persist attack through attacker transformation

[drbergman]

Rebased onto current development and inverted relative to this PR's original direction: transformation now keeps an in-progress attack running instead of ending it.

Background

• pAttackTarget lives in phenotype.cell_interactions, so phenotype = cd.phenotype in convert_to_cell_definition drops it.
• The other two halves of the attack link — the target's state.attacked_by and the shared spring — live in state, which transformation leaves alone.

Since #422 landed the symmetric attack link, convert_to_cell_definition resolved that mismatch by calling remove_self_from_attacked() to end the attack. This PR resolves it the other way: hold pAttackTarget across the assignment and put it back. All three parts of the link then stay consistent, and the spring is never touched.

This also makes transformation symmetric — attacks against a transforming cell already persisted, while attacks by it were ended.

total_damage_delivered is carried across too (second commit). It also lives in phenotype.cell_interactions, so transformation was resetting the attacker's lifetime damage tally to zero — visibly wrong now that the attack it was counting keeps running afterwards.

Verification

Built and ran interaction-sample with CD8+ T cell -> neutrophil transformation enabled and a 30 min attack duration. At each transformation of an attacking cell, all three parts of the link survive:

TRANSFORM_WITH_ATTACK now=24 newtype=neutrophil target=bacteria \
  pAttackTarget_kept=1 target_lists_me=1 spring_mine=1 spring_theirs=1

interactions and rules_sample are the only sample configs with both nonzero attack and transformation rates, and neither is SVG-diff-checked in CI, so no snapshot changes are expected.

Known gap

The same run shows the case worth flagging:

newtype=neutrophil attack_rate_vs_target=0 attack_damage_rate=1 attack_duration=0.1

The transformed cell keeps attacking a cell type its new definition has a zero attack rate against — an attack it could never have started — and with unset attack parameters the defaults (attack_damage_rate = 1.0, attack_duration = 30 min) let that run on.

Handled separately in #432, which warns about exactly this case without changing the simulation.

[drbergman]

Here's an example demonstrating the need. The attacker (grey cell) attacks the defender (red cell). The attacker transforms into an exhausted cell (yellow cell). Meanwhile, the defender is chemotaxing to the left and the spring attachment between the two persists even after the attack has ceased. I have also included the config and cells.csv to recreate from the template project.

https://github.com/user-attachments/assets/1ca37563-35df-4959-8374-b652a27bf8b4
PhysiCell_settings.txt
cells.csv

[Copilot]

Copilot reviewed 1 out of 1 changed files in this pull request and generated no comments.

Comments suppressed due to low confidence (1)

core/PhysiCell_cell.cpp:1134

• [nitpick] Consider using 'nullptr' instead of 'NULL' to align with modern C++ practices.

if ( phenotype.cell_interactions.pAttackTarget != NULL )

[drbergman]

Came across this again and adding some more concrete code details.

1. Attack target is part of the attacking cell's phenotype. See here
2. On transformation, only a subset of phenotype fields are protected from conversion. See here
3. This does not include phenotype.cell_interactions (again, this is where the pTarget info is stored)

Thus, the copied Cell_Definition& cd phenotype with a NULL pTarget erases our transformed cell's pTarget, but the this.state is unchanged, and this is where spring_attachments is located.

[drbergman]

Can we make attack a conserved vs not property?

[drbergman]

Need a new PR where the attack persists through transformation.