Aura Designer: a candidate-filter gate at the engine chokepoint, both container lanes - #252
Open
Maelareth wants to merge 2 commits into
Open
Aura Designer: a candidate-filter gate at the engine chokepoint, both container lanes#252Maelareth wants to merge 2 commits into
Maelareth wants to merge 2 commits into
Conversation
…, both container lanes Adds the machinery that lets a feature darken its own aura displays on a switch - built for the Power Infusion Helper (following PR) but written as engine capability. The chokepoint: every engine write of candidate filters funnels through one of two lanes. Group/overlay/row containers pass recordCandidateFilters, which substitutes a populated match-nothing map when the config carries the ownership mark and the gate is shut (or the unit's role is excluded). Stored config is never touched: a rebuild produces something already gated rather than something corrected afterwards, which removes the re-assert race a config-swapping design would carry against every future rebuild path. The slot lane: placed indicators (AcquireSlot) never passed that funnel - their map went straight from config into the engine at six sites, so a gated placed indicator was gated when it fell back to the per-indicator container (in combat) and ungated when the slot path served it. SlotHandle:_cf() derives the same substitution at read time; routed through it are the create-time declare, the tuning push, the cine and death latch re-parses, the recovery re-push, and the combat-exit replay. The one deliberate exception is the identity-gate vulnerability recompute, which must see the real selection or gated slots would leave the cinematic-latch population and return from a cutscene fail-open. The accessor is nil-faithful: never nil for a real selection (the recovery edge burns once and treats nil as nothing-to-repair), never a value for a nil one. Ownership is config.dfGate, stamped by whoever builds a config - one stampGate helper at every call site so a missed site reads as an absent line in a known list, plus self-stamps in the builders that already hold the effect config. A gateOwns predicate centralises which marks the gate owns. An earlier design carried a synthetic spell id in real filter data; rejected because profiles travel and get decoded, the resolved maps are cached and shared between consumers, and debuff groups carry no config-wide candidate set to mark. The broadcast walks both registries narrowed to owned entries - the native setters have no equality guard, so an unnarrowed walk would re-parse every container in the addon per flip. In lockdown the slot lane defers through the standard regen replay. Role exclusion resolves per unit at the same chokepoint (player's own role from spec, others from the group, failing open on NONE), with an exported predicate so non-container consumers (sound) can give the same answer.
- The gate broadcast skips parked slots: a parked slot renders nothing, so pushing candidates at it is pure work, and it stays in the registry by necessity because Restore has to find it again. Restore now re-pushes candidates through the accessor, so a gate edge that flipped while the slot was parked is picked up on un-park rather than lost. - The container debug dump reports the raw selection instead of the gated one. It read through the gating accessor, so a gated row logged the dead match-nothing placeholder rather than the user's real filter - on the one diagnostic line most likely to be read while debugging that exact gate. - The gate's combat-exit backstop registers its five events only while a consumer exists, matching the pattern its sibling watcher already used. Driven by a single "does a consumer exist" test so the two registrations cannot disagree.
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
1 participant
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
What
Engine capability that lets a feature darken its own aura displays on a switch, built for the Power Infusion Helper (follow-up PR) but written as general machinery. Two files:
Frames/AuraContainer.lua,AuraDesigner/Factory.lua.recordCandidateFilterssubstitutes a populated match-nothing map when a config carriesdfGateand the gate is shut (or the unit's role is excluded). Stored config is never touched: a rebuild produces something already gated, which removes the re-assert race a config-swapping design would carry against every future rebuild path (per the thread discussion and your sign-off on the chokepoint approach).SlotHandle:_cf()derives the same substitution at read time; routed through the create-time declare, the tuning push, both latch re-parses, the recovery re-push, and the combat-exit replay (five stash reads, not four — the replay was the dangerous one). The identity-gate vulnerability recompute deliberately stays raw, or gated slots would leave the cine-latch population and return from a cutscene fail-open. The accessor is nil-faithful by construction — the recovery edge burns once and treats nil as nothing-to-repair.config.dfGatevia onestampGatehelper at every call site (your forget-one condition), self-stamps where the builder already holds the effect config, and agateOwnspredicate centralising which marks the gate owns. The sentinel-in-filter-data design is gone for the reasons that surfaced while mapping: profiles travel, the resolved maps are cached and shared, and debuff groups had nothing to mark.Why
Closes the intermittency you called out: the same placed indicator was gated when built out of combat (the Create fallback) and ungated when the slot path served it — "plainly missing is fine; intermittent is a bug report we'd never reproduce."
Testing
Field-verified across two group sessions and solo rounds on PTR: the gate darkening and reopening across every display family including a square built mid-combat arriving at combat's end already gated; role exclusion per unit; ordinary (non-gated) placed indicators regression-checked unchanged. Full history on lab thread
d04fe999.