event_manager_update (272 B) gated exact on first compile. code_000cb850
(768 B) needed three stacked levers: (1) the no-post arm is a NEGATED
else-if -- 'else if (!(crossed || crossed || repeat)) post = FALSE; else
{ post body }' -- which lays both xor-post blocks inline as fallthrough
stubs and keeps them un-cross-jumped, fixing the 16-byte size gap;
(2) 'boolean post = TRUE;' at declaration, so the live range crosses
system_milliseconds and post colors into BL, denying EBX to the time
local (January homes time in [ebp-4] and reloads at every use);
(3) one shared 'if (post) { insert }' tail instead of per-arm early
returns -- the type-0/3 paths constant-fold the test away, matching
January's straight jump into the insert.
Struct recoveries: event_record.data is a union (point2d stick /
{index,value} button), second field controller_index attested by
ui_widget_event_handler_functions' private copy; event_manager_globals
__unknown108 resolved as stick_event_times[2][4] +
previous_stick_axes[2][2][4] (0x108..0x167). Stick deadzone 29490
(0x7332), repeat window 250 ms, clamp writes via chained assignment
'x = event->data.stick.x = 32767' selected by a switch on
'x >= 0 ? 1 : -1'.
Inert alternatives measured: post declared before time (no move);
post = FALSE init at top (memory-homes post, restructures prologue);
guard-inverted arms without the negation (720 B collapse).
Board: 179/619 objects, 3,609 fns, 486,407 bytes.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Change static const short terminal_tab_stops[] to const short
rdata_002712d0[] so the symbol is external (storage 0x2) and
matches the target COFF's symbol name and linkage. The 6-byte
tab-stops array was previously reported as a .rdata mismatch
(96.36% of 176 bytes matching). All 13 code functions remain
unchanged and verified EQUAL via the hardened comparator.
Jonas Volman