Clustering by erincatto · Pull Request #1047 · erincatto/box2d

erincatto · 2026-04-07T06:17:43Z

Experiment with clustering (cell) solver

Copilot

Pull request overview

This PR experiments with a new cluster-based (“cell”) constraint solver pipeline, replacing the prior constraint-graph/coloring approach and refactoring how awake body state is packed and accessed during solving.

Changes:

Introduces cluster computation/classification and new cluster-oriented solver stages.
Refactors solver sets to store body IDs (vs. body sim/state arrays) and updates wake/sleep transfer logic accordingly.
Updates multiple joint/contact solver paths and temporarily disables determinism assertions/tests.

Reviewed changes

Copilot reviewed 40 out of 40 changed files in this pull request and generated 9 comments.

Show a summary per file

File	Description
`src/cluster.c`	Computes body clusters and packs awake `b2BodyState` for the step.
`src/solver.c`	Integrates cluster-based stage scheduling and bullet processing buffers.
`src/solver.h`	Updates step context (cluster data, borders, bullet body list type, stage enums).
`src/solver_set.h`	Changes solver sets to store `bodyIds` array.
`src/solver_set.c`	Updates wake/sleep/merge/transfer logic for new solver set storage.
`src/contact_solver.h`	Defines scalar contact-constraint pipeline and contact constraint indexing contract.
`src/contact_solver.c`	Prepares/warm-starts/solves contact constraints using packed states.
`src/body.c`	Updates body creation and runtime state storage on `b2Body`.
`src/wheel_joint.c`	Switches wheel joint preparation/solve to use `stateIndex` indirection.
`src/weld_joint.c`	Switches weld joint preparation/solve to use `stateIndex` indirection; adds extra asserts.
`test/main.c`	Disables running determinism test in the test runner.
`test/test_determinism.c`	Comments out determinism validation checks.
`test/test_container.c`	Removes redundant array declaration from the container test.

Comments suppressed due to low confidence (2)

src/contact_solver.c:111

Warm-start currently does int indexA = constraint->indexA - 1; / indexB = ... - 1; and then treats index == B2_NULL_INDEX as the dummy state. With the current prepare path storing raw stateIndex, this makes stateIndex==0 become -1 and any other index off-by-one. This needs to be consistent with how indices are stored in b2PrepareContactConstraints (either base-1 everywhere or base-0 everywhere).

		b2ContactConstraint* constraint = constraints + i;

		int indexA = constraint->indexA - 1;
		int indexB = constraint->indexB - 1;

		b2BodyState* stateA = indexA == B2_NULL_INDEX ? &dummyState : states + indexA;
		b2BodyState* stateB = indexB == B2_NULL_INDEX ? &dummyState : states + indexB;

src/solver.c:1559

b2StepContext::bulletBodies is declared as b2Body** (solver.h), but this allocates awakeBodyCount * sizeof(int). This will under-allocate on 64-bit and corrupt memory when storing pointers. Allocate awakeBodyCount * sizeof(b2Body*) (or change the field back to int* consistently).

💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.

Copilot · 2026-04-07T06:29:17Z

src/cluster.c

+	// First-time or too few bodies: full k-means with seeding
+	if ( manager->initialized == false || awakeCount < B2_CLUSTER_COUNT )
+	{
+		int seedCount = b2MinInt( awakeCount, B2_CLUSTER_COUNT );
+		for ( int i = 0; i < seedCount; ++i )
+		{
+			b2Body* body = b2BodyArray_Get( &world->bodies, bodyIds[i] );
+			clusters[i].center = body->center;
+			b2Array_Push( clusters[i].bodyIds, bodyIds[i] );
+			body->clusterIndex = (int16_t)i;
+		}
+
+		if ( awakeCount < B2_CLUSTER_COUNT )
+		{
+			return;
+		}


b2ComputeClusters returns early when awakeCount < B2_CLUSTER_COUNT after seeding, leaving context->states unallocated and leaving body->stateIndex unset for the awake bodies. Later solver/joint/contact code appears to assume context->states is valid and uses body->stateIndex, which would be stale/uninitialized in this path. Allocate/copy context->states (and set body->stateIndex) even in the small-awake-count case, or otherwise ensure downstream code is gated off when context->states is NULL.

Copilot · 2026-04-07T06:29:17Z

src/body.c

@@ -298,14 +276,15 @@ b2BodyId b2CreateBody( b2WorldId worldId, const b2BodyDef* def )
 	body->islandId = B2_NULL_INDEX;
 	body->islandIndex = B2_NULL_INDEX;
 	body->bodyMoveIndex = B2_NULL_INDEX;
-	body->id = bodyId;
 	body->mass = 0.0f;
 	body->inertia = 0.0f;
 	body->sleepThreshold = def->sleepThreshold;
 	body->sleepTime = 0.0f;
+	body->minExtent = B2_HUGE;
+	body->maxExtent = 0.0f;
 	body->type = def->type;
-	body->flags = bodySim->flags;
 	body->enableSleep = def->enableSleep;
+	body->clusterIndex = B2_NULL_INDEX;


b2CreateBody is not initializing several new b2Body runtime fields (e.g., linearVelocity, angularVelocity, force, torque, and stateIndex). Since bodies are allocated from a pool and can be reused, these fields can contain stale data; additionally, the initial velocity from b2BodyDef is currently ignored. Initialize these fields explicitly (and set stateIndex to B2_NULL_INDEX until b2ComputeClusters assigns it).

Copilot · 2026-04-07T06:29:18Z

src/contact_solver.c

+		b2ContactSim* contactSim = contacts[i];
+		b2Body* bodyA = bodies + contactSim->bodyIdA;
+		b2Body* bodyB = bodies + contactSim->bodyIdB;
+		int indexA = bodyA->stateIndex;
+		int indexB = bodyB->stateIndex;
+
 		const b2Manifold* manifold = &contactSim->manifold;
 		int pointCount = manifold->pointCount;
-
 		B2_ASSERT( 0 < pointCount && pointCount <= 2 );

-		int indexA = contactSim->bodySimIndexA;
-		int indexB = contactSim->bodySimIndexB;
-
-#if B2_ENABLE_VALIDATION
-		b2Body* bodyA = bodies + contactSim->bodyIdA;
-		int validIndexA = bodyA->setIndex == b2_awakeSet ? bodyA->localIndex : B2_NULL_INDEX;
-		B2_ASSERT( indexA == validIndexA );
-
-		b2Body* bodyB = bodies + contactSim->bodyIdB;
-		int validIndexB = bodyB->setIndex == b2_awakeSet ? bodyB->localIndex : B2_NULL_INDEX;
-		B2_ASSERT( indexB == validIndexB );
-#endif
-
 		b2ContactConstraint* constraint = constraints + i;
-
-		// 0 is null
-		constraint->indexA = indexA + 1;
-		constraint->indexB = indexB + 1;
+		constraint->indexA = indexA;
+		constraint->indexB = indexB;
 		constraint->normal = manifold->normal;
 		constraint->friction = contactSim->friction;


b2PrepareContactConstraints assigns constraint->indexA/indexB directly from body->stateIndex, but b2ContactConstraint documents these indices as “base-1, 0 for null”. Downstream warm-start/solve subtracts 1, so an awake body with stateIndex == 0 will be treated as null. Store stateIndex + 1 (and ensure static bodies store 0) to match the base-1 convention, or remove the base-1 convention consistently.

Copilot · 2026-04-07T06:29:18Z

src/wheel_joint.c

 	b2Body* bodyA = b2BodyArray_Get( &world->bodies, idA );
 	b2Body* bodyB = b2BodyArray_Get( &world->bodies, idB );

-	B2_ASSERT( bodyA->setIndex == b2_awakeSet || bodyB->setIndex == b2_awakeSet );
-	b2SolverSet* setA = b2SolverSetArray_Get( &world->solverSets, bodyA->setIndex );
-	b2SolverSet* setB = b2SolverSetArray_Get( &world->solverSets, bodyB->setIndex );
+	base->stateIndexA = bodyA->stateIndex;
+	base->stateIndexB = bodyB->stateIndex;

-	int localIndexA = bodyA->localIndex;
-	int localIndexB = bodyB->localIndex;
-
-	b2BodySim* bodySimA = b2BodySimArray_Get( &setA->bodySims, localIndexA );
-	b2BodySim* bodySimB = b2BodySimArray_Get( &setB->bodySims, localIndexB );
-
-	float mA = bodySimA->invMass;
-	float iA = bodySimA->invInertia;
-	float mB = bodySimB->invMass;
-	float iB = bodySimB->invInertia;
-
-	base->invMassA = mA;
-	base->invMassB = mB;
-	base->invIA = iA;
-	base->invIB = iB;
+	float mA = bodyA->invMass;
+	float iA = bodyA->invInertia;
+	float mB = bodyB->invMass;
+	float iB = bodyB->invInertia;


base->stateIndexA/stateIndexB are copied from body->stateIndex for all bodies. For static bodies (and any body not in the awake cluster state pack), body->stateIndex must be B2_NULL_INDEX to avoid indexing into context->states with a stale/garbage value. Ensure b2Body.stateIndex is initialized to B2_NULL_INDEX for non-awake bodies (e.g., in body creation and when moving between solver sets), or compute these indices here based on body type/set membership instead of trusting body->stateIndex.

Copilot · 2026-04-07T06:29:18Z

src/weld_joint.c

 	b2Body* bodyA = b2BodyArray_Get( &world->bodies, idA );
 	b2Body* bodyB = b2BodyArray_Get( &world->bodies, idB );

-	B2_ASSERT( bodyA->setIndex == b2_awakeSet || bodyB->setIndex == b2_awakeSet );
-	b2SolverSet* setA = b2SolverSetArray_Get( &world->solverSets, bodyA->setIndex );
-	b2SolverSet* setB = b2SolverSetArray_Get( &world->solverSets, bodyB->setIndex );
+	base->stateIndexA = bodyA->stateIndex;
+	base->stateIndexB = bodyB->stateIndex;

-	int localIndexA = bodyA->localIndex;
-	int localIndexB = bodyB->localIndex;
-
-	b2BodySim* bodySimA = b2BodySimArray_Get( &setA->bodySims, localIndexA );
-	b2BodySim* bodySimB = b2BodySimArray_Get( &setB->bodySims, localIndexB );
-
-	float mA = bodySimA->invMass;
-	float iA = bodySimA->invInertia;
-	float mB = bodySimB->invMass;
-	float iB = bodySimB->invInertia;
-
-	base->invMassA = mA;
-	base->invMassB = mB;
-	base->invIA = iA;
-	base->invIB = iB;
+	float iA = bodyA->invInertia;
+	float iB = bodyB->invInertia;


base->stateIndexA/stateIndexB are taken from body->stateIndex, but static/sleeping bodies won’t be part of context->states and must use B2_NULL_INDEX so warm-start/solve can safely use dummyState. Make sure b2Body.stateIndex is reliably set to B2_NULL_INDEX for non-awake bodies (including newly-created statics and any bodies moved out of the awake set), or derive the state index here instead of copying it blindly.

Copilot · 2026-04-07T06:29:18Z

src/weld_joint.c

+	B2_ASSERT( b2IsValidVec2( vA ) );
+	B2_ASSERT( b2IsValidFloat( wA ) );
+	B2_ASSERT( b2IsValidVec2( vB ) );
+	B2_ASSERT( b2IsValidFloat( wB ) );


These assertions are duplicated back-to-back, which adds noise without increasing coverage. Remove the repeated block to keep validation concise.

Copilot · 2026-04-07T06:29:18Z

test/main.c

+	// RUN_TEST( DeterminismTest );
 	RUN_TEST( DistanceTest );
 	RUN_TEST( DynamicTreeTest );
 	RUN_TEST( IdTest );


DeterminismTest is commented out from the main unit test runner. Given this PR is a major solver refactor, disabling determinism coverage makes regressions much harder to catch. Prefer keeping this test enabled (or gating it behind an explicit build flag / platform check) rather than commenting it out unconditionally.

Copilot · 2026-04-07T06:29:19Z

test/test_determinism.c

 	}

 	enkiDeleteTaskScheduler( scheduler );

-	ENSURE( data.sleepStep == EXPECTED_SLEEP_STEP );
-	ENSURE( data.hash == EXPECTED_HASH );
+	//ENSURE( data.sleepStep == EXPECTED_SLEEP_STEP );
+	//ENSURE( data.hash == EXPECTED_HASH );



The determinism test no longer validates the expected sleep step/hash because the ENSURE checks are commented out. If the expected values changed due to solver updates, it’s better to update the constants or explicitly gate the check with a flag, rather than silently skipping the assertions (which makes the test always pass).

Copilot · 2026-04-07T06:29:19Z

test/test_determinism.c

 		done = UpdateFallingHinges( worldId, &data );
 	}

-	ENSURE( data.sleepStep == EXPECTED_SLEEP_STEP );
-	ENSURE( data.hash == EXPECTED_HASH );
+	//ENSURE( data.sleepStep == EXPECTED_SLEEP_STEP );
+	//ENSURE( data.hash == EXPECTED_HASH );



CrossPlatformTest’s determinism validation is currently disabled by commenting out the ENSURE checks. This makes the subtest non-verifying. Consider updating EXPECTED_* (if the new solver legitimately changes them) or gating the check, but keep an assertion so the test can fail on nondeterministic output.

Copilot

Pull request overview

Copilot reviewed 40 out of 40 changed files in this pull request and generated 6 comments.

💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.

Copilot · 2026-04-07T07:02:06Z

src/body.c

 	{
 		return;
 	}



b2Body_SetLinearVelocity no longer wakes a sleeping body when a non-zero velocity is set. This can make the call a no-op for sleeping bodies (they won’t be simulated) and is a behavioral regression from typical Box2D semantics. Consider waking the body when setting a meaningful non-zero velocity (or documenting/renaming if intentional).

Suggested change

if ( b2LengthSquared( linearVelocity ) > 0.0f )

{

b2WakeBody( world, body );

}

Copilot · 2026-04-07T07:02:07Z

src/body.c

@@ -808,18 +750,7 @@ void b2Body_SetAngularVelocity( b2BodyId bodyId, float angularVelocity )
 		return;
 	}



b2Body_SetAngularVelocity no longer wakes a sleeping body when a non-zero angular velocity is set, which can prevent the body from ever advancing. Consider waking the body when setting a meaningful non-zero angular velocity (or clearly documenting the new behavior).

Suggested change

if ( body->setIndex != b2_awakeSet )

{

float maxVelocity = b2AbsFloat( angularVelocity ) * body->maxExtent;

if ( maxVelocity >= body->sleepThreshold )

{

b2WakeBody( world, body );

}

}

Copilot · 2026-04-07T07:02:07Z

src/solver.h

 	// Array of bullet bodies that need continuous collision handling
-	int* bulletBodies;
+	b2Body** bulletBodies;
 	b2AtomicInt bulletBodyCount;


b2StepContext::bulletBodies is now a b2Body** (pointer array). Any arena allocations and indexing must use sizeof(b2Body*) and store b2Body* elements; allocating as sizeof(int) (or treating it as int*) will cause memory corruption on 64-bit builds.

Copilot · 2026-04-07T07:02:08Z

src/physics_world.c

+	fprintf( file, "body id: %d\n", bodyIdCapacity * (int)sizeof( int ) );
 	fprintf( file, "joint sim: %d\n", jointSimCapacity * (int)sizeof( b2JointSim ) );
 	fprintf( file, "contact sim: %d\n", contactSimCapacity * (int)sizeof( b2ContactSim ) );
 	fprintf( file, "island sim: %d\n", islandSimCapacity * (int)sizeof( islandSimCapacity ) );


The memory stats print uses sizeof(islandSimCapacity) (an int) instead of the element type stored in set->islandSims. This under-reports memory and is likely unintended; use sizeof(b2IslandSim) (or the correct island-sim type) instead.

Suggested change

fprintf( file, "island sim: %d\n", islandSimCapacity * (int)sizeof( islandSimCapacity ) );

fprintf( file, "island sim: %d\n", islandSimCapacity * (int)sizeof( b2IslandSim ) );

Copilot · 2026-04-07T07:02:08Z

src/physics_world.c

+		bool drawClusterColor = true;
+		if ( drawClusterColor && body->clusterIndex != B2_NULL_INDEX )
+		{
+			B2_ASSERT( 0 <= body->clusterIndex && body->clusterIndex < B2_CLUSTER_COUNT );
+			color = b2_clusterColors[body->clusterIndex];


Cluster-color rendering is hard-coded on (drawClusterColor = true) and overrides custom shape colors for clustered bodies. Consider wiring this to a debug draw option (similar to the removed drawGraphColors) or defaulting it off so existing visual/debug semantics aren’t changed unexpectedly.

Copilot · 2026-04-07T07:02:08Z

src/contact_solver.c

-	b2TracyCZoneEnd( store_impulses );
 }

+#if 0


Large SIMD contact solver implementation is now kept in the file but disabled with #if 0. This adds substantial dead code/maintenance surface. Consider deleting it, moving it behind a real feature flag (e.g., B2_ENABLE_SIMD_CONTACT_SOLVER), or splitting it into a separate file to keep the active solver path clear.

Suggested change

#if 0

// Optional SIMD contact solver implementation. Disabled by default unless explicitly enabled.

#if defined( B2_ENABLE_SIMD_CONTACT_SOLVER ) && B2_ENABLE_SIMD_CONTACT_SOLVER

…clustering

Copilot

Pull request overview

Copilot reviewed 45 out of 45 changed files in this pull request and generated 4 comments.

💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.

Copilot · 2026-04-11T07:06:45Z

src/arena_allocator.c

@@ -38,7 +40,7 @@ void* b2AllocateArenaItem( b2ArenaAllocator* alloc, int size, const char* name )
 	b2ArenaEntry entry;
 	entry.size = size32;
 	entry.name = name;
-	if ( alloc->index + size32 > alloc->capacity )
+	if ( alloc->index + size32 > alloc->capacity || B2_FORCE_HEAP == 1 )
 	{
 		// fall back to the heap (undesirable)
 		entry.data = b2Alloc( size32 );


B2_FORCE_HEAP is hard-coded to 1, which forces all arena allocations to fall back to heap allocations (b2Alloc). This defeats the purpose of the arena allocator (performance, fragmentation control) and changes memory accounting/behavior globally. Suggest removing this override or gating it behind a debug-only compile flag (e.g., #if defined(B2_DEBUG) && B2_FORCE_HEAP).

Copilot · 2026-04-11T07:06:45Z

src/physics_world.c

 	fprintf( file, "solver sets\n" );
-	fprintf( file, "body sim: %d\n", bodySimCapacity * (int)sizeof( b2BodySim ) );
-	fprintf( file, "body state: %d\n", bodyStateCapacity * (int)sizeof( b2BodyState ) );
+	fprintf( file, "body id: %d\n", bodyIdCapacity * (int)sizeof( int ) );
 	fprintf( file, "joint sim: %d\n", jointSimCapacity * (int)sizeof( b2JointSim ) );
 	fprintf( file, "contact sim: %d\n", contactSimCapacity * (int)sizeof( b2ContactSim ) );
 	fprintf( file, "island sim: %d\n", islandSimCapacity * (int)sizeof( islandSimCapacity ) );


The memory stat for island sims multiplies by sizeof(islandSimCapacity) (i.e., sizeof(int)), which is unrelated to the element type and will report incorrect sizes. This should use sizeof(b2IslandSim) (or the actual island-sim element type stored in set->islandSims).

Copilot · 2026-04-11T07:06:45Z

test/test_determinism.c

+	printf( "workers=%d sleepStep=%d hash=0x%08X\n", workerCount, data.sleepStep, data.hash );
+
+	//ENSURE( data.sleepStep == EXPECTED_SLEEP_STEP );
+	//ENSURE( data.hash == EXPECTED_HASH );


The multithreaded determinism test no longer enforces determinism because the ENSURE checks are commented out. This will allow non-deterministic behavior to pass CI silently. Suggest re-enabling the assertions (and updating expected values if necessary) and keeping the printf behind a debug flag if you still want the diagnostic output.

Copilot · 2026-04-11T07:06:46Z

samples/main.cpp

 	s_context.Load();
 	s_context.workerCount = b2MinInt( 8, (int)enki::GetNumHardwareThreads() / 2 );
+	s_context.workerCount = 1;



s_context.workerCount is computed from hardware threads and then immediately overridden to 1. If this was for debugging, it will unintentionally disable multithreading in the sample app for everyone. Suggest removing the forced assignment or making it conditional (e.g., a command-line flag / debug option).

…clustering

erincatto added 10 commits September 26, 2025 13:44

wip

af64933

Merge remote-tracking branch 'origin/main' into clustering

b21afdb

fixes

9ce95fa

wip

1c3f8bd

wip

b569ad7

wip

796937e

wip

9f016a8

removed constraint graph

48df827

wip

5a840ed

fixes

2de06c7

erincatto marked this pull request as draft April 7, 2026 06:18

erincatto requested a review from Copilot April 7, 2026 06:20

Copilot started reviewing on behalf of erincatto April 7, 2026 06:21 View session

Copilot AI reviewed Apr 7, 2026

View reviewed changes

erincatto added 2 commits April 6, 2026 23:40

fixes

7dfcbae

fixes

4cfe494

erincatto requested a review from Copilot April 7, 2026 06:54

Copilot started reviewing on behalf of erincatto April 7, 2026 06:55 View session

Copilot AI reviewed Apr 7, 2026

View reviewed changes

erincatto added 8 commits April 7, 2026 10:44

fixes

8560ff8

merge integration into clusters

ff2e73c

work stealing

79e7f36

prefetch

5e0d3a7

Update imgui

30f5f84

Merge branch 'clustering' of https://github.com/erincatto/box2d into …

0358516

…clustering

clustering wip

af03132

Merge branch 'clustering' of github.com:erincatto/box2d into clustering

93c3b11

erincatto requested a review from Copilot April 11, 2026 07:03

Copilot started reviewing on behalf of erincatto April 11, 2026 07:04 View session

Copilot AI reviewed Apr 11, 2026

View reviewed changes

erincatto added 6 commits April 11, 2026 00:53

unused code

c83396c

Merge branch 'clustering' of https://github.com/erincatto/box2d into …

928dadc

…clustering

fix for unitialized stages

9d5f4d0

Merge branch 'clustering' of https://github.com/erincatto/box2d into …

57f3f2d

…clustering

fix for non-determinism

e05c263

restored unit tests

312436c

+	if ( b2LengthSquared( linearVelocity ) > 0.0f )
+	{
+		b2WakeBody( world, body );
+	}

		@@ -808,18 +750,7 @@ void b2Body_SetAngularVelocity( b2BodyId bodyId, float angularVelocity )
		return;
		}

+	if ( body->setIndex != b2_awakeSet )
+	{
+		float maxVelocity = b2AbsFloat( angularVelocity ) * body->maxExtent;
+		if ( maxVelocity >= body->sleepThreshold )
+		{
+			b2WakeBody( world, body );
+		}
+	}

	fprintf( file, "island sim: %d\n", islandSimCapacity * (int)sizeof( islandSimCapacity ) );
	fprintf( file, "island sim: %d\n", islandSimCapacity * (int)sizeof( b2IslandSim ) );

	#if 0
	// Optional SIMD contact solver implementation. Disabled by default unless explicitly enabled.
	#if defined( B2_ENABLE_SIMD_CONTACT_SOLVER ) && B2_ENABLE_SIMD_CONTACT_SOLVER

Uh oh!

Conversation

erincatto commented Apr 7, 2026

Uh oh!

Copilot AI left a comment

Choose a reason for hiding this comment

Pull request overview

Reviewed changes

Uh oh!

Copilot AI Apr 7, 2026

Choose a reason for hiding this comment

Uh oh!

Copilot AI Apr 7, 2026

Choose a reason for hiding this comment

Uh oh!

Copilot AI Apr 7, 2026

Choose a reason for hiding this comment

Uh oh!

Copilot AI Apr 7, 2026

Choose a reason for hiding this comment

Uh oh!

Copilot AI Apr 7, 2026

Choose a reason for hiding this comment

Uh oh!

Copilot AI Apr 7, 2026

Choose a reason for hiding this comment

Uh oh!

Copilot AI Apr 7, 2026

Choose a reason for hiding this comment

Uh oh!

Copilot AI Apr 7, 2026

Choose a reason for hiding this comment

Uh oh!

Copilot AI Apr 7, 2026

Choose a reason for hiding this comment

Uh oh!

Copilot AI left a comment

Choose a reason for hiding this comment

Pull request overview

Uh oh!

Copilot AI Apr 7, 2026

Choose a reason for hiding this comment

Uh oh!

Copilot AI Apr 7, 2026

Choose a reason for hiding this comment

Uh oh!

Copilot AI Apr 7, 2026

Choose a reason for hiding this comment

Uh oh!

Copilot AI Apr 7, 2026

Choose a reason for hiding this comment

Uh oh!

Copilot AI Apr 7, 2026

Choose a reason for hiding this comment

Uh oh!

Copilot AI Apr 7, 2026

Choose a reason for hiding this comment

Uh oh!

Copilot AI left a comment

Choose a reason for hiding this comment

Pull request overview

Uh oh!

Copilot AI Apr 11, 2026

Choose a reason for hiding this comment

Uh oh!

Copilot AI Apr 11, 2026

Choose a reason for hiding this comment

Uh oh!

Copilot AI Apr 11, 2026

Choose a reason for hiding this comment

Uh oh!

Copilot AI Apr 11, 2026

Choose a reason for hiding this comment

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development