refactor(app): compose session and player startup

Move streaming, live-session, hydration, local-player, combat, and teleport construction behind the typed Phase-7 boundary. Add exact-owner runtime bindings and focused spawn-claim classification so partial startup rolls back without retaining old session targets while preserving the accepted construction and frame dependencies.

Co-authored-by: Codex <codex@openai.com>
This commit is contained in:
Erik 2026-07-22 18:28:32 +02:00
parent 3573da12e1
commit 7771c07fb6
31 changed files with 1759 additions and 532 deletions

View file

@ -97,15 +97,17 @@ runtime-target binding without replay, and one current/toon settings owner.
`GameWindow` is 3,663 lines / 162 fields / 37 methods at G. Checkpoint H adds
sole terrain-atlas and sky-shader roots, one retained Host/runtime lease, an
atomic update/render frame-root pair, prepared portal fallback/transfer, and
checked retryable GL leaf construction and state mutation. Checkpoint I.1I.6
checked retryable GL leaf construction and state mutation. Checkpoint I.1I.7
now provide the executable ordered startup pipeline through retained UI and
live presentation. The canonical live runtime/world, presentation/effect
controllers, selection/radar, landblock publishers, and portal/sky/particle
resources have transactional phase ownership and exact-owner late bindings.
`GameWindow` is now 2,840 raw lines. I.7I.9, shutdown, canonical soak
snapshots, and closeout checkpoints JL remain. The App gate passes 3,407 tests
/ 3 intentional skips, the complete Release suite passes 7,779 tests / 5
skips, and all three I.6 corrected-diff reviews are clean.
I.7 adds the complete streaming/session/hydration/local-player/teleport phase,
including explicit reversible ownership for every late edge. `GameWindow` is
now 2,479 raw lines. I.8I.9, shutdown, canonical soak snapshots, and closeout
checkpoints JL remain. The App gate passes 3,420 tests / 3 intentional skips,
the complete Release suite passes 7,792 tests / 5 skips, and all three I.7
corrected-diff reviews are clean.
This is a behavior-preserving structural program. Severe regressions still get
root-cause fixes in separate commits; ordinary feature work resumes with M4