* fix: duel teams start on their own arena points, in formation
MoveInside clamped the team index with Math.Min instead of Math.Max, so
every team started on the west edge point. The formation rotation also
reused the already-rotated X when computing Y, stacking players on the
north and south edges, and the SW/NE corners swapped axes, which leaves
the corner formation pointing out of the arena.
* refactor: one start point index clamp for GetBaseStartPoint and MoveInside
The clamp was written twice, and only one copy was correct.
* refactor: leave the placement loop as it was
Only the rotation lines in the loop belong to the fix.
Fixes#2682. Thanks @DatMoshu for the detailed report and repro.
While fixing it I audited the other map, item and multi enumerators for the same class of defect. None repeats #2682 itself, but four others turned up. All are fixed here.
## Fixes
- **`StaticTileEnumerator` (#2682)**: `SetMulti` returned `SetTile()` for the first multi whose bounds contained the point. If that multi had no tile at the cell, the enumeration ended and every later multi in the sector was skipped. Movement and line of sight lost a floor that stood inside another multi's bounding box. It now moves on to the next multi.
- **`MultiBoundsEnumerator`**: deduplication used one static `HashSet` that every new query cleared. A query run inside another made the outer query skip multis or return them twice. Each multi is now yielded only from the sector that holds its `Location`; the bounds filter guarantees that sector is visited. A per-enumerator set covers a multi that isn't registered in that sector (components changed after placement). Normal queries allocate nothing.
- **`Item`/`Mobile`/`ClientDistanceEnumerator`**: ring sectors were bounded only by `sx >= _sectorStartX`. Ring cells past the map edge resolved to the shared invalid sector, once per cell. Entities at off-map coordinates were returned more than once, and sectors outside the query were created lazily. The check now covers all four sides. Results match the bounds enumerators, which never returned off-map entities.
- **`BaseBoat.MovingEntitiesEnumerator`**: `GetMovingEntities` handles a null map, but the enumerator constructor still called `boat.Map.GetMobilesInBounds` and threw. A boat with no map now gets empty enumerators.
- **`BaseDoor.ChainEnumerator`**: the walk stopped only on returning to the start door. Links are set in pairs, so `[ChainLink A B C` followed by `[Link B C` leaves A→B→C→B, and enumerating that chain never ended. The walk now stops on any door it has already yielded. Only reachable when `UseChainedFunctionality` is true, which is off in base and in all shipped content.
## Testing
- A regression test for each fix except the boat null guard. Each failed before its fix and passes after.
- Server.Tests 910/910. UOContent.Tests 1185 passed, 2 skipped.
- Test note: the `MultiComponentList(List<MultiTileEntry>)` constructor drops entries after the first whose `Flags` are 0. The new test multis use non-zero flags.
`TryStealItem` doesn't check for a locked container, so an item that can still be targeted inside one (the gump stays open when it gets locked, or the client targets by serial) is stolen without picking the lock.
It now refuses with 501747 ("It appears to be locked."); GMs are exempt, and stealing the locked container itself is unchanged. It checks the lock directly rather than calling `CheckLift`, which would also refuse items in deco containers.
## Summary
`Find_TwoPathfindsOverlappingChunks_PromoteToBuildOnSecondFind` fails intermittently in CI with `second Find through overlapping chunks must promote (got 0 builds)`.
**Cause:** `SecondTouchAfterWindow_RestartsTheCount` and `TwoFindsAcrossTheWindow_DoNotPromote` set `StepCache.Instance.MissPromotionWindowMs = 1` and never restore it, and `Clear()` does not reset it. Both classes share the `Sequential Pathfinding Tests` collection, so when a lifecycle test runs first, the promotion test's two Finds have to land within 1 ms of each other. On Linux, `Environment.TickCount` moves in coarse steps of several ms, so even a sub-millisecond test can see the window lapse. The second Find then restarts the miss count instead of promoting. On Windows the ~15.6 ms tick nearly always reads the same value for both Finds, which is why it didn't reproduce locally.
**Fix (tests only):**
- `StepCacheLifecycleTests` implements `IDisposable` and restores the window after each test.
- The promotion test sets the window it depends on (30 s) instead of inheriting it.
## Test plan
- [x] Reproduced deterministically: with a temporary 20 ms sleep between the two Finds, the test fails whenever the lifecycle tests run first and passes when run alone
- [x] With the sleep still in, the test passes with the Dispose restore alone and with both fixes
- [x] All pathfinding tests pass (123/123) with the sleep removed
## Summary
**Followers of a deleted character.** `PlayerMobile.OnDelete` deleted stabled pets and left every active follower pointing at the deleted mobile. Pets stayed controlled by nobody, and after a restart they loaded as owned pets with no owner. Now, when a character is deleted:
- bonded or dead pets and summons (controlled or not) are deleted;
- unbonded pets are released: they go wild and despawn on the abandoned-pet timer;
- escorts and hirelings are left alone, because they already notice a deleted master and walk off on their own.
**Familiars and animated dead that never unregistered.** `BaseCreature` unregistered them in `OnAfterDelete`, after `OnDelete` had already cleared the master. Both calls did nothing, so `SummonFamiliarSpell.Table` and the animated-dead registry kept deleted creatures until the caster's next cast. The unregisters now run in `OnDelete`, before the master is cleared.
## Test plan
- [x] `DeletedMasterTests` (6 cases, all failing on main): unbonded pet released; bonded and bonded-dead pets deleted; controlled and uncontrolled summons deleted; familiar unregistered on delete
- [x] UOContent.Tests (1184 passed, 2 skipped) and Server.Tests (907) green
