fix(engine): retain Sakashima's own abilities when it copies a creature (#6009)

[nghetienhiep]

fix(engine): retain Sakashima's own abilities when it copies a creature (#6009)

Sakashima of a Thousand Faces reads "except it has Sakashima's other
abilities" — a CR 707.9a exception that keeps ALL of Sakashima's own
printed abilities (the legend-rule exemption static, Partner) on the
copy, not just a single indexed ability. The existing "except it has
this ability" handling only retains the one ability containing the
BecomeCopy effect, so Sakashima's legend-rule exemption and Partner
were silently dropped whenever it copied a creature.

Added RetainAllOtherAbilitiesFromSource, an unbounded sibling of
RetainPrintedTriggerFromSource/RetainPrintedAbilityFromSource that
pulls the source's entire activated/triggered/static/keyword surface
into the copy's Layer 1 copiable values instead of a single index.

Closes #6009

[gemini-code-assist]

Code Review

This pull request implements support for ContinuousModification::RetainAllOtherAbilitiesFromSource to correctly handle CR 707.9a (e.g., Sakashima of a Thousand Faces), ensuring that the entire ability surface of the source is retained on the copy. Feedback on the changes suggests avoiding cloning the retained_other_abilities tuple inside the hot path loop over affected_ids in layers.rs to prevent unnecessary allocations, recommending borrowing the tuple instead.

Important

The consumer version of Gemini Code Assist on GitHub is being sunset. Starting June 18, 2026, new organization installations will be blocked, and all code review activity will officially cease on July 17, 2026.
For more details on the timeline and next steps, please review the Help Documentation.

[gemini-code-assist]

Avoid cloning the entire retained_other_abilities tuple of vectors inside the loop over affected_ids. Since layer evaluation is a hot path, cloning these vectors for every affected object introduces unnecessary allocations. Instead, borrow the tuple and clone the individual elements only when they are actually pushed.

                if let Some((abilities, triggers, statics, keywords)) = &retained_other_abilities {
                    let obj_abilities = Arc::make_mut(&mut obj.abilities);
                    for ability in abilities.iter() {
                        if !obj_abilities.contains(ability) {
                            obj_abilities.push(ability.clone());
                        }
                    }
                    for trigger in triggers.iter() {
                        if !obj.trigger_definitions.iter_all().any(|t| t == trigger) {
                            obj.trigger_definitions.push(trigger.clone());
                        }
                    }
                    for static_def in statics.iter() {
                        if !obj.static_definitions.iter_all().any(|s| s == static_def) {
                            obj.static_definitions.push(static_def.clone());
                        }
                    }
                    for keyword in keywords {
                        if !obj.keywords.contains(keyword) {
                            obj.keywords.push(keyword.clone());
                        }
                    }
                }

[github-actions]

Parse changes introduced by this PR

✓ No card-parse changes detected.

[matthewevans]

Maintainer fixup applied; hold for one production regression.

I pushed dee203eabe to complete the new ContinuousModification match in phase-ai (the previous head did not compile) and to remove the per-affected-object clone in the Layer 1 loop identified in the review thread.

Before this can be approved, please add a regression that drives Sakashima's actual parsed as-enters replacement through the entry/choice pipeline and proves that selecting a copy preserves both Partner and the controller-scoped legend-rule exemption. The current tests validate parser shape and a hand-built ResolvedAbility, but do not exercise the production replacement-enter path.

Recommendation: add that pipeline regression, then request re-review on the resulting head.

[matthewevans]

Request changes — the current implementation still lacks a production-path regression.

The maintainer fixup restores the exhaustive AI match and removes the Layer 1 clone, but the PR still only proves constructed ability data. Please add a scenario using parsed Sakashima Oracle text that drives the replacement-enter/copy pipeline and verifies all retained source abilities through Layer 1. Until that test reaches the production path, this class remains unproven.

[matthewevans]

Approved: the current head exercises the parsed replacement-enter pipeline, retains the relevant source ability surface at Layer 1, and preserves that behavior in copiable values.