.CanExecute(null));
worker.RaisePlanningMergeAborted("task-drain-2");
Assert.True(vm.CancelReviewCommand.CanExecute(null));
}
[Fact]
public void CancelReview_UnaffectedByMergeEvents_ForADifferentTask()
{
var worker = new RecordingWorkerClient();
var vm = BuildVm(worker);
vm.Bind(new TaskRowViewModel { Id = "task-drain-3", Status = TaskStatus.WaitingForReview });
worker.RaisePlanningMergeStarted("some-other-parent", "main");
Assert.True(vm.CancelReviewCommand.CanExecute(null));
}
private sealed class ThrowingCancelWorkerClient : StubWorkerClient
{
public override bool IsConnected => true;
public string ExceptionMessage { get; init; } =
"A merge is in progress for this task; wait for it to finish before cancelling.";
public override Task CancelReviewAsync(string taskId) => throw new Exception(ExceptionMessage);
}
[Fact]
public async Task CancelReview_WhenWorkerThrows_RaisesErrorReported()
{
var worker = new ThrowingCancelWorkerClient();
var vm = BuildVm(worker);
vm.Bind(new TaskRowViewModel { Id = "task-cancel-err-1", Status = TaskStatus.WaitingForReview });
string? reportedError = null;
vm.ErrorReported += msg => reportedError = msg;
await vm.CancelReviewCommand.ExecuteAsync(null);
Assert.Equal(worker.ExceptionMessage, reportedError);
}
}
@@ -20,7 +20,8 @@ public static class TaskStateServiceBuilder
CountingQueueWaker Waker,
RunCancellationRegistry RunCancels);
public static Built Build(IDbContextFactory<ClaudeDoDbContext> dbFactory)
public static Built Build(
IDbContextFactory<ClaudeDoDbContext> dbFactory, Func<IActiveMergeState>? mergeState = null)
{
var hub = new CapturingHubContext();
var broadcaster = new HubBroadcaster(hub);
@@ -35,12 +36,19 @@ public static class TaskStateServiceBuilder
waker,
chain,
runCancels,
mergeState ?? (() => NoActiveMergeState.Instance),
NullLogger<TaskStateService>.Instance);
return new Built(state, chain, hub, () => waker.Count, waker, runCancels);
}
}
file sealed class NoActiveMergeState : IActiveMergeState
{
public static readonly NoActiveMergeState Instance = new();
public bool HasActiveMerge(string taskId) => false;
}
public sealed class CountingQueueWaker : IQueueWaker
{
private int _count;
@@ -1,6 +1,7 @@
using ClaudeDo.Data;
using ClaudeDo.Data.Models;
using ClaudeDo.Data.Repositories;
using ClaudeDo.Worker.Planning;
using ClaudeDo.Worker.State;
using ClaudeDo.Worker.Tests.Infrastructure;
using Microsoft.EntityFrameworkCore;
@@ -8,6 +9,12 @@ using TaskStatus = ClaudeDo.Data.Models.TaskStatus;
namespace ClaudeDo.Worker.Tests.State;
file sealed class FakeActiveMergeState : IActiveMergeState
{
public HashSet<string> ActiveTaskIds { get; } = new();
public bool HasActiveMerge(string taskId) => ActiveTaskIds.Contains(taskId);
}
public sealed class TaskStateServiceTests : IDisposable
{
private readonly DbFixture _db = new();
@@ -319,6 +326,34 @@ public sealed class TaskStateServiceTests : IDisposable
Assert.Equal(TaskStatus.Done, await GetStatusAsync(id));
}
[Fact]
public async Task CancelAsync_WhileUnitMergeDraining_Rejects_AndDoesNotMutate()
{
var mergeState = new FakeActiveMergeState();
var built = TaskStateServiceBuilder.Build(_factory, () => mergeState);
var id = await SeedTaskAsync(TaskStatus.WaitingForReview);
mergeState.ActiveTaskIds.Add(id);
var result = await built.State.CancelAsync(id, DateTime.UtcNow, default);
Assert.False(result.Ok);
Assert.Equal(TaskStatus.WaitingForReview, await GetStatusAsync(id));
}
[Fact]
public async Task CancelAsync_AfterUnitMergeDrainCompletes_Succeeds()
{
var mergeState = new FakeActiveMergeState();
var built = TaskStateServiceBuilder.Build(_factory, () => mergeState);
var id = await SeedTaskAsync(TaskStatus.WaitingForReview);
// Merge drain finished (or never started) for this task — cancel behaves as before.
var result = await built.State.CancelAsync(id, DateTime.UtcNow, default);
Assert.True(result.Ok);
Assert.Equal(TaskStatus.Cancelled, await GetStatusAsync(id));
}
// ─── ResetToIdleAsync ─────────────────────────────────────────────────
[Fact]