Integrate verified source and data progress; exclude unproven COMMON ownership, held UB/asm forms, and address-named data credit. Stable sweep: 23 exact functions gained, zero lost.
- encounters: _encounter_update_status (stay_active flag, direct
accumulations in January's store order), _encounters_test_activation
(January's non-short-circuit activation OR chain, HCEA-corroborated;
bound parent and link locals), and park reopens
_encounter_compute_activation_cluster_bit_vector (named parent locals,
store-once dormant desire) and _encounter_update_respawn (bounded
respawn while loop)
- network_connection: _network_connection_write (real
notify_traffic_event calls replace a hand copy; January's double
would-block loop test; single exit) and _network_connection_new
(single-exit success flag replaces the goto teardown; January assert
texts) - both park reopens
- _symbol_name_from_address: reads through symbol_table directly
Stable diff vs 9062b1e96: 0 regressions; active parks 195 -> 191.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
Owner-unblocked units, reviewed reconstructions, each strict exact in its
whole-TU gate with zero inherited losses, guard passing, no new owners:
- vehicles: _vehicle_update with cross/dot helpers on the object fields
(removes two invented static carriers), vehicle-type switch, PIN
- network_connection: read_unreliable/read_reliable single-exit bodies
with January's literals (including the '#d' typo) and a reset flag,
read assert text, get_address per-path clears, idle and
idle_server_reliable_endpoint (six park reopens)
- transport_endpoint_winsock: connect_endpoint_async, write_to_endpoint,
read_from_endpoint single-exit results (three park reopens)
- transport_endpoint_set_winsock: remove_endpoint_from_set single exit
Rejected: _poll_endpoint_set (negative-result clamp chosen to steer block
order), _add_endpoint_to_set (identical FD_SET branches),
_action_obey_command_begin (redundant store), _actor_action_try_to_dive
(units.h prototype contradicting its definition).
Stable diff vs 9062b1e96: 0 regressions; parks 250 -> 240.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
Owner lifted the reservation on the concurrent-work units for this
isolated lane. Per-function donor harvest (upstream punpckhdq PR50 and
the unmerged root/publish-opus5-safe-20260913 reconciliation), each
landed as a house-rule-clean minimum body, strict exact in its whole-TU
gate with zero inherited losses:
- transport_network_available: boolean result declared and initialized
from the link-status mask (January masks the byte result)
- transport_nonce_is_equal: boolean comparison of the memcmp result,
keeping the owner void const * prototype
- winsock_error_to_string: January's switch case order
- read_endpoint: single-exit result with one error store
- network_connection_connected/disconnect: boolean expressions over the
client-flag and endpoint tests
Donor steering forms were not imported (boolean truncation casts,
byte * parameters). Stable diff vs 9062b1e96: 0 regressions; parks
261 -> 255.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
Reconstruct _main_loop from the exact Claude packet using semantic helper names, authentic types, and owner-header prototypes. Keep collision timing declarations behind the collision debug/log boundary so units.obj remains exact.
