BFM-decomp/docs/cookbook-index.md

Cookbook index — find the entry by the SYMPTOM you are looking at

Generated by tools/cookbook_index.py — do not hand-edit (R33). Regenerate after adding a cookbook section.

docs/matching-cookbook.md is ~716 KB / 651 sections. Grepping it blind is how three P30 wave-1 agents each "discovered" an idiom that was already written down. Start here, then read the section. A section appears under every symptom it addresses.

How to use: name what you SEE in the diff (a stolen delay slot, an extra la, a swapped register pair, a conflicting types error), find that symptom below, read those sections first. If nothing fits, THEN grind — and add a section when you win.

Start here — the symptoms that come up most, with the section that fixes them

Hand-curated from what agents actually hit and, in several cases, RE-DERIVED because keyword matching alone did not surface the entry (P30 wave-2 feedback). If your symptom is here, read the named section before anything else.

• wrong branch sense / arms swapped / beq where the target has bne (match_one prints this class as BRANCH-POLARITY) → §3-T4 (invert the source condition) + §32.2; for a trailing return 0 vs an early return see the shared-ret0 note (cookbook L1344)
• conflicting types on YOUR OWN function's definition, where the fleet canon is (void) → §73 (the PARAMS axis) + §42 — keep the (void) signature and read the incoming arg via register s32 a0v __asm__("$4"); §42 is otherwise indexed only under regalloc
• a shared global declared at a conflicting type (u8 vs s32, signedness) blocking your draft → §37 asm-label alias extern T X __asm__("D_x") — beats *(T*)&X, whose address-of perturbs regalloc. Works for FUNCTION definitions too (P30 wave 2)
• target reuses ONE address register across two different offsets of the same global → §20 (the pointer-var-to-the-global bullet, cookbook L1907-1918) — take T *p = &D_x; and index off p, never the bare symbol twice
• an extra la / the address hoisted into a callee-saved register across calls → §20 + gcc-2.7.2-map/cse_expr.md §H — *(T*)&sym force_regs the address; the asm-label alias (§37) keeps the direct %lo mem form
• andi $x,0xFF folded away in your output but present in the target → §1/I2 + §12 name the family; if the prescribed & 0xff at the use folds, hold the masked byte in a u16 local so only a QI->HI extend survives (P30 wave 2, byte-tested)
• slti where the target has sltiu (or vice versa) → §35 — a separate SIGNED int copy of an unsigned load keeps slti; chained bounds get range-folded, so write each bound as its own if/goto (§21)
• you are about to hand-derive a body that some overlay already matched → §71 — grep the callees/globals for an already-matched SIBLING first; in wave 2 this alone produced iteration-1 MATCHes on 4 of 19 targets
• gcc stole an instruction into a branch delay slot that the target leaves as nop → §5a + the zero-byte __asm__("") fence (§34 toolkit) — reorg.c stop_search_p halts the eager filler on an asm insn
• void vs s32 return — is promoting it byte-neutral? → NO, not always: §41d (byte-proven; a void body with no return gains an instruction). P30 adds a second mechanism: an s32 return keeps $v0 live-out and blocks dbr from filling a loop-back delay slot
• match_one says SIZE-MISMATCH/short and the target has a frame-pointer prologue (addu $fp,$sp,$zero / 21F0A003) → The target is -O0. Pass --o0 to match_one/rtu_match — nothing else will ever match (§116: opt level is a property of the FILE). Known -O0 regions: boot, ov_SC01_077_o0*, the whale _o0b, and 0x8013B568..0x8013C98C — a target outside all four still needs checking
• same instruction multiset, one contiguous window, loads/registers ROTATED inside it → pure sched1 statement-order (gcc-2.7.2-map/sched.md §S1/§S4). Cheapest lever: brute-force it — N independent statements, script all N! orders through match_one (24 runs ≈ 2 min) instead of reasoning about rank_for_schedule
• a two-constant if/else or ?: result lands in $v1 where the target reuses the condition's $v0 → §76 — fold the condition into a NAMED local and overwrite that SAME variable with the two constants (an s16 temp, a fresh temp, an inner scope, and both ternary polarities all stay $v1; only reusing one s32 local coalesces onto $v0). §76 reads as behemoth-only; it is not
• a load HOISTED above a store (match_one may call it a WIDTH class) → usually §76 regalloc, not a width bug: a missing WAR dependency lets sched2 hoist it. Same fix — one variable for the compare temp and the result
• ori $v0,0xffd8 in yours vs addiu $v0,-0x28 in the target → a NEGATIVE constant stored into an unsigned narrow local materialises via zero-extended ori; make the local signed to get addiu. It does NOT cost you the lhu on readback — gcc-2.7.2 emits lhu for any plain HImode load feeding an sh (P30 wave 4, byte-tested)
• LENGTH-DRIFT −1 and the missing instruction is a reg-to-reg COPY in a jal delay slot → a narrow prototyped param (plain ANSI s16 arg3) — §43 is indexed as the K&R-definition lever, but the ANSI form is the fix as often
• your narrow load lost its load-delay nop right after an inline block move → an lwl/lwr+swl/swr pair is a delay-slot SPONGE for the following load — the inverse of the §5a/§34 fence case. Align-1 4×u8 struct assign emits the inline form (§38 covers only the -O0 memcpy-call form)
• your e = param_1 copy VANISHED (target addresses every field off a copy at a join block) → cse.c make_regs_eqv keeps a param copy only when the new pseudo's live range escapes the cse block AND outlives the param's last mention — reach that by REUSING the same variable in a later block. Copy-in-the-arms gets hoisted into the delay slot; copy-assigned-in-an-arm goes global allocno → callee-saved + 2 prologue insns
• ⚠️ the .run/ghidra_c/<fn>.c seed looks like a DIFFERENT function entirely → it may be — overlays share VAs, so a seed can be decompiled from another overlay mapped at the same address. Trust the target .s, not the seed (P30 wave 4)

By symptom

delay slots & branches (21)

• §3-T4 — Branch polarity: invert the source condition to flip gcc's chosen branch _L90
• §5a — Cross-jump tail-merge — gcc collapses two byte-identical blocks the original kept separate (FIX FOUND) _L211
• §3-D. — THE CROSS-JUMP RATCHET (the sharpest new trap — func_80131340 L-C) _L3508
• §3-C. — Inner-block declaration does NOT delay slot allocation — BYTE-REFUTED _L10068
• §162 — CROSS-JUMP DIRECTION: the surviving copy is always the LATER one, so a BACKWARD j into a sibling arm is a source goto (P30 S48) _L11266
• §162 — The cross-jump "CALL veto" is a COUNT law, not a CALL law (BOUNDS §88a; P30 S48, func_80189540) _L11302
• §177 — 🔴 THE EPILOGUE RETURN-DELAY SLOT IS DECIDED BY YOUR SAVED-REGISTER SET, NOT BY SCHEDULING _L17159
• §176-A — "SCHEDULE / DELAY-SLOT / LENGTH-DRIFT ±1" ⇒ check STATEMENT ORDER around the call first _L17616
• §186 — CROSS-JUMPING RUNS AFTER SCHEDULING, SO NO C-LEVEL BARRIER CAN STEER IT _L18063
• §194-H — §164-29's WAR fence runs again at SCHED2 ON HARD REGISTERS: an in-place v &= K before a branch fences the store that reads v, and when TWO independent stores compete for the one delay slot the fence picks the winner at ZERO length drift (a WIDTH/sh!=sw swap, not a nop). Corrects §167-42's reconstruction (pseudo arm → hard-reg arm) and supplies its first re-runnable A/B. _L19282
• §194-M — A STORE in a CONDITIONAL branch's delay slot proves its C statement DOMINATES the branch — reorg can never pull a store out of either thread (gcc-2.7.2, -mips1) _L19541
• §195-D — §195 — masked_diff.mask_for returns 0 for EVERY j/jal word, so an internal j destination is invisible to match_one, the permuter scorer AND every similarity tier: a control-flow semantic error (which calls execute) surfaces as a 1-instruction residual mislabelled DELAY-SLOT / profile=schedule, or as an outright false MATCH _L19852
• §195-E — A 0/1 materialised at a JOIN immediately before the controlling beqz/bnez proves the source NAMED the condition — a truth expression in an if's controlling position reaches do_jump, which has no value path _L19920
• §195-F — fold-const.c:4825 canonicalises a ?: whose THEN arm is zero (or constant against a non-constant ELSE) by inverting the condition and swapping the arms — so a ternary's written arm order is byte-inert there, and c ? 0 : X is unspellable: it always compiles as !c ? X : 0 _L19978
• §195-G — §NEW — TWO ARMS CALLING THE SAME CALLEE MERGE INTO ONE jal UNLESS EACH ARM'S OWN CODE CONSUMES THE RESULT: the C dial that keeps two call sites apart is WHERE the consumer test lives, and the branch SENSE you spell it with is byte-inert (jump.c:1737 canonicalises it) _L20030
• §195-J — GTE / PsyQ op CALL-vs-INLINE is a PER-SITE SOURCE FACT, not a TU style — both forms coexist in one TU (measured in 2 TUs), and the tell is the target's own opcodes (lwc2/sqr/swc2 vs jal), not the sibling. Costs -8 ins on func_8018505C. Corollary: the game's inline sqr macro emits TWO hazard nops, the SDK's Square0 body emits ONE — so the inline form is provably not Square0 inlined (a second instance of §187's SDK-vs-game GTE nop divergence). _L20190
• §195-N — In a call-bearing chain of N≥2 if (f(...)) return 1; tests closed by return 0;, the LAST test must stay in STATEMENT form — the value form (return f() != 0; / ? 1 : 0 / !!f()) costs +1 j and empties the other N−1 delay slots. The cause is REORG block placement, not jump.c's delete_jump. _L20382
• §197-B — A REPEATED COMPARE OF ONE VALUE AGAINST ONE CONSTANT IS DELETED BY cse's qty_comparison_code CHANNEL (the non-EQ complement of §165-03) — and a front-end-opaque mask on EITHER compare is a pure-C dial that keeps the target's second branch _L20582
• §199-F — §164-36b — THE TARGET-HEAD FENCE IS A DELAY-SLOT THREAD SELECTOR, AND IT ONLY FIRES WHEN mostly_true_jump > 0 (amendment to §164-36; its "−1 instruction" tell is falsified) _L20942
• §199-G — At TWO case nodes the switch-vs-if oracle is not blind — but the tell is ALL tests positive + a trailing j default, NOT the first test's polarity _L20996
• §201-C — §X — A CALL'S OWN DELAY SLOT AND THE UPSTREAM CONDITIONAL BRANCH'S SLOT COMPETE FOR ONE INSN (the call's argument copy), AND ONE STATEMENT'S POSITION RELATIVE TO THE CALL DECIDES BOTH — the residual is visible at the BRANCH, not at the call _L21233

instruction scheduling (36)

• §3-T2 — Source statement order drives instruction scheduling _L78
• §3 — When a diff is pure scheduling → decomp-permuter (harness built, Phase 6) _L107
• §10 — Closing the regalloc/scheduling hard tail by hand (LZSS, Phase 7 session F — the full close) _L835
• §25 — The "schedule" class is mostly COALESCING (pin-crackable), not scheduling; + the gate two-stage + h_exact over-counts ×134 (Phase 21, cont.6) _L2132
• §3-The — crack: the residual was gcc-2.7.2 COPY-COALESCING, not the instruction scheduler _L2144
• §3-The — genuine scheduler — rank_for_schedule (sched.c), for when it IS scheduling _L2160
• §3-The — genuine schedule WALLS (do NOT re-grind — stub) _L2169
• §32 — The region-a CAMERA-GIANT idiom set: struct-base hoisting + 4 sibling levers (Phase 24 T7, Fable5-cracked on func_80129CF8 191 ins, match_one MATCH; transferable to the 6 sibling giants) _L2422
• §34 — The func_80138ED0 giant crack: gcc-2.7.2's 3-qty sort bug + the zero-byte asm allocation toolkit + the giv-init fence (Phase 24 T5; Opus→close=21, Fable5→MATCH ×134) _L2454
• §35 — The region-a sibling-giant harvest: difficulty ≠ $s-reg count (it's global-array hoisting) + the banking recipe + new loop idioms (Phase 24 T7, 5 parallel Opus-Max agents) _L2471
• §3-A4 — SINK THE CONSUMER CALL INTO THE ARMS (the inverse of A1; func_8016AB6C, byte-proven) _L3444
• §49 — The LUID DIAL: a zero-byte SCHEDULING dial (the sched.c analogue of §47) — func_8017A4AC (536 ins ×134), Phase 26 session 8 _L3527
• §66d-5 — residual_class's "structural ⇒ permuter CPU is waste" is WRONG for schedule permutations (measured, Phase 29 SESSION-18) _L5491
• §3-The — attribution primitive (use this before calling anything a scheduling residual) _L6057
• §3-The — scheduling rules (refining §135-2 and §135-4) _L8912
• Consequence — for the family (a real scheduling decision) _L10095
• §148 — The loop.c hoisting THRESHOLD is arithmetic you can compute, and the ?: clamp that folds to MIN_EXPR (P30 S42, func_8017C6F4, 947 ins) _L10145
• §3-A. — move_movables hoists iff threshold × savings × lifetime ≥ insn_count — and you can read it _L10151
• §16Z — SHARPENS (sharpens §37 "the /s-DEP LATTICE", §136-13, §136-14, §16Xy, §162q, gcc-2.7.2-map/sched.md §64) _L14340
• §175 — A CALLER-SAVED REGISTER PIN CAN BE A CORRECTNESS BUG, NOT JUST A SCHEDULING CHOICE (P31 wave H, 2026-08-15) _L16712
• §177 — 🔴 THE EPILOGUE RETURN-DELAY SLOT IS DECIDED BY YOUR SAVED-REGISTER SET, NOT BY SCHEDULING _L17159
• §3-B. — A return <const> IS A PRIORITY-1 HARD-REG SET THE SCHEDULER PLACES FIRST (func_8001BE30, 92 ins) _L17244
• §176-A — "SCHEDULE / DELAY-SLOT / LENGTH-DRIFT ±1" ⇒ check STATEMENT ORDER around the call first _L17616
• §176-C — 🔴 WALL REFUTATION: a hard-register pin CANNOT schedule around a call, because of a genuine gcc-2.7.2 bug _L17680
• §186 — CROSS-JUMPING RUNS AFTER SCHEDULING, SO NO C-LEVEL BARRIER CAN STEER IT _L18063
• §193-C — gcc-2.7.2 cross_jump merges the SCHEDULED common SUFFIX only — there is no prefix/head merge, so §8/§48-A1's "duplicate into both arms and cross_jump refunds it" is a TAIL-only lever _L18518
• §193-D — A COMPILER-GENERATED ARGUMENT COPY IS A optimize_reg_copy_1 TRIGGER: when the pointer's LAST use in a block precedes a call that takes the pointer as a register, sched1 hoists the implicit move $aN,$sN to the block top and local-alloc re-bases the WHOLE block's memory operands onto $aN. The only C dial is a label (the §165-24 goto-join) between the block and the call. _L18552
• §193-F — §148-A2 — The threshold -= 3 STAIRCASE: move_movables admits a COUNT of invariants, not a boolean — two identical merged constants split hoisted/not-hoisted by LIST ORDER, and insn_count picks the rank cutoff _L18653
• §194-A — A zero-byte scheduling fence goes AFTER the defining statement to make that computation emit FIRST in its block — and the barrier predicate is volatile-or-colon-less, not the "memory" clobber (COMPLEMENTS §165-40; does not refute it) _L18921
• §194-D — Declared width of a computed-value local is a sched1 dial (count-neutral) — a replication + attribution correction of §165-17, NOT a new argument-register law _L19088
• §194-H — §164-29's WAR fence runs again at SCHED2 ON HARD REGISTERS: an in-place v &= K before a branch fences the store that reads v, and when TWO independent stores compete for the one delay slot the fence picks the winner at ZERO length drift (a WIDTH/sh!=sw swap, not a nop). Corrects §167-42's reconstruction (pseudo arm → hard-reg arm) and supplies its first re-runnable A/B. _L19282
• §194-K — Blind sched1's alias oracle with a second SET of a pointer pseudo — the first zero-byte, non-volatile, dependence-CREATING lever (corrects §167-05's "volatile is the only door"; fourth consumer of reg_n_sets) _L19415
• §194-N — §193-D's C dial is misstated: the lever is a SURVIVING CODE_LABEL (a label with a real incoming edge), not "a label between the block and the call" — a bare label, or a goto L; L: pair whose target is the next active insn, is deleted by jump1 (jump.c:663-669 → delete_insn → jump.c:3458-3461, and jump.c:243 for the bare case) long before sched1/local-alloc, and costs exactly zero bytes _L19596
• §195-D — §195 — masked_diff.mask_for returns 0 for EVERY j/jal word, so an internal j destination is invisible to match_one, the permuter scorer AND every similarity tier: a control-flow semantic error (which calls execute) surfaces as a 1-instruction residual mislabelled DELAY-SLOT / profile=schedule, or as an outright false MATCH _L19852
• §199-A — §189-A's asm→source inference is byte-FALSE: an interloper between a split constant's lui/ori is a SCHEDULE fact, not a source fact — and the separator is the BIRTHING BOOST, not a "priority floor" (§189-A's split-timing half survives; its "no statement order / no pin" absolute and the candidate's own forward-scheduler narrative both fall) _L20681
• §199-F — §164-36b — THE TARGET-HEAD FENCE IS A DELAY-SLOT THREAD SELECTOR, AND IT ONLY FIRES WHEN mostly_true_jump > 0 (amendment to §164-36; its "−1 instruction" tell is falsified) _L20942

register allocation & pins (68)

• §10 — Closing the regalloc/scheduling hard tail by hand (LZSS, Phase 7 session F — the full close) _L835
• Residual — A — commutative |/&/+ result lands in the wrong source-operand register _L856
• Register-allocation — ORDER (call-crossing $s0/$s1 swap) → FORCE it with register pins (byte-proven) _L1384
• §3-The — LOOSE-TYPING wall is real for narrow params (Phase 16, reconfirmed) _L1428
• §3-THE — CAP: at this tail the match_one→gate gap is the LOOSE-TYPING CALL-GRAPH wall (not cheap plumbing) _L1637
• §22 — DEF-side loose-typing recovery + grinder blacklist (Phase 21) _L2005
• §24 — The ov_SC01_077_a.c split-file vein: split-aware propagation, but loose-typing-gated (Phase 21) _L2085
• §25 — The "schedule" class is mostly COALESCING (pin-crackable), not scheduling; + the gate two-stage + h_exact over-counts ×134 (Phase 21, cont.6) _L2132
• §42 — The F-band ≤28 regalloc crack wave: register-pin/DENSITY levers beat the permuter (Phase 25 T7 F-band, 2026-07-10; Ultracode 9-worker wave, 4/9 banked byte-identical, 266 swept ×134) _L2900
• §43 — The K&R s16-param definition DISSOLVES the "narrow-param wall" for by-value register args (Phase 25 task A, Fable5 crack of the 369-ins giant func_80166994 ×134, 2026-07-11) _L3195
• §45 — The flagship func_80133CD4 crack (399 ins ×134): the merged-variable permutation-breaker + the 1-death local-alloc gate (Phase 25 task A giant escalation, Fable5 gdb-on-cc1, 2026-07-11) _L3298
• §47 — The live-length SLIDER: splitting a global.c allocno-priority TIE with one zero-byte asm (Phase 26 session 8, Fable5 Max, byte-proven on func_8017BEBC 952 ins ×113) _L3365
• §48 — The 12-core jr crack wave: the ALLOCNO-PRICING dials and the EBB rule (Phase 26 session 8, Ultracode, 9/12 MATCH first pass) _L3405
• §3-A. — ALLOCNO-PRICING DIALS — move a value into the register you want, byte-neutrally _L3413
• §52 — The WALKER-FAMILY skeleton: 6 regalloc-order levers + a deeply-characterized intrinsic wall (func_80178004, 165 ins ×134; Phase 26, Fable5, 2026-07-15) _L3913
• §52a — The regalloc sibling wave: new levers + two new wall classes (cheap-Opus applying §52, 2026-07-15) _L3959
• §52b — Sibling wave 2: more de-pin levers, a third wall class, and the match_one→whole-binary gap at scale (2026-07-15) _L3997
• §66d-1 — What transfers between giants is the LOOP, not the PIN _L5280
• §70 — The giv-init base register: walk the PARAMETER, not a copy of it (Phase 29 SESSION-18, func_801777BC) _L5619
• §71 — Before mapping a giant, look for an already-matched SIBLING beside it (Phase 29 SESSION-18, func_8017D960, 3,338 ins) _L5660
• §72 — A register __asm__ pin is a PREFERENCE, not a reservation (Phase 29 SESSION-18, func_8017F510) _L5726
• §74 — Auditing a pinned draft: the §72 hazard is CALLER-SAVED pins spanning a call, and only the disassembly can tell you (Phase 29 SESSION-19, func_8017D960 b2, 5 pins) _L5822
• §76 — The allocno CLASS (local vs global) is the dominant regalloc lever, and C reaches it ONLY through declaration scope and variable reuse (Phase 29 SESSION-19, behemoth #3 func_8017F510 1,511 ins, 97 → MATCH, pin-free) _L6028
• §78 — A LENGTH drift can be a register grant in disguise; and fold never leaves a literal first in an | chain (Phase 29 SESSION-19, behemoth #2 func_8017D960 3,338 ins, 1806 → 0, pin-free) _L6174
• §76 — confirmed at scale, and a pin nuance _L6251
• §80 — A do-not-re-buy entry is scoped to its BASE, not to the function; and the pin's hidden cost is an unconditional qty_phys_sugg (Phase 29 SESSION-19, func_8017BF14 45 → 0) _L6270
• §3-The — pin's hidden cost, with the citation _L6291
• §83 — The parameterised-repeat law, the spill-area trap, and why a per-case edit cannot move a per-case symptom (Phase 29 SESSION-20, func_80183814 5,122 ins, cold-ish start → 36 structural / 99.3%) _L6412
• §83c — TRAP: a "dead local" in a prior draft may be gcc's OWN spill area _L6445
• §86 — Pinned-exemplar templatability is a PER-FAMILY property, not a per-member rate; and the §42e pin guard is now over-conservative (Phase 29 SESSION-20) _L6589
• §3-Two — further notes worth keeping _L8248
• §137 — REGALLOC-PERM is a TWO-COMPILE ARITHMETIC PROBLEM, not a permuter job _L9344
• §3-The — same swallow, twice more, in the integration spine _L9709
• §3-B. — A ?: on MEMORY operands costs ~16 bytes of invisible frame; on REGISTER operands, zero _L10057
• §3-D. — A lone $t8/$t9 in the target is RELOAD SCRATCH — reproduce the spill, don't pin the register _L10073
• §3-C. — A zero-byte ALLOCNO-PRIORITY slider _L10184
• §150 — A register ROTATION across symmetric blocks is VARIABLE-IDENTITY evidence, not an allocator tie (P30 S43, func_8017C6F4, 947 ins ×4) _L10312
• §152 — BYTE SIZE is the family key that name- and h_seq-grouping both miss (P30 S43, the 0xECC family: 1 crack → 12 overlays → 11,364 ins) _L10418
• §155 — hi/lo literal scanning MUST track base registers (S45) _L10558
• §158 — The RANGE-EXTENDER: a fourth zero-emission asm lever completes the allocno toolkit (P30 S46 tier-3, func_8017CE58, 733 ins) _L10712
• Bonus — facts worth keeping _L10757
• §156 — THE PREFERENCE-DONOR MERGE: cross-region variable reuse is what fills a0-a3, and a call-arg use in ONE region steers the fill in ALL of them (P30 S46 tier-3, func_80186E24, 611 ins: 236-off "S11 regalloc-order" → MATCH, zero new pins) _L10772
• §175 — A CALLER-SAVED REGISTER PIN CAN BE A CORRECTNESS BUG, NOT JUST A SCHEDULING CHOICE (P31 wave H, 2026-08-15) _L16712
• §176j — STOPPING A WAVE MID-FLIGHT COSTS THE IN-FLIGHT TAIL (and how much is recoverable) _L17096
• §177 — 🔴 THE EPILOGUE RETURN-DELAY SLOT IS DECIDED BY YOUR SAVED-REGISTER SET, NOT BY SCHEDULING _L17159
• §179-A — 🔴 A LOOP-WALKED POINTER PARAMETER HANDS ITS ARGUMENT REGISTER TO THE GIV (9 byte-proofs) _L17313
• §179-F — PINNING A LOOP-WALKED POINTER IS A TOTAL OFF-SWITCH FOR STRENGTH REDUCTION _L17509
• §179-G — 🟡 A PIN CAN CREATE A COMBINE LOG_LINK AND DELETE AN andi (sixth RC-5 channel, n=1) _L17532
• §176 — SEVEN LEVERS FROM THE P31 OVERNIGHT WAVES (2026-08-15): statement order, false regalloc, and the pin that fights back _L17610
• §176-B — "REGALLOC-PERM, 1-4 instructions off" ⇒ it is usually NOT register allocation _L17640
• §176-C — 🔴 WALL REFUTATION: a hard-register pin CANNOT schedule around a call, because of a genuine gcc-2.7.2 bug _L17680
• §3-18 — of 20 reconciled while keeping the match. The two that did not are mechanism, not effort. _L17933
• §186 — CROSS-JUMPING RUNS AFTER SCHEDULING, SO NO C-LEVEL BARRIER CAN STEER IT _L18063
• §186c — WHERE A VALUE IS LOADED DECIDES WHICH ALLOCATOR OWNS IT, AND THEREFORE ITS REGISTER _L18092
• §193-B — A NARROW CAST OF A WIDE PARAMETER COSTS ONE EXTRA INSTRUCTION WHEN ITS STATEMENT SITS AFTER AN INTERVENING jal — AND THE DECIDER IS combine.c:929's CROSS-CALL GUARD, NOT REGISTER ALLOCATION _L18480
• §193-D — A COMPILER-GENERATED ARGUMENT COPY IS A optimize_reg_copy_1 TRIGGER: when the pointer's LAST use in a block precedes a call that takes the pointer as a register, sched1 hoists the implicit move $aN,$sN to the block top and local-alloc re-bases the WHOLE block's memory operands onto $aN. The only C dial is a label (the §165-24 goto-join) between the block and the call. _L18552
• §194-B — A addu $rA,$rB,$zero copy feeding ≥2 sh stores is a SECOND, 16-BIT-DECLARED local — the width, not the clamp or the join liveness, is what keeps the copy (and func_80183094's pin + "memory" clobber are both provably inert) _L18987
• §194-C — A CALLER-SAVED loop counter proves its live range crosses ZERO calls — so it cannot share a pseudo with any value LIVE ACROSS a call (but it may freely share one with values that merely sit between calls) _L19025
• §194-D — Declared width of a computed-value local is a sched1 dial (count-neutral) — a replication + attribution correction of §165-17, NOT a new argument-register law _L19088
• §194-G — Reading the integer half of a 16.16 stack aggregate: REGISTER-LIVENESS, not the C spelling, decides sra vs lhu — and the break is count-neutral _L19239
• §194-H — §164-29's WAR fence runs again at SCHED2 ON HARD REGISTERS: an in-place v &= K before a branch fences the store that reads v, and when TWO independent stores compete for the one delay slot the fence picks the winner at ZERO length drift (a WIDTH/sh!=sw swap, not a nop). Corrects §167-42's reconstruction (pseudo arm → hard-reg arm) and supplies its first re-runnable A/B. _L19282
• §194-N — §193-D's C dial is misstated: the lever is a SURVIVING CODE_LABEL (a label with a real incoming edge), not "a label between the block and the call" — a bare label, or a goto L; L: pair whose target is the next active insn, is deleted by jump1 (jump.c:663-669 → delete_insn → jump.c:3458-3461, and jump.c:243 for the bare case) long before sched1/local-alloc, and costs exactly zero bytes _L19596
• §195-B — A CALL_INSN does not start a basic block in gcc-2.7.2 — so a call-crossing temp can be a LOCAL-alloc quantity (the missing precondition under §48-A2 / §52 / regalloc.md K8) _L19762
• §195-F — fold-const.c:4825 canonicalises a ?: whose THEN arm is zero (or constant against a non-constant ELSE) by inverting the condition and swapping the arms — so a ternary's written arm order is byte-inert there, and c ? 0 : X is unspellable: it always compiles as !c ? X : 0 _L19978
• §197-C — Fix A1 (operand order) cannot move a commutative destination whose .greg conflict set already contains BOTH operand hard registers — split the accumulate so the destination IS the load's pseudo _L20610
• §199-A — §189-A's asm→source inference is byte-FALSE: an interloper between a split constant's lui/ori is a SCHEDULE fact, not a source fact — and the separator is the BIRTHING BOOST, not a "priority floor" (§189-A's split-timing half survives; its "no statement order / no pin" absolute and the candidate's own forward-scheduler narrative both fall) _L20681
• §199-B — A permutation sweep that holds ANY statement fixed is not a sweep: the statement an agent pins as "obviously load-bearing" is the one carrying the signal, and the partial sweep returns a FLAT residual that reads as proof of order-invariance _L20730
• §199-E — §189-A BOUNDED AND CORRECTED — the discriminator is INSN_PRIORITY, not "is the interloper a constant": an insn between a lui/ori pair proves NOTHING about the source spelling unless it TIES the ori on priority, and on the pinned -mcpu=3000 triple a dependent load never does _L20883

CSE / redundancy / rematerialization (17)

• §46 — The func_80178D40 crack (890 ins ×134, the heaviest core in the game): four LOOP-STRUCTURE levers cheap-Opus found by reading loop.c/jump.c/cse.c (Phase 26 session 8, 2026-07-13) _L3313
• §83d — CSE's quantity budget is WHOLE-FUNCTION, so a local rewrite cannot fix a local symptom _L6452
• §153 — THE ADDRESS-REMATERIALISATION LAUNDER: a third zero-emission asm lever (P30 S43, func_8018D98C, 710 ins) _L10466
• §176-D — CSE-class levers used in reverse (two sharpenings of §153 and cse_expr §2) _L17706
• §193-E — A varying-address (pointer) load is re-emitted once per CSE-LIVE INTERVAL, and naming it in a C local is the only C-level lever over that count — no store SPELLING has any reach _L18623
• §193-F — §148-A2 — The threshold -= 3 STAIRCASE: move_movables admits a COUNT of invariants, not a boolean — two identical merged constants split hoisted/not-hoisted by LIST ORDER, and insn_count picks the rank cutoff _L18653
• §193-H — A pointer-derived base load *(s32*)(p+K) is uncacheable across ANY memory write or call (cse's third kill disjunct) — so the target emits one lw per store-separated RUN of spellings, and a run of stores sharing one lw is a C local you must declare _L18778
• §194-A — A zero-byte scheduling fence goes AFTER the defining statement to make that computation emit FIRST in its block — and the barrier predicate is volatile-or-colon-less, not the "memory" clobber (COMPLEMENTS §165-40; does not refute it) _L18921
• §194-J — Back-to-back identical stores: flow.c's last_mem_set deletes the first, and only volatile saves it _L19363
• §194-K — Blind sched1's alias oracle with a second SET of a pointer pseudo — the first zero-byte, non-volatile, dependence-CREATING lever (corrects §167-05's "volatile is the only door"; fourth consumer of reg_n_sets) _L19415
• §195-H — §165-27g — ON A FIXED-SYMBOL GLOBAL, extern T D[]; + D[0] IS THE CSE RELOAD DIAL, AT ZERO ADDRESSING COST, AND THE ELEMENT COUNT IS INERT (§165-27's T v[2] CAVEAT IS A LOCAL-FRAME FACT AND DOES NOT TRANSFER) _L20078
• §195-I — §145(b) AMENDED — the pointer-bump addiu is saved by cse, not combine: ANY second SET of a pseudo in the address's equivalence chain (in-place p += K, §145(b)'s p = r; copy, or an asm re-tie) kills the fold; the pass is cse and the gate is invalidate's reg_tick++ _L20125
• §195-L — The cse store-re-seed does not cross a JOIN LABEL: per-arm stores + a join read keep the reload that one join store deletes (bounds §193-E BOUND 1/BOUND 3 with §48-B's EBB boundary) _L20286
• §197-A — A NARROW SIGNED MEMORY READ FEEDING A CONSTANT >> LOSES ITS lh, AND THE CURE IS AN ASM RE-TIE (attribution CONTESTED: cse vs combine) _L20544
• §197-B — A REPEATED COMPARE OF ONE VALUE AGAINST ONE CONSTANT IS DELETED BY cse's qty_comparison_code CHANNEL (the non-EQ complement of §165-03) — and a front-end-opaque mask on EITHER compare is a pure-C dial that keeps the target's second branch _L20582
• §199-D — A narrow UNSIGNED value compared in an ordered if emits sltiu/sltu; the only dial is the WIDTH of a real object, and a widening (s32) cast is inert — AMENDS §35 (whose stated "only CSE-reuse canonicalizes" mechanism is byte-wrong) and does NOT apply inside a switch _L20823
• §201-E — §194-J-2 — The last_mem_set deletion window is measured in SOURCE/expand order, not in the printed stream; a foreign store moved between the pair in C source is a real lever, and volatile is not always the better one _L21331

loops & induction variables (14)

• §3-T1 — Loop pointer: top-of-body for addu induction, not constant-folded addiu _L71
• §34 — The func_80138ED0 giant crack: gcc-2.7.2's 3-qty sort bug + the zero-byte asm allocation toolkit + the giv-init fence (Phase 24 T5; Opus→close=21, Fable5→MATCH ×134) _L2454
• §35 — The region-a sibling-giant harvest: difficulty ≠ $s-reg count (it's global-array hoisting) + the banking recipe + new loop idioms (Phase 24 T7, 5 parallel Opus-Max agents) _L2471
• §46 — The func_80178D40 crack (890 ins ×134, the heaviest core in the game): four LOOP-STRUCTURE levers cheap-Opus found by reading loop.c/jump.c/cse.c (Phase 26 session 8, 2026-07-13) _L3313
• §66d — The permuter⇄reader loop: alternate a random search with a byte-verified idiom, and let residual_class decide whose turn it is (Phase 29 SESSION-17) _L5245
• §66d-1 — What transfers between giants is the LOOP, not the PIN _L5280
• §70 — The giv-init base register: walk the PARAMETER, not a copy of it (Phase 29 SESSION-18, func_801777BC) _L5619
• §145 — Three loop/combine levers from the S40 wave-2 drafters (16/16 match_one) _L9927
• §148 — The loop.c hoisting THRESHOLD is arithmetic you can compute, and the ?: clamp that folds to MIN_EXPR (P30 S42, func_8017C6F4, 947 ins) _L10145
• §171a — THE MECHANICAL A-PROP DRAFT (P30 S50): 256 members banked with no agent in the loop _L16420
• §179-A — 🔴 A LOOP-WALKED POINTER PARAMETER HANDS ITS ARGUMENT REGISTER TO THE GIV (9 byte-proofs) _L17313
• §179-E — A >2*MAX_MOVE_BYTES BLOCK COPY IS A STRUCT ASSIGNMENT, NOT A HAND LOOP _L17483
• §179-F — PINNING A LOOP-WALKED POINTER IS A TOTAL OFF-SWITCH FOR STRENGTH REDUCTION _L17509
• §194-C — A CALLER-SAVED loop counter proves its live range crosses ZERO calls — so it cannot share a pseudo with any value LIVE ACROSS a call (but it may freely share one with values that merely sit between calls) _L19025

structs, block moves & memcpy (51)

• §3-T2 — Source statement order drives instruction scheduling _L78
• §5 — Known hard-residual classes (instruction-identical, one byte-exact blocker) _L199
• §13 — Add a location overlay — the canonical runbook (Phase 13; the Phase-15 fleet recipe) _L1056
• §15 — Struct-heavy shared-core pipeline (Phase 16) — empirical determinations (S0) _L1313
• §16 — Guided hand-matching the struct-heavy core (Phase 17 — beats the §15 brute-force) _L1335
• §3-The — crack: the residual was gcc-2.7.2 COPY-COALESCING, not the instruction scheduler _L2144
• §28 — Banking a "close=0 gate-rejected" giant: the canonical-extern recovery (Phase 22 T1, byte-proven on func_8015126C ×134) _L2325
• §28b — The struct-walled close=0 giant (§28 case #3) IS bankable: the engine_types.h type-lift (Phase 22 T2, byte-proven on func_80156B74 ×134) _L2357
• §30 — Fable5Max cracks §20/§10 "unsteerable" from the gcc SOURCE: store-vs-load is a /s aliasing flag, the def-side wall has a macro escape, + the birthing-boost (Phase 23, byte-proven on giant func_8014EE14 248 ins ×134) _L2382
• §32 — The region-a CAMERA-GIANT idiom set: struct-base hoisting + 4 sibling levers (Phase 24 T7, Fable5-cracked on func_80129CF8 191 ins, match_one MATCH; transferable to the 6 sibling giants) _L2422
• §38 — The WHALE func_80144B9C (770 ins): the -O0 struct-assign memcpy idiom + the -O0 reach-134 ×134 rollout (Phase 24 T7 §G, cheap Opus — no Fable5, no calls.c) _L2527
• §40 — Structural families: the MECHANICAL symbol-remap (crack one exemplar → remap the rest, ~0 tokens) (Phase 25 T3/T7, 2026-07-08) _L2553
• §41 — The DEF-SIDE canonical-sig wall: mechanically banking a drafted giant past conflicting types (Phase 25 T5b batch-2, 2026-07-09; tools/canon_sig_reconcile.py, byte-proven on func_8013B274) _L2710
• §42a — addendum — wave 2 (residuals + 29-100 band): iso-MATCH ≠ real-TU bank, the memcpy→struct-assign fix, +5 levers (2026-07-10b) _L2953
• §44 — The Phase-25 cheap-Opus giant batch: 5 structural levers + the §43 extension (2026-07-11, 6 crackers over the frontier giants) _L3242
• §46 — The func_80178D40 crack (890 ins ×134, the heaviest core in the game): four LOOP-STRUCTURE levers cheap-Opus found by reading loop.c/jump.c/cse.c (Phase 26 session 8, 2026-07-13) _L3313
• §3-Why — 0/8 was structural, and predictable from two words _L4045
• §54 — --fix-def-sig: the member's CANONICAL DECLARATION is a build step too (tiny-IMM mega-pools, +4,801, Phase 29 T6, 2026-07-16) _L4163
• §56b — PROPAGATING an h_seq giant: the exemplar's externs MUST be fleet-canonical, not the draft's types _L4232
• §65a — The blast-radius taxonomy (makes §61's law structural instead of remembered) _L5030
• §66d-5 — residual_class's "structural ⇒ permuter CPU is waste" is WRONG for schedule permutations (measured, Phase 29 SESSION-18) _L5491
• §83 — The parameterised-repeat law, the spill-area trap, and why a per-case edit cannot move a per-case symptom (Phase 29 SESSION-20, func_80183814 5,122 ins, cold-ish start → 36 structural / 99.3%) _L6412
• §84 — The DERIVED-OFFSET remap bug: a hand-computed literal that encodes the DISTANCE between two per-overlay symbols, and why match_one is structurally blind to it (Phase 29 SESSION-20, func_8013D53C) _L6472
• §3-The — construct _L6477
• §91 — A structure-TRANSFER is only valid where the structure corresponds: the --like role trap (Phase 29 SESSION-21, func_8012AAAC ×137) _L6842
• §3-The — three-hypothesis trail, because two of them were wrong and the wrongness is instructive _L6871
• §129 — Post-carve, rtu_match/match_one COUNT THE JUMP TABLE AS INSTRUCTIONS; and a carve must never be committed without its owner (P30 S28, func_8013BD74) _L8437
• §129a — the target instruction count is INFLATED after a carve _L8441
• §3-The — DEFINITION-side alias is the only escape when the fleet canon disagrees on a promoting param _L9473
• §3-Two — errors of mine, both instructive _L10009
• §160 — THE REACH-15 WAVE HARVEST: an align-1 block move, and the instrument that called it a wall (P30 S47) _L10893
• §16Xy — SHARPENS (sharpens §136d-3, §37 /s-DEP LATTICE, §135-2, §136-13, §162q) _L12300
• §3-The — /s drop clause is in ALL THREE dependence predicates, so the FIXED-ADDRESS STORE is what floats (P30 S48, func_8017CA18, ov_MAIN_012) _L12302
• §16Z — SHARPENS (sharpens §37 "the /s-DEP LATTICE", §136-13, §136-14, §16Xy, §162q, gcc-2.7.2-map/sched.md §64) _L14340
• §3-The — ADDRESS-CLASS TABLE: which load/store pairs even REACH the /s clause (P30 S48 wave 4, func_80185B44, ov_SC03_014) _L14342
• §176g — SIZE A WAVE BY INSTRUCTIONS, NOT BY CARDS (P31 S52 — the adopted doctrine) _L16930
• §3-F. — MEM_IN_STRUCT_P ASYMMETRY IN true_dependence (func_80037144, 124 ins) _L17281
• §179-E — A >2*MAX_MOVE_BYTES BLOCK COPY IS A STRUCT ASSIGNMENT, NOT A HAND LOOP _L17483
• §176-B — "REGALLOC-PERM, 1-4 instructions off" ⇒ it is usually NOT register allocation _L17640
• §189 — FIVE COMPILER LAWS MINED FROM THE WAVE R/S JOURNALS (P31 S53), each source-cited and re-derived by a second agent _L18197
• §193-A — The wave card ships only OPEN-set pointers and discards the atlas's BANKED one — exemplar/sibs are stubs 100% by construction (tools/atlas.py:96/657), while seed.ref (matched pool, atlas.py:505-536) is dropped at build_wave_atlas.py:143 _L18413
• §193-B — A NARROW CAST OF A WIDE PARAMETER COSTS ONE EXTRA INSTRUCTION WHEN ITS STATEMENT SITS AFTER AN INTERVENING jal — AND THE DECIDER IS combine.c:929's CROSS-CALL GUARD, NOT REGISTER ALLOCATION _L18480
• §193-I — A DECLARED AGGREGATE LOCAL HAS AN 8-BYTE FRAME STRIDE — CEIL(size,8), NOT size. N ARRAYS THEREFORE COST 8·k MORE THAN ONE HAND-LAID STRUCT (k = how many of them have size mod 8 ≠ 0), AND THE TELL IS IDENTICAL INSTRUCTION COUNT WITH EVERY $sp DISPLACEMENT SHIFTED BY THE SAME CONSTANT. _L18818
• §194-E — The wave card's exemplar is the TARGET ITSELF on 42/73 wave-U and 36/71 wave-T cards — a construction consequence of atlas.py:657 (exemplar = max-nins OPEN member) meeting build_wave_atlas.py:166 (one card per gid); and the shipped §193-A seed_ref is same-binary on 0/51, so no card field can ever name a destination-TU sibling _L19129
• §194-H — §164-29's WAR fence runs again at SCHED2 ON HARD REGISTERS: an in-place v &= K before a branch fences the store that reads v, and when TWO independent stores compete for the one delay slot the fence picks the winner at ZERO length drift (a WIDTH/sh!=sw swap, not a nop). Corrects §167-42's reconstruction (pseudo arm → hard-reg arm) and supplies its first re-runnable A/B. _L19282
• §195-D — §195 — masked_diff.mask_for returns 0 for EVERY j/jal word, so an internal j destination is invisible to match_one, the permuter scorer AND every similarity tier: a control-flow semantic error (which calls execute) surfaces as a 1-instruction residual mislabelled DELAY-SLOT / profile=schedule, or as an outright false MATCH _L19852
• §195-F — fold-const.c:4825 canonicalises a ?: whose THEN arm is zero (or constant against a non-constant ELSE) by inverting the condition and swapping the arms — so a ternary's written arm order is byte-inert there, and c ? 0 : X is unspellable: it always compiles as !c ? X : 0 _L19978
• §195-G — §NEW — TWO ARMS CALLING THE SAME CALLEE MERGE INTO ONE jal UNLESS EACH ARM'S OWN CODE CONSUMES THE RESULT: the C dial that keeps two call sites apart is WHERE the consumer test lives, and the branch SENSE you spell it with is byte-inert (jump.c:1737 canonicalises it) _L20030
• §195-K — At -O2 the §18 array-of-struct lever is a FRAME lever, not a length lever — but only when the N reads carry DISTINCT index expressions; with a SHARED index §18's +2-instruction residual is still alive at -O2 (the submitted "memory-loaded narrow index" precondition and unconditional length-neutrality are both falsified) _L20234
• §199-D — A narrow UNSIGNED value compared in an ordered if emits sltiu/sltu; the only dial is the WIDTH of a real object, and a widening (s32) cast is inert — AMENDS §35 (whose stated "only CSE-reuse canonicalizes" mechanism is byte-wrong) and does NOT apply inside a switch _L20823
• §199-F — §164-36b — THE TARGET-HEAD FENCE IS A DELAY-SLOT THREAD SELECTOR, AND IT ONLY FIRES WHEN mostly_true_jump > 0 (amendment to §164-36; its "−1 instruction" tell is falsified) _L20942

types, signedness & load/store width (49)

• §3-I1 — Unsigned range check: (x - lo) < (hi-lo) → addiu+sltiu _L41
• §3-I2 — Byte mask forces andi even after lbu _L47
• §3-T3 — Types _L85
• §7 — PsyQ SDK types & symbols (the library-call prerequisite) _L303
• §14c — Callee-signature-aware harvest — breaking the extern-type-conflict wall (Phase 15, T6) _L1213
• Phase-20 — RESOLUTION — tools/cast_call_sites.py BUILT + the cap re-diagnosed (R14, byte-proven) _L1701
• §23 — Giant func_80153E00 cracked (scalar-data CAST); BUT most giant drafts are STALE+INCOMPLETE, and a diag MUST remove artifacts (Phase 21 — byte-proven + a self-correction) _L2032
• §28b — The struct-walled close=0 giant (§28 case #3) IS bankable: the engine_types.h type-lift (Phase 22 T2, byte-proven on func_80156B74 ×134) _L2357
• §33 — Automating the giant decl-reconcile: tools/reconcile_decls.py (the DATA analog of §20's cast_call_sites) + a fleet-majority type oracle (Phase 24 T7b, byte-proven on func_80129CF8) _L2434
• §40a — The DECL-RECONCILE pass (type-lift so remapped drafts compile in the sibling TU) (Phase 25 T7.2, 2026-07-08) _L2601
• §41 — The DEF-SIDE canonical-sig wall: mechanically banking a drafted giant past conflicting types (Phase 25 T5b batch-2, 2026-07-09; tools/canon_sig_reconcile.py, byte-proven on func_8013B274) _L2710
• §41d — void→s32 is NOT always byte-neutral: gate the RAW draft FIRST (Phase 26 session 6, byte-proven) _L2878
• §42b — addendum — wave 3 (Max, 2026-07-10c): THE STALE-OBJECT GATE TRAP + the read-global &-cast drift + fix _L2994
• §43 — The K&R s16-param definition DISSOLVES the "narrow-param wall" for by-value register args (Phase 25 task A, Fable5 crack of the 369-ins giant func_80166994 ×134, 2026-07-11) _L3195
• §3-C. — TYPE- AND SHAPE-DRIVEN CODEGEN (the C type literally selects the addressing mode) _L3484
• §56b — PROPAGATING an h_seq giant: the exemplar's externs MUST be fleet-canonical, not the draft's types _L4232
• §64 — The §20 type-lift's three laws: fold the tagged typedef, check VISIBILITY, and strip only what is TEXTUALLY IDENTICAL (lift_types.py, Phase 29 SESSION-14, 2026-07-23) _L4908
• §64a — VARIANT types: UNIQUIFY the camps, do not reconcile them (uniquify_type.py, Phase 29 SESSION-14) _L4957
• §65g — Where the cheap levers STOP: the local-type and in-TU-self-decl classes did not yield _L5139
• §68 — A comment-only line halted the extern scan, and the skip label blamed the type cap (Phase 29 SESSION-18) _L5524
• §73 — A def-side self-decl conflict has TWO axes: RETURN (fleet widen, T2) and PARAMS (casts at each use, T0). Diagnose which before paying for the expensive one (Phase 29 SESSION-19, func_8014F3E8 + func_8014D4C0) _L5783
• §75c — Class-B's remedy is the FULL §17a-1 PAIR (decl and call-site cast); a decl-only fix moves the error and looks like a new wall (Phase 29 SESSION-19, func_8012F14C) _L5997
• §3-The — defect this exposed: a shared type that is present but invisible _L6354
• §92 — Conforming a DECLARATION: pointer changes are caller-neutral, scalar-WIDTH changes are not (Phase 29 SESSION-21, tools/conform_decls.py) _L6891
• §94 — A family sweep's 0/N is a TYPE-CARRY failure until proven otherwise: lift the exemplar's local types, ALL of them, transitively (Phase 29 SESSION-21, func_8016B6BC 0/137 → 137/137) _L6947
• §100 — Prefer the DRAFT-LOCAL fix: a type only one function uses belongs in its BODY, not in a shared header (Phase 29 SESSION-22, func_80175DA8 0/137 → 137/137) _L7216
• §120 — Uniquify draft-defined TYPE names; and check which of N staging sites you actually patched (Phase 29 T93) _L8026
• §3-Do — NOT "strip the duplicate typedef" — it breaks the extern that uses it _L8037
• §3-The — type-form rules _L8879
• §143 — cast_call_sites read a RETURN STATEMENT as a prototype and deleted it. A 0/39 sweep became 18/39. (P30 S40) _L9834
• Then — propagation returned 0/137 TWICE — both times a missing TYPE _L10000
• §154 — Reading a disc payload: the module-id word, static base derivation, and "type 1 = uncompressed overlay" (P30 S44) _L10522
• §3-C. — PAC type 1 = the same payload class as type 4, just NOT compressed _L10548
• §155b — check the TYPE your oracle returns before comparing against it (S45 p5) _L10587
• §3-B. — Typedef handling — the only strategy that survives contact _L16994
• §3-D. — A NARROW TYPE BLOCKS COPY ELISION (func_8001D3FC — new idiom) _L17265
• §185 — EDIT THE SIDE THAT IS CHEAP TO VERIFY, AND CHECK A TU RETYPE AT ITS USE SITES _L18031
• §186b — A NO-SAVE 16-BYTE FRAME IN A LEAF FUNCTION MEANS s16 LOCALS, NOT A HIDDEN CALL _L18084
• §193-B — A NARROW CAST OF A WIDE PARAMETER COSTS ONE EXTRA INSTRUCTION WHEN ITS STATEMENT SITS AFTER AN INTERVENING jal — AND THE DECIDER IS combine.c:929's CROSS-CALL GUARD, NOT REGISTER ALLOCATION _L18480
• §193-G — §164-54's "scope to ≥4 arms" bound is byte-wrong — the dispatch-topology oracle goes live at THREE case nodes (balance_case_nodes splits at i > 2), but only for a signed-after-promotion index _L18717
• §193-H — A pointer-derived base load *(s32*)(p+K) is uncacheable across ANY memory write or call (cse's third kill disjunct) — so the target emits one lw per store-separated RUN of spellings, and a run of stores sharing one lw is a C local you must declare _L18778
• §194-B — A addu $rA,$rB,$zero copy feeding ≥2 sh stores is a SECOND, 16-BIT-DECLARED local — the width, not the clamp or the join liveness, is what keeps the copy (and func_80183094's pin + "memory" clobber are both provably inert) _L18987
• §194-D — Declared width of a computed-value local is a sched1 dial (count-neutral) — a replication + attribution correction of §165-17, NOT a new argument-register law _L19088
• §194-G — Reading the integer half of a 16.16 stack aggregate: REGISTER-LIVENESS, not the C spelling, decides sra vs lhu — and the break is count-neutral _L19239
• §194-H — §164-29's WAR fence runs again at SCHED2 ON HARD REGISTERS: an in-place v &= K before a branch fences the store that reads v, and when TWO independent stores compete for the one delay slot the fence picks the winner at ZERO length drift (a WIDTH/sh!=sw swap, not a nop). Corrects §167-42's reconstruction (pseudo arm → hard-reg arm) and supplies its first re-runnable A/B. _L19282
• §194-I — §16N+2's magic-per-odd-part ladder has exactly one broken row — read the divisor arithmetically instead: d = round(2^(32 + post_shift) / magic_read_as_unsigned) _L19320
• §197-A — A NARROW SIGNED MEMORY READ FEEDING A CONSTANT >> LOSES ITS lh, AND THE CURE IS AN ASM RE-TIE (attribution CONTESTED: cse vs combine) _L20544
• §199-D — A narrow UNSIGNED value compared in an ordered if emits sltiu/sltu; the only dial is the WIDTH of a real object, and a widening (s32) cast is inert — AMENDS §35 (whose stated "only CSE-reuse canonicalizes" mechanism is byte-wrong) and does NOT apply inside a switch _L20823
• §201-D — THE SIGNEDNESS OF A div/mod MAGIC IS DECIDED BY THE STATIC TYPE OF THE DIVIDEND TREE AFTER get_narrower STRIPS WIDENING CONVERSIONS — NEVER BY A PROVABLE RANGE. AN & 0xFF IS NOT A CONVERSION, SO IT NEVER FLIPS THE MAGIC; A DECLARED-UNSIGNED LOCAL OR A NARROWING CAST WRITTEN AT THE DIVIDE BOTH DO. _L21284

declarations, prototypes & K&R (75)

• §3-T4 — Branch polarity: invert the source condition to flip gcc's chosen branch _L90
• §8c — Splitting a TU means rebuilding its DECLARATION ENVIRONMENT, not moving text (Phase 26 session 6) _L437
• §8d — Templating a body INTO a TU must not CHANGE its declaration environment — demote the carried data externs (Phase 26 session 8, byte-proven on func_8015AE2C ×133) _L483
• §13 — Add a location overlay — the canonical runbook (Phase 13; the Phase-15 fleet recipe) _L1056
• §14c — Callee-signature-aware harvest — breaking the extern-type-conflict wall (Phase 15, T6) _L1213
• §3-The — pipeline gotcha — match_one ≠ the gate; sig_unify is MANDATORY _L1421
• §3-The — gate two-stage — sig_unify is a FALLBACK, not unconditional (gate_stage.py, §19 folded in) _L2189
• §28 — Banking a "close=0 gate-rejected" giant: the canonical-extern recovery (Phase 22 T1, byte-proven on func_8015126C ×134) _L2325
• §28d — The "macro-extern-injection" lever: freeing reach-134 inline matches dedup_propagate skips as "not self-contained" (Phase 23, tools/inject_capped_externs.py) _L2366
• §33 — Automating the giant decl-reconcile: tools/reconcile_decls.py (the DATA analog of §20's cast_call_sites) + a fleet-majority type oracle (Phase 24 T7b, byte-proven on func_80129CF8) _L2434
• §40a — The DECL-RECONCILE pass (type-lift so remapped drafts compile in the sibling TU) (Phase 25 T7.2, 2026-07-08) _L2601
• §41 — The DEF-SIDE canonical-sig wall: mechanically banking a drafted giant past conflicting types (Phase 25 T5b batch-2, 2026-07-09; tools/canon_sig_reconcile.py, byte-proven on func_8013B274) _L2710
• §41b — addendum — M4 "reconcile_decls" tier is ALSO a probe over-count: 0/8 mechanical (Phase 25 T7-M4, 2026-07-10) _L2858
• §41d — void→s32 is NOT always byte-neutral: gate the RAW draft FIRST (Phase 26 session 6, byte-proven) _L2878
• §43 — The K&R s16-param definition DISSOLVES the "narrow-param wall" for by-value register args (Phase 25 task A, Fable5 crack of the 369-ins giant func_80166994 ×134, 2026-07-11) _L3195
• §51g — When the thing you are scanning has a GRAMMAR, parse the grammar (tools/cdecl.py) _L3769
• §54 — --fix-def-sig: the member's CANONICAL DECLARATION is a build step too (tiny-IMM mega-pools, +4,801, Phase 29 T6, 2026-07-16) _L4163
• §56 — Banking a hand-drafted GIANT into its exemplar TU: self-contained draft vs live-TU decls, 4 reconciliation tactics (func_8013FAF8, 312 ins ×138, Phase 29 T4, 2026-07-17) _L4187
• §56b — PROPAGATING an h_seq giant: the exemplar's externs MUST be fleet-canonical, not the draft's types _L4232
• §57 — The SELF-decl normalize: the sibling's OWN caller declares the templated fn divergently (the third §17a-1 direction; tools/normalize_self_decls.py, func_801670E4 ×137, Phase 29, 2026-07-18) _L4256
• §61 — Task 14: the gate ladder's missing stage is the ARITY pre-pass, and it is TU-side not draft-side (Phase 29, 2026-07-21) _L4545
• §63 — The fresh-138 DEF-SIDE blocker: fix the HEADER decl, not the draft (fix_header_decl.py, Phase 29 SESSION-13, 2026-07-23) _L4877
• §63 — UPDATE (Phase 29 SESSION-14) — fix_header_decl's "SAFE" verdict is FLEET-BLIND; it MUST be R22-validated _L5000
• §65c — rtu_match MATCH → bank held 13/13 on self-decl, but broke on the FIRST callee-decl case _L5081
• §65g — Where the cheap levers STOP: the local-type and in-TU-self-decl classes did not yield _L5139
• §66a — The widest write in a pipeline is the one most likely to be UNDECLARED _L5181
• §68 — A comment-only line halted the extern scan, and the skip label blamed the type cap (Phase 29 SESSION-18) _L5524
• §73 — A def-side self-decl conflict has TWO axes: RETURN (fleet widen, T2) and PARAMS (casts at each use, T0). Diagnose which before paying for the expensive one (Phase 29 SESSION-19, func_8014F3E8 + func_8014D4C0) _L5783
• §75 — A propagation cap is usually a MINORITY-SPELLING SOURCE OVERLAY, not a wall: census the carried extern before believing the exclusion message (Phase 29 SESSION-19, func_8014F3E8 ×4 → ×138) _L5866
• §75b — A body's preamble can carry #defines, not just externs; extraction lifts only the externs (Phase 29 SESSION-19, func_80165CA0 ×3 → fleet) _L5948
• §75c — Class-B's remedy is the FULL §17a-1 PAIR (decl and call-site cast); a decl-only fix moves the error and looks like a new wall (Phase 29 SESSION-19, func_8012F14C) _L5997
• §76 — The allocno CLASS (local vs global) is the dominant regalloc lever, and C reaches it ONLY through declaration scope and variable reuse (Phase 29 SESSION-19, behemoth #3 func_8017F510 1,511 ins, 97 → MATCH, pin-free) _L6028
• §79 — For a 0-callee giant, fingerprint by DATA symbols (§71 cannot fire); and the STACK-SLOT ORDER is a declaration-order oracle (Phase 29 SESSION-19, func_8017BF14 4,763 ins, cold start → 45/4763) _L6220
• §3-NEW — LEVER — the frame layout reads back the original declaration order _L6240
• §3-1. — The inlined-helper signature _L6375
• §92 — Conforming a DECLARATION: pointer changes are caller-neutral, scalar-WIDTH changes are not (Phase 29 SESSION-21, tools/conform_decls.py) _L6891
• §95 — reconcile_tu dropped the SIBLING declarators of a multi-symbol extern line (Phase 29 SESSION-21, func_80176218) _L6983
• §96 — The same rewrite, one shape down: reconcile_tu matched statements to lines by TEXT, so every COMMENTED declaration was silently skipped (Phase 29 SESSION-22, func_80176218 banked) _L7018
• §98 — conform_decls had three defects, and only the third needed R22 to find (Phase 29 SESSION-22, func_8014CF04) _L7113
• §99 — The narrow-param wall is a DEF-side problem with a ZERO-blast-radius fix: convert the definition to K&R (Phase 29 SESSION-22, func_80175AB8 + func_80175DA8) _L7167
• §102 — A PLUMBING verdict can MASK a DIFF; and K&R is not always a codegen change (Phase 29 SESSION-22, func_8016EC0C) _L7272
• §103 — A FILE-scope extern in a shared overlay TU is a GLOBAL constraint on every LATER function; move the DECL, not the draft (Phase 29 T48/T51, func_80135260 — the fleet-wide half) _L7307
• §104 — Two silent-skip defects in one scan: match on MASKED text, emit from the ORIGINAL (Phase 29 T53, gather_externs) _L7395
• §112 — A macro-scoped declaration only collides where the macro is INSTANTIATED (Phase 29 T67/T69, audit_header_sigs.py) _L7730
• §113 — An ARITY blocker only exists if the macro CALLS the function; an address-taken use has no call site (Phase 29 T72) _L7776
• §114 — The THIRD decl axis: a CALLEE the draft declares differently from the target TU (Phase 29 T76/T77) _L7808
• §3-Do — NOT "strip the duplicate typedef" — it breaks the extern that uses it _L8037
• §121 — Synthesise externs for macro-DEFINED callees from the macro's own definition head (Phase 29 T95) _L8061
• §3-The — declaration surface (integration, not codegen) _L8941
• §136f — Two declaration sub-cases the reconcile lane surfaced (lane now 15/15 lifetime) _L9186
• §138 — The propagation lanes: a gate refusal is a DECLARATION, and which lever you owe depends on blast radius _L9415
• Reconciling — a gate-refused draft: which way you edit depends on WHERE the TU's decl is _L9565
• §141 — The §134 class is CLOSED: every line-shape decision now routes through cdecl._mask (P30 S39) _L9740
• §143 — cast_call_sites read a RETURN STATEMENT as a prototype and deleted it. A 0/39 sweep became 18/39. (P30 S40) _L9834
• §3-C. — Inner-block declaration does NOT delay slot allocation — BYTE-REFUTED _L10068
• §159 — THE DECLARATION AXIS: conform to byte-truth, and make every guard state its COVERAGE (P30 S47; ~10,930 sites across 8 axes, fleet byte-identical) _L10828
• §168 — THE COUSIN TIER (P30 S49, 2026-08-12): h_seq's exact-hash brittleness, measured — and the similarity map above it _L16211
• §176f — THE DECLARATION FORM IS A MATCHING LEVER, SO RECONCILE TOWARD THE FORM THE MATCH NEEDS (P31 S52) _L16899
• §183 — THE DECLARATION-RECONCILIATION PLAYBOOK (P31 S53, measured on 20 byte-verified drafts) _L17932
• §193-H — A pointer-derived base load *(s32*)(p+K) is uncacheable across ANY memory write or call (cse's third kill disjunct) — so the target emits one lw per store-separated RUN of spellings, and a run of stores sharing one lw is a C local you must declare _L18778
• §193-I — A DECLARED AGGREGATE LOCAL HAS AN 8-BYTE FRAME STRIDE — CEIL(size,8), NOT size. N ARRAYS THEREFORE COST 8·k MORE THAN ONE HAND-LAID STRUCT (k = how many of them have size mod 8 ≠ 0), AND THE TELL IS IDENTICAL INSTRUCTION COUNT WITH EVERY $sp DISPLACEMENT SHIFTED BY THE SAME CONSTANT. _L18818
• §194-B — A addu $rA,$rB,$zero copy feeding ≥2 sh stores is a SECOND, 16-BIT-DECLARED local — the width, not the clamp or the join liveness, is what keeps the copy (and func_80183094's pin + "memory" clobber are both provably inert) _L18987
• §194-D — Declared width of a computed-value local is a sched1 dial (count-neutral) — a replication + attribution correction of §165-17, NOT a new argument-register law _L19088
• §195-A — §167-08's "an $aN READ before the jal is scratch" has a byte-proven FALSE-NEGATIVE class: an argument that DIES at the call is allocated straight into $aN, so its only def is a plain load far above the jal and every intervening use reads $aN — there is no positive tell in either direction, only the two-arity A/B _L19691
• §195-C — A call-argument %hi/%lo pair sitting at the block head, far above its jal, is a load-delay-gap filler chosen by SOURCE STATEMENT ORDER — swap the two independent statements nearest the call; a single-use void *p = &SYM; call-arg temp is OUTPUT-inert against it (but NOT expand-stream-inert) _L19804
• §195-D — §195 — masked_diff.mask_for returns 0 for EVERY j/jal word, so an internal j destination is invisible to match_one, the permuter scorer AND every similarity tier: a control-flow semantic error (which calls execute) surfaces as a 1-instruction residual mislabelled DELAY-SLOT / profile=schedule, or as an outright false MATCH _L19852
• §195-F — fold-const.c:4825 canonicalises a ?: whose THEN arm is zero (or constant against a non-constant ELSE) by inverting the condition and swapping the arms — so a ternary's written arm order is byte-inert there, and c ? 0 : X is unspellable: it always compiles as !c ? X : 0 _L19978
• §195-G — §NEW — TWO ARMS CALLING THE SAME CALLEE MERGE INTO ONE jal UNLESS EACH ARM'S OWN CODE CONSUMES THE RESULT: the C dial that keeps two call sites apart is WHERE the consumer test lives, and the branch SENSE you spell it with is byte-inert (jump.c:1737 canonicalises it) _L20030
• §195-H — §165-27g — ON A FIXED-SYMBOL GLOBAL, extern T D[]; + D[0] IS THE CSE RELOAD DIAL, AT ZERO ADDRESSING COST, AND THE ELEMENT COUNT IS INERT (§165-27's T v[2] CAVEAT IS A LOCAL-FRAME FACT AND DOES NOT TRANSFER) _L20078
• §196 — PUT ON THE CARD WHAT THE TREE ALREADY KNOWS: the fleet's declaration consensus (P31 S54) _L20475
• §199-D — A narrow UNSIGNED value compared in an ordered if emits sltiu/sltu; the only dial is the WIDTH of a real object, and a widening (s32) cast is inert — AMENDS §35 (whose stated "only CSE-reuse canonicalizes" mechanism is byte-wrong) and does NOT apply inside a switch _L20823
• §199-G — At TWO case nodes the switch-vs-if oracle is not blind — but the tell is ALL tests positive + a trailing j default, NOT the first test's polarity _L20996
• §200 — THE ALIAS IS THE UNIVERSAL DECLARATION ESCAPE: stop negotiating with the TU's spelling (P31 S55) _L21062
• §201-A — §150-B applies to decl_prior's DEF row: for an overlay-window symbol the banked "definition" is usually another overlay's function, and the card ranks it ABOVE the destination TU _L21121
• §201-D — THE SIGNEDNESS OF A div/mod MAGIC IS DECIDED BY THE STATIC TYPE OF THE DIVIDEND TREE AFTER get_narrower STRIPS WIDENING CONVERSIONS — NEVER BY A PROVABLE RANGE. AN & 0xFF IS NOT A CONVERSION, SO IT NEVER FLIPS THE MAGIC; A DECLARED-UNSIGNED LOCAL OR A NARROWING CAST WRITTEN AT THE DIVIDE BOTH DO. _L21284

jump tables & switches (31)

• §8 — rodata island (compiler jump tables) — the .data→.rodata→.data sandwich (Phase 7) _L320
• §8a — rodata island in a flat OVERLAY — the tail sandwich, per matched jr-function (Phase 26 — PoC PROVEN) _L342
• §8a-pad — a trailing .word 0x00000000 under a jtbl dlabel is .align PAD, not an entry (Phase 26 session 6, byte-proven) _L380
• §8b — MULTI-jtbl per overlay — the ld_interleave --order sandwich + the same-subseg cases (Phase 26 session 4) _L404
• §8e — The jtbl ALIGNMENT LAW + the pad-spec filter — multi-table .rodata spans (Phase 29, byte-proven; .run/probe_jtbl/verdict.md) _L530
• §3-The — clean object-level metric (use this, not the permuter score, for jtbl/rodata functions) _L844
• §41b — T7 execution: the object-only probe OVER-counts BANKABLE by two link/rodata classes (Phase 25 T7-M1, 2026-07-10) _L2809
• §48 — The 12-core jr crack wave: the ALLOCNO-PRICING dials and the EBB rule (Phase 26 session 8, Ultracode, 9/12 MATCH first pass) _L3405
• §53 — SWEEP A FAMILY WITH THE TOOL ITS EXEMPLAR NEEDED: the jr/switch carve, and how omitting it manufactured the "families don't template" doctrine (Phase 28 T1, 2026-07-15) _L4028
• §61b — The jtbl gate stage: built, and the ORDERING law it exposed (Phase 29 Task-14 stage 4, 2026-07-21) _L4658
• §61c — The jtbl bank is INCREMENTALLY valid and CLEAN-INVALID (Phase 29, 2026-07-21) — the blocking finding _L4713
• §62 — The jtbl RECONCILE must also follow the carve: the post-carve draft reconcile (harvest_verify._jtbl_reconcile, Phase 29 SESSION-11, 2026-07-22) _L4833
• §81 — Banking a jr (jump-table) function: the 3-step carve chain, and why match_one cannot see the problem (Phase 29 SESSION-19, func_8017C954) _L6323
• §88d — BANKING ORDER: run the §81 carve chain BEFORE banking, never after _L6701
• §97 — The gate's own tree hygiene: a refused carve, an unchecked recovery, and a snapshot that captured a dirty tree (Phase 29 SESSION-22) _L7066
• §105 — A gate's revert must survive an EXCEPTION, not just a failure (Phase 29 T53, jtbl_family_bank) _L7431
• §125 — Split the CARVE from the BODY before calling a jr residue a wall — and measure it by SHA from a CLEAN tree (P30 SESSION-28; this section's first draft was WRONG and the method caught it) _L8206
• §126 — The carve-within-a-carve: an ADDRESS RANGE is not an OPTIMIZATION REGION (P30 T2, byte-proven end-to-end) _L8264
• §129 — Post-carve, rtu_match/match_one COUNT THE JUMP TABLE AS INSTRUCTIONS; and a carve must never be committed without its owner (P30 S28, func_8013BD74) _L8437
• §129a — the target instruction count is INFLATED after a carve _L8441
• §129b — never commit a carve whose owner is still a stub (it strands the carve) _L8462
• §130 — An INCREMENTAL build can report BYTE-IDENTICAL for a change the CLEAN build cannot even LINK (P30 S28, the jr pair) _L8486
• §131 — The jtbl OVER-SPAN: sltiu N is ground truth in BOTH directions, and the zero-word rule only guards one (P30 S28, func_80191C50) _L8533
• §132 — The JR-PAIR-IN-ONE-O0-OBJECT "wall" was TWO instrument defects: a merged-double span the carve could not see, and a truncated object no rule deleted (P30 S29, func_8013B83C + func_8013BD74) _L8570
• §132b — When the span's already-matched owner is ITSELF multi-switch: --span-rel (P30 S1, func_8014032C) _L8660
• §139 — A GATE THAT GREPS FOR VERDICTS MUST ASSERT 1:1 ACCOUNTING; and a --src filter must not survive a carve (P30 S38, wave 6: 10 of 16 drafts vanished) _L9594
• §161 — THE RETRY-WAVE HARVEST: a jump table indexed from zero, and two allocator traps (P30 S47) _L10969
• §179-F — PINNING A LOOP-WALKED POINTER IS A TOTAL OFF-SWITCH FOR STRENGTH REDUCTION _L17509
• §188 — 🔴 THE jr $ra + addiu $sp TAIL IS AN ASSEMBLER ARTIFACT, NOT A FRAME SHAPE _L18147
• §199-D — A narrow UNSIGNED value compared in an ordered if emits sltiu/sltu; the only dial is the WIDTH of a real object, and a widening (s32) cast is inert — AMENDS §35 (whose stated "only CSE-reuse canonicalizes" mechanism is byte-wrong) and does NOT apply inside a switch _L20823
• §199-G — At TWO case nodes the switch-vs-if oracle is not blind — but the tell is ALL tests positive + a trailing j default, NOT the first test's polarity _L20996

optimisation level (-O0/-O2) (13)

• §6 — Per-module optimization mixing — the -O0 boot module (Phase 7) _L246
• Detecting — the opt level (do this first) _L254
• Build — mechanism — per-file opt override (splat resegmentation) _L288
• §18 — Per-file -O0 split inside an overlay/blob (Phase 19 T1) _L1519
• §38 — The WHALE func_80144B9C (770 ins): the -O0 struct-assign memcpy idiom + the -O0 reach-134 ×134 rollout (Phase 24 T7 §G, cheap Opus — no Fable5, no calls.c) _L2527
• §39 — The ×1→×134 giant-endgame: propagate a matched -O2 giant via the NATIVE DEFINE-macro path (Phase 24 T7 §G close, 2026-07-08) _L2537
• §116 — Optimization level is a property of the FILE, not the function: read a family 0/N against the member's stub HOME (Phase 29 T79) _L7877
• §3-The — finding: opt level is per FILE, so the file's contents must be opt-HOMOGENEOUS _L8279
• §127 — The -O0 regime: the CONSTANT-OFFSET FOLD, and why -O0 needs its own idiom set (P30 T3 wave, 15 targets) _L8349
• §3-The — rest of the -O0 regime (write PLAIN C, and mean it) _L8367
• §127a — §71 (sibling-first) is the strongest -O0 lever, and it beats the index _L8377
• §132 — The JR-PAIR-IN-ONE-O0-OBJECT "wall" was TWO instrument defects: a merged-double span the carve could not see, and a truncated object no rule deleted (P30 S29, func_8013B83C + func_8013BD74) _L8570
• §195-K — At -O2 the §18 array-of-struct lever is a FRAME lever, not a length lever — but only when the N reads carry DISTINCT index expressions; with a SHARED index §18's +2-instruction residual is still alive at -O2 (the submitted "memory-loaded narrow index" precondition and unconditional length-neutrality are both falsified) _L20234

family propagation & sweeps (91)

• §8d — Templating a body INTO a TU must not CHANGE its declaration environment — demote the carried data externs (Phase 26 session 8, byte-proven on func_8015AE2C ×133) _L483
• §11 — Cross-binary dedup & code-sharing (Phase 11 — "one match unlocks many") _L908
• §3-The — mechanism: game-code dedup is SOURCE-LEVEL, not an object swap (R-D1, the key lesson) _L926
• §3-The — flat-blob overlay config (what the template encodes) _L1071
• §3-THE — NON-4-ALIGNED-OVERLAY GOTCHA (≈75% of the fleet; fixed in the template + Makefile, automatic) _L1079
• §14 — Propagate a matched function across the fleet — tools/dedup_propagate.py (Phase 15) _L1126
• §19 — Scaling the toolkit waves — the recovery PIPELINE + the propagation CAP (Phase 19 T3) _L1597
• §24 — The ov_SC01_077_a.c split-file vein: split-aware propagation, but loose-typing-gated (Phase 21) _L2085
• §25 — The "schedule" class is mostly COALESCING (pin-crackable), not scheduling; + the gate two-stage + h_exact over-counts ×134 (Phase 21, cont.6) _L2132
• §3-h — _exact OVER-COUNTS ×134 — verify shareability before crediting a class's "reach-134" (R14) _L2177
• Where — this leaves the reach-134 tail (cont.6 option-3, now CONFIRMED byte-backed) _L2233
• §28 — Banking a "close=0 gate-rejected" giant: the canonical-extern recovery (Phase 22 T1, byte-proven on func_8015126C ×134) _L2325
• §28c — The close=0 recovery is NOT fully exhausted (§26 corrected, R14); + the dedup_propagate registry-skip recovery (Phase 22 T2) _L2352
• §28b — The struct-walled close=0 giant (§28 case #3) IS bankable: the engine_types.h type-lift (Phase 22 T2, byte-proven on func_80156B74 ×134) _L2357
• §28d — The "macro-extern-injection" lever: freeing reach-134 inline matches dedup_propagate skips as "not self-contained" (Phase 23, tools/inject_capped_externs.py) _L2366
• §30 — Fable5Max cracks §20/§10 "unsteerable" from the gcc SOURCE: store-vs-load is a /s aliasing flag, the def-side wall has a macro escape, + the birthing-boost (Phase 23, byte-proven on giant func_8014EE14 248 ins ×134) _L2382
• §34 — The func_80138ED0 giant crack: gcc-2.7.2's 3-qty sort bug + the zero-byte asm allocation toolkit + the giv-init fence (Phase 24 T5; Opus→close=21, Fable5→MATCH ×134) _L2454
• §38 — The WHALE func_80144B9C (770 ins): the -O0 struct-assign memcpy idiom + the -O0 reach-134 ×134 rollout (Phase 24 T7 §G, cheap Opus — no Fable5, no calls.c) _L2527
• §39 — The ×1→×134 giant-endgame: propagate a matched -O2 giant via the NATIVE DEFINE-macro path (Phase 24 T7 §G close, 2026-07-08) _L2537
• §40 — Structural families: the MECHANICAL symbol-remap (crack one exemplar → remap the rest, ~0 tokens) (Phase 25 T3/T7, 2026-07-08) _L2553
• §40a — The DECL-RECONCILE pass (type-lift so remapped drafts compile in the sibling TU) (Phase 25 T7.2, 2026-07-08) _L2601
• §40b — The reloc-tracker blind spot: the indexed-global idiom that hid the "reach-1 tail" (Phase 26 Task 1, 2026-07-11, byte-verified V0/V1) _L2634
• §40c — The h_seq per-sibling reconcile: templating a reconcile-class crack ×134 (Phase 26 Task 8, 2026-07-12, byte-proven) _L2676
• §41c — T7-M2: the ×134 def-side-wall sweep via per-sibling RE-reconcile (Phase 25, 2026-07-10; 4,389 banks, ~0 agent tokens) _L2839
• §42 — The F-band ≤28 regalloc crack wave: register-pin/DENSITY levers beat the permuter (Phase 25 T7 F-band, 2026-07-10; Ultracode 9-worker wave, 4/9 banked byte-identical, 266 swept ×134) _L2900
• §42e — propagating a CRACK ×134: the def-finder bug + the byte-drift residual (the "remap-fail" misdiagnosis) _L3121
• §43 — The K&R s16-param definition DISSOLVES the "narrow-param wall" for by-value register args (Phase 25 task A, Fable5 crack of the 369-ins giant func_80166994 ×134, 2026-07-11) _L3195
• §45 — The flagship func_80133CD4 crack (399 ins ×134): the merged-variable permutation-breaker + the 1-death local-alloc gate (Phase 25 task A giant escalation, Fable5 gdb-on-cc1, 2026-07-11) _L3298
• §46 — The func_80178D40 crack (890 ins ×134, the heaviest core in the game): four LOOP-STRUCTURE levers cheap-Opus found by reading loop.c/jump.c/cse.c (Phase 26 session 8, 2026-07-13) _L3313
• §47 — The live-length SLIDER: splitting a global.c allocno-priority TIE with one zero-byte asm (Phase 26 session 8, Fable5 Max, byte-proven on func_8017BEBC 952 ins ×113) _L3365
• §49 — The LUID DIAL: a zero-byte SCHEDULING dial (the sched.c analogue of §47) — func_8017A4AC (536 ins ×134), Phase 26 session 8 _L3527
• §52 — The WALKER-FAMILY skeleton: 6 regalloc-order levers + a deeply-characterized intrinsic wall (func_80178004, 165 ins ×134; Phase 26, Fable5, 2026-07-15) _L3913
• §3-The — 6 levers (the "walker-family skeleton" — apply to the 12 siblings; levers 1-5 retire ~half the residual) _L3924
• §53 — SWEEP A FAMILY WITH THE TOOL ITS EXEMPLAR NEEDED: the jr/switch carve, and how omitting it manufactured the "families don't template" doctrine (Phase 28 T1, 2026-07-15) _L4028
• §55c — Sizing the propagate: a TARGETED propagate is ~4 min/core, and "it's slow" was a BROKEN-TREE ARTIFACT _L4141
• §54 — --fix-def-sig: the member's CANONICAL DECLARATION is a build step too (tiny-IMM mega-pools, +4,801, Phase 29 T6, 2026-07-16) _L4163
• §56 — Banking a hand-drafted GIANT into its exemplar TU: self-contained draft vs live-TU decls, 4 reconciliation tactics (func_8013FAF8, 312 ins ×138, Phase 29 T4, 2026-07-17) _L4187
• §56b — PROPAGATING an h_seq giant: the exemplar's externs MUST be fleet-canonical, not the draft's types _L4232
• §57 — The SELF-decl normalize: the sibling's OWN caller declares the templated fn divergently (the third §17a-1 direction; tools/normalize_self_decls.py, func_801670E4 ×137, Phase 29, 2026-07-18) _L4256
• §57a — Two NSD corrections + the SURGICAL-ONLY law + the honest broad-sweep yield (Phase 29, 2026-07-18) _L4303
• §59 — Three h_seq sweep-residual classes match_one/the-exemplar-bank don't reveal (Phase 29 crack-wave close, 2026-07-18) _L4377
• §3-Two — corollaries worth remembering _L4450
• §65a — The blast-radius taxonomy (makes §61's law structural instead of remembered) _L5030
• §3-Giv — record order (the §70 family) _L5760
• §75 — A propagation cap is usually a MINORITY-SPELLING SOURCE OVERLAY, not a wall: census the carried extern before believing the exclusion message (Phase 29 SESSION-19, func_8014F3E8 ×4 → ×138) _L5866
• §75a — The exclusion classes, enumerated with named causes (Phase 29 SESSION-19, the 134-binary dedup_extend sweep) _L5925
• §75b — A body's preamble can carry #defines, not just externs; extraction lifts only the externs (Phase 29 SESSION-19, func_80165CA0 ×3 → fleet) _L5948
• §76 — The allocno CLASS (local vs global) is the dominant regalloc lever, and C reaches it ONLY through declaration scope and variable reuse (Phase 29 SESSION-19, behemoth #3 func_8017F510 1,511 ins, 97 → MATCH, pin-free) _L6028
• §77 — Every extraction tool carries a NARROW hard-coded preamble set; anything outside it silently caps the body's reach. Diff the preamble before you gate. (Phase 29 SESSION-19 — three variants in one session, two different tools) _L6084
• §84 — The DERIVED-OFFSET remap bug: a hand-computed literal that encodes the DISTANCE between two per-overlay symbols, and why match_one is structurally blind to it (Phase 29 SESSION-20, func_8013D53C) _L6472
• §86 — Pinned-exemplar templatability is a PER-FAMILY property, not a per-member rate; and the §42e pin guard is now over-conservative (Phase 29 SESSION-20) _L6589
• §3-THE — LAW: all-or-nothing PER FAMILY _L6600
• §3-Why — the two live families differ from the three dead ones — the open question _L6627
• §3-The — cheap discriminator, before spending a sweep _L6666
• §89b — the parallel gate farm existed; the family path could not reach it (tools/sweep_parallel.py) _L6745
• §91 — A structure-TRANSFER is only valid where the structure corresponds: the --like role trap (Phase 29 SESSION-21, func_8012AAAC ×137) _L6842
• §94 — A family sweep's 0/N is a TYPE-CARRY failure until proven otherwise: lift the exemplar's local types, ALL of them, transitively (Phase 29 SESSION-21, func_8016B6BC 0/137 → 137/137) _L6947
• §105 — A gate's revert must survive an EXCEPTION, not just a failure (Phase 29 T53, jtbl_family_bank) _L7431
• §107 — A lever wired into ONE gate path is a lever most families cannot reach (Phase 29 T56, func_80144090 0/136 → 136/136) _L7513
• §108 — Diagnosing a family 0/N: the four causes, and the third opt-in lever (Phase 29 T59) _L7551
• §111 — The distinct-code metric is not noisy: a family pays it only if its members are byte-VARIANT (Phase 29 T66) _L7681
• §116 — Optimization level is a property of the FILE, not the function: read a family 0/N against the member's stub HOME (Phase 29 T79) _L7877
• §123 — PROPAGATE A FAMILY WITH THE TOOL ITS TIER NEEDS: dedup_propagate is h_exact-only; its refusals are statements about the TOOL (P30 wave 1, 2026-07-30) _L8116
• §124 — A "not matched" verdict can mean the definition is there under a DIFFERENT C NAME: the asm-label alias blind spot (P30 SESSION-28, func_8016191C ×137) _L8153
• §124a — a family sweep's 0 matched-exemplar families may be a FILTER, not a wall _L8198
• §134 — MULTI-LINE BLINDNESS: one root cause, four faces, in family_remap's preamble scanner (P30 S6, 190 zero-crack families) _L8711
• §136c — SIBLING-FIRST is a DERIVATION shortcut, not just a conflict fix (the fastest route in a family wave) _L9069
• §138 — The propagation lanes: a gate refusal is a DECLARATION, and which lever you owe depends on blast radius _L9415
• §142 — An open stub whose h_exact class is MATCHED elsewhere is FREE. Propagate the body; do not gate a draft. (P30 S39, +7,710 ins in two commands) _L9778
• §143 — cast_call_sites read a RETURN STATEMENT as a prototype and deleted it. A 0/39 sweep became 18/39. (P30 S40) _L9834
• Then — propagation returned 0/137 TWICE — both times a missing TYPE _L10000
• §3-A. — The frame has THREE strata, and stratum 3 is unreachable from C _L10041
• §3-E. — A qty_compare TIE is not spelling-reachable — recognise it and stop _L10083
• Consequence — for the family (a real scheduling decision) _L10095
• §150 — A register ROTATION across symmetric blocks is VARIABLE-IDENTITY evidence, not an allocator tie (P30 S43, func_8017C6F4, 947 ins ×4) _L10312
• §151 — THE GHOST WEDGE: when a load-before-store transposition is unreachable by ANY statement order (P30 S43, func_8017EF68, 969 ins) _L10367
• When — to reach for it _L10407
• §152 — BYTE SIZE is the family key that name- and h_seq-grouping both miss (P30 S43, the 0xECC family: 1 crack → 12 overlays → 11,364 ins) _L10418
• §156 — an ORPHANED reconcile poisons the fleet: dedup_propagate's kept edit (S45 p6/p7) _L10640
• §160 — THE REACH-15 WAVE HARVEST: an align-1 block move, and the instrument that called it a wall (P30 S47) _L10893
• §162 — S48 WAVE-1 HARVEST (P30, 2026-08-11): the reach-ordered sibling campaign's first 12 targets _L11014
• §163z — THE UNVETTED REMAINDER (do not cite as law; each needs a dedupe pass) _L11828
• §3-The — ADDRESS-CLASS TABLE: which load/store pairs even REACH the /s clause (P30 S48 wave 4, func_80185B44, ov_SC03_014) _L14342
• §168 — THE COUSIN TIER (P30 S49, 2026-08-12): h_seq's exact-hash brittleness, measured — and the similarity map above it _L16211
• §171a — THE MECHANICAL A-PROP DRAFT (P30 S50): 256 members banked with no agent in the loop _L16420
• §171b — THREE CARRIES THE MECHANICAL DRAFT NEEDS (P30 S50, banking the top-reach families) _L16470
• §3-G. — TWO MODELLING TRAPS THAT COST THESE AGENTS SWEEPS OF HUNDREDS OF COMPILES _L17287
• §193-E — A varying-address (pointer) load is re-emitted once per CSE-LIVE INTERVAL, and naming it in a C local is the only C-level lever over that count — no store SPELLING has any reach _L18623
• §194-E — The wave card's exemplar is the TARGET ITSELF on 42/73 wave-U and 36/71 wave-T cards — a construction consequence of atlas.py:657 (exemplar = max-nins OPEN member) meeting build_wave_atlas.py:166 (one card per gid); and the shipped §193-A seed_ref is same-binary on 0/51, so no card field can ever name a destination-TU sibling _L19129
• §195-E — A 0/1 materialised at a JOIN immediately before the controlling beqz/bnez proves the source NAMED the condition — a truth expression in an if's controlling position reaches do_jump, which has no value path _L19920
• §199-B — A permutation sweep that holds ANY statement fixed is not a sweep: the statement an agent pins as "obviously load-bearing" is the one carrying the signal, and the partial sweep returns a FLAT residual that reads as proof of order-invariance _L20730

integration / TU plumbing (48)

• §8c — Splitting a TU means rebuilding its DECLARATION ENVIRONMENT, not moving text (Phase 26 session 6) _L437
• §8d — Templating a body INTO a TU must not CHANGE its declaration environment — demote the carried data externs (Phase 26 session 8, byte-proven on func_8015AE2C ×133) _L483
• §9.3 — Make it the build: resegment + swap + resolve (tools/psyq_integrate.py, libcd DONE) _L685
• §9.4 — Integrating a SECOND library (libgs block 6 after libcd) — multi-library gotchas _L712
• §9.5 — Integrating a WHOLE multi-block library in one call (full libgs — Phase 7 session G) _L740
• §17a — The TOOLKIT at WAVE scale (Phase-18 Step-3b/Step-1 — measured) + the pipeline-integration gotchas _L1445
• §3-THE — CAP: at this tail the match_one→gate gap is the LOOSE-TYPING CALL-GRAPH wall (not cheap plumbing) _L1637
• Diagnostic — lesson: a failed in-TU build leaves a STALE .o _L1654
• §28d — The "macro-extern-injection" lever: freeing reach-134 inline matches dedup_propagate skips as "not self-contained" (Phase 23, tools/inject_capped_externs.py) _L2366
• §30a — §30 generalizes to the FULL near-miss backlog (via STANDARD Opus agents, not just Fable5) + 2 more steer levers + the mechanical-integration throughput unlock (Phase 23 (a)) _L2394
• §40a — The DECL-RECONCILE pass (type-lift so remapped drafts compile in the sibling TU) (Phase 25 T7.2, 2026-07-08) _L2601
• §42a — addendum — wave 2 (residuals + 29-100 band): iso-MATCH ≠ real-TU bank, the memcpy→struct-assign fix, +5 levers (2026-07-10b) _L2953
• §42c — addendum — wave 3 (Max orchestrator + CORRECTED Ultracode fan-out, 2026-07-10c): the real-TU-faithful parallel harness (rtu_match) + 7/9 crack, ZERO iso-drift _L3033
• §56 — Banking a hand-drafted GIANT into its exemplar TU: self-contained draft vs live-TU decls, 4 reconciliation tactics (func_8013FAF8, 312 ins ×138, Phase 29 T4, 2026-07-17) _L4187
• §61 — Task 14: the gate ladder's missing stage is the ARITY pre-pass, and it is TU-side not draft-side (Phase 29, 2026-07-21) _L4545
• §61a — The Task-5 wave: 11/12 MATCH, 0 banked — three DISTINCT integration walls, each now named (Phase 29, 2026-07-21) _L4606
• §63 — The fresh-138 DEF-SIDE blocker: fix the HEADER decl, not the draft (fix_header_decl.py, Phase 29 SESSION-13, 2026-07-23) _L4877
• §63 — UPDATE (Phase 29 SESSION-14) — fix_header_decl's "SAFE" verdict is FLEET-BLIND; it MUST be R22-validated _L5000
• §65 — The stranded-draft recovery: BLAST-RADIUS TIERS, and the per-overlay de-macroize that refutes §20's DEF-conflict wall (tools/demacroize.py + tools/blocker_probe.py, Phase 29 SESSION-16, 2026-07-24) _L5015
• §65b — The escape: de-macroize the instantiation, don't touch the shared header _L5045
• §65f — The de-macroize lever's BOUNDARY, measured: byte-neutral 14 times of 15, and the 15th is why the gate exists _L5118
• §65g — Where the cheap levers STOP: the local-type and in-TU-self-decl classes did not yield _L5139
• §85 — The RETURN-axis fleet widen is ALL-OR-NOTHING: widening the shared header alone guarantees a conflict in the source overlay (Phase 29 SESSION-20, func_8012CC88 / func_8014D12C) _L6539
• §95 — reconcile_tu dropped the SIBLING declarators of a multi-symbol extern line (Phase 29 SESSION-21, func_80176218) _L6983
• §96 — The same rewrite, one shape down: reconcile_tu matched statements to lines by TEXT, so every COMMENTED declaration was silently skipped (Phase 29 SESSION-22, func_80176218 banked) _L7018
• §3-Two — reconcile_tu bugs found underneath, one introduced while fixing the other _L7190
• §100 — Prefer the DRAFT-LOCAL fix: a type only one function uses belongs in its BODY, not in a shared header (Phase 29 SESSION-22, func_80175DA8 0/137 → 137/137) _L7216
• §102 — A PLUMBING verdict can MASK a DIFF; and K&R is not always a codegen change (Phase 29 SESSION-22, func_8016EC0C) _L7272
• §103 — A FILE-scope extern in a shared overlay TU is a GLOBAL constraint on every LATER function; move the DECL, not the draft (Phase 29 T48/T51, func_80135260 — the fleet-wide half) _L7307
• §109 — Conforming a definition to a shared header: fix the NAMES, then check the RETURN precondition (Phase 29 T60) _L7600
• §112 — A macro-scoped declaration only collides where the macro is INSTANTIATED (Phase 29 T67/T69, audit_header_sigs.py) _L7730
• §114 — The THIRD decl axis: a CALLEE the draft declares differently from the target TU (Phase 29 T76/T77) _L7808
• §3-The — integration idioms (these decide whether a byte-correct draft BANKS) _L8804
• §3-The — declaration surface (integration, not codegen) _L8941
• Reconciling — a gate-refused draft: which way you edit depends on WHERE the TU's decl is _L9565
• §3-The — same swallow, twice more, in the integration spine _L9709
• §3-A. — Payload word0 is a global MODULE ID; code starts after the header _L10527
• §162 — THE LICM PAIR: what makes an address a movable AT ALL, and why the preheader order is the body order (P30 S47, ov_MAIN_012) _L11143
• §166 — THE DESTINATION-TU ORACLE (P30 S48): the seven-attempt bug that was never codegen _L14902
• §173 — THE STORED-PLUMBING RECOVERY RECIPE (P31 T6): symfix-first, per-group isolation, and where the verdicts have no drafts _L16606
• §176d — THE CONFLICT TABLE MUST BE SEEDED FROM THE TU, AND KEYED PER FILE (P31 S52, 2026-08-15) _L16803
• §180b — WHAT THE PRE-GATE LADDER ACTUALLY FINDS IN A COLD PILE (the shape of integration debt) _L17813
• §185 — EDIT THE SIDE THAT IS CHEAP TO VERIFY, AND CHECK A TU RETYPE AT ITS USE SITES _L18031
• §194-E — The wave card's exemplar is the TARGET ITSELF on 42/73 wave-U and 36/71 wave-T cards — a construction consequence of atlas.py:657 (exemplar = max-nins OPEN member) meeting build_wave_atlas.py:166 (one card per gid); and the shipped §193-A seed_ref is same-binary on 0/51, so no card field can ever name a destination-TU sibling _L19129
• §195-J — GTE / PsyQ op CALL-vs-INLINE is a PER-SITE SOURCE FACT, not a TU style — both forms coexist in one TU (measured in 2 TUs), and the tell is the target's own opcodes (lwc2/sqr/swc2 vs jal), not the sibling. Costs -8 ins on func_8018505C. Corollary: the game's inline sqr macro emits TWO hazard nops, the SDK's Square0 body emits ONE — so the inline form is provably not Square0 inlined (a second instance of §187's SDK-vs-game GTE nop divergence). _L20190
• §199-D — A narrow UNSIGNED value compared in an ordered if emits sltiu/sltu; the only dial is the WIDTH of a real object, and a widening (s32) cast is inert — AMENDS §35 (whose stated "only CSE-reuse canonicalizes" mechanism is byte-wrong) and does NOT apply inside a switch _L20823
• §200 — THE ALIAS IS THE UNIVERSAL DECLARATION ESCAPE: stop negotiating with the TU's spelling (P31 S55) _L21062
• §201-A — §150-B applies to decl_prior's DEF row: for an overlay-window symbol the banked "definition" is usually another overlay's function, and the card ranks it ABOVE the destination TU _L21121

build graph, splat & the harness (145)

• §4 — Flag/toolchain gotchas _L190
• Build — mechanism — per-file opt override (splat resegmentation) _L288
• §8c — Splitting a TU means rebuilding its DECLARATION ENVIRONMENT, not moving text (Phase 26 session 6) _L437
• §9 — Link real PsyQ library objects byte-exact (Phase 7 — GO proven) _L613
• §9.1 — Generalised per-object linker — tools/psyq_link.py (+ psyq_link_lib.py), 18/18 libcd byte-exact _L642
• §9.2 — Wire a library region into the build with NOLOAD — no data carving (tools/psyq_link_region.py) _L666
• §9.3 — Make it the build: resegment + swap + resolve (tools/psyq_integrate.py, libcd DONE) _L685
• §9.6 — Scaling library linking to the whole EXE (Phase 8 — 8 libs linked, 20%→50% byte-identical) _L766
• §9.7 — Binary-agnostic toolchain refactor (Phase 9) — the reusable pattern for Gen2 _L802
• §3-The — clean object-level metric (use this, not the permuter score, for jtbl/rodata functions) _L844
• §3-The — mechanism: game-code dedup is SOURCE-LEVEL, not an object swap (R-D1, the key lesson) _L926
• Per-binary — toolchain provenance (R24) _L959
• §12 — Ultracode harvest — parallel-draft + byte-gate at scale (Phase 12, resident: 1.4%→71.7% in one session) _L964
• One-command — onboarding — tools/new_overlay.sh <SCxx> <FILE_nnn> _L1062
• §3-THE — NON-4-ALIGNED-OVERLAY GOTCHA (≈75% of the fleet; fixed in the template + Makefile, automatic) _L1079
• Fleet — build — make build-all / make check-all _L1120
• §14 — Propagate a matched function across the fleet — tools/dedup_propagate.py (Phase 15) _L1126
• §14b — Harvesting the UNMATCHED shared core — the match_one wall (Phase 15) _L1192
• §14c — Callee-signature-aware harvest — breaking the extern-type-conflict wall (Phase 15, T6) _L1213
• §17 — The compiler-quirk wall — the matching TOOLKIT (Phase 18; gcc-2.7.2 source + byte-gated) _L1376
• §3-The — pipeline gotcha — match_one ≠ the gate; sig_unify is MANDATORY _L1421
• §17a — The TOOLKIT at WAVE scale (Phase-18 Step-3b/Step-1 — measured) + the pipeline-integration gotchas _L1445
• §19 — Scaling the toolkit waves — the recovery PIPELINE + the propagation CAP (Phase 19 T3) _L1597
• §20 — The wave-at-scale GATE CAP + residual-class verdicts (Phase 20) _L1632
• §3-THE — CAP: at this tail the match_one→gate gap is the LOOSE-TYPING CALL-GRAPH wall (not cheap plumbing) _L1637
• Diagnostic — lesson: a failed in-TU build leaves a STALE .o _L1654
• Phase-20 — RESOLUTION — tools/cast_call_sites.py BUILT + the cap re-diagnosed (R14, byte-proven) _L1701
• §24 — The ov_SC01_077_a.c split-file vein: split-aware propagation, but loose-typing-gated (Phase 21) _L2085
• §25 — The "schedule" class is mostly COALESCING (pin-crackable), not scheduling; + the gate two-stage + h_exact over-counts ×134 (Phase 21, cont.6) _L2132
• §3-The — gate two-stage — sig_unify is a FALLBACK, not unconditional (gate_stage.py, §19 folded in) _L2189
• §28 — Banking a "close=0 gate-rejected" giant: the canonical-extern recovery (Phase 22 T1, byte-proven on func_8015126C ×134) _L2325
• §28c — The close=0 recovery is NOT fully exhausted (§26 corrected, R14); + the dedup_propagate registry-skip recovery (Phase 22 T2) _L2352
• §28d — The "macro-extern-injection" lever: freeing reach-134 inline matches dedup_propagate skips as "not self-contained" (Phase 23, tools/inject_capped_externs.py) _L2366
• §29 — Reasoning-model (GLM5.2) DEF-side reconciliation idioms + the wall's hard limit (Phase 23 T10.7, tools/glm_reconcile.py) _L2374
• §33 — Automating the giant decl-reconcile: tools/reconcile_decls.py (the DATA analog of §20's cast_call_sites) + a fleet-majority type oracle (Phase 24 T7b, byte-proven on func_80129CF8) _L2434
• §34 — The func_80138ED0 giant crack: gcc-2.7.2's 3-qty sort bug + the zero-byte asm allocation toolkit + the giv-init fence (Phase 24 T5; Opus→close=21, Fable5→MATCH ×134) _L2454
• §35 — The region-a sibling-giant harvest: difficulty ≠ $s-reg count (it's global-array hoisting) + the banking recipe + new loop idioms (Phase 24 T7, 5 parallel Opus-Max agents) _L2471
• §37 — The T7 §G giant endgame: 6/8 cracked, meta-laws + transferable levers (Phase 24, 2026-07-07; FULL byte-verified detail + gcc-2.7.2 line cites in docs/gcc-2.7.2-map/t7g-giant-harvest.md) _L2506
• §39 — The ×1→×134 giant-endgame: propagate a matched -O2 giant via the NATIVE DEFINE-macro path (Phase 24 T7 §G close, 2026-07-08) _L2537
• §41 — The DEF-SIDE canonical-sig wall: mechanically banking a drafted giant past conflicting types (Phase 25 T5b batch-2, 2026-07-09; tools/canon_sig_reconcile.py, byte-proven on func_8013B274) _L2710
• §41a — v3.1: the def-side wall was ~71% TOOL-shaped — the five measured defects + the laws that dissolve them (Phase 25 T6, Fable5, 2026-07-09) _L2754
• §41b — T7 execution: the object-only probe OVER-counts BANKABLE by two link/rodata classes (Phase 25 T7-M1, 2026-07-10) _L2809
• §41d — void→s32 is NOT always byte-neutral: gate the RAW draft FIRST (Phase 26 session 6, byte-proven) _L2878
• §42b — addendum — wave 3 (Max, 2026-07-10c): THE STALE-OBJECT GATE TRAP + the read-global &-cast drift + fix _L2994
• §45 — The flagship func_80133CD4 crack (399 ins ×134): the merged-variable permutation-breaker + the 1-death local-alloc gate (Phase 25 task A giant escalation, Fable5 gdb-on-cc1, 2026-07-11) _L3298
• §51 — TOOLING INTEGRITY: the silent skip, and how to hunt it _L3630
• §51b — Why the byte-gate cannot save you _L3652
• §51f — Checklist for any new corpus-scanning tool _L3755
• §51g — When the thing you are scanning has a GRAMMAR, parse the grammar (tools/cdecl.py) _L3769
• §53 — SWEEP A FAMILY WITH THE TOOL ITS EXEMPLAR NEEDED: the jr/switch carve, and how omitting it manufactured the "families don't template" doctrine (Phase 28 T1, 2026-07-15) _L4028
• §55 — Core-crack wave levers + the GATE-ORCHESTRATION law (Phase 29 T3, 13-agent ultracode wave, 2026-07-17) _L4093
• §55b — THE GATE-ORCHESTRATION LAW (3 traps, ~3.5h lost; all recovered, 0 data lost) _L4122
• §55c — Sizing the propagate: a TARGETED propagate is ~4 min/core, and "it's slow" was a BROKEN-TREE ARTIFACT _L4141
• §54 — --fix-def-sig: the member's CANONICAL DECLARATION is a build step too (tiny-IMM mega-pools, +4,801, Phase 29 T6, 2026-07-16) _L4163
• §57 — The SELF-decl normalize: the sibling's OWN caller declares the templated fn divergently (the third §17a-1 direction; tools/normalize_self_decls.py, func_801670E4 ×137, Phase 29, 2026-07-18) _L4256
• §61 — Task 14: the gate ladder's missing stage is the ARITY pre-pass, and it is TU-side not draft-side (Phase 29, 2026-07-21) _L4545
• §61b — The jtbl gate stage: built, and the ORDERING law it exposed (Phase 29 Task-14 stage 4, 2026-07-21) _L4658
• §61d — The undo was eating the tree: two tools, one defect, invisible to the byte-gate (Phase 29, 2026-07-22) _L4779
• §62 — The jtbl RECONCILE must also follow the carve: the post-carve draft reconcile (harvest_verify._jtbl_reconcile, Phase 29 SESSION-11, 2026-07-22) _L4833
• §65 — The stranded-draft recovery: BLAST-RADIUS TIERS, and the per-overlay de-macroize that refutes §20's DEF-conflict wall (tools/demacroize.py + tools/blocker_probe.py, Phase 29 SESSION-16, 2026-07-24) _L5015
• §65d — Existing-ladder baseline, measured (do this before building a recovery stage) _L5091
• §65e — Two oracles, and the disagreement is the finding (R34 in practice) _L5101
• §65f — The de-macroize lever's BOUNDARY, measured: byte-neutral 14 times of 15, and the 15th is why the gate exists _L5118
• §66b — A metric parsed out of another tool's prose goes NULL silently when the label changes _L5200
• §67a — Run the symbol-set guard BEFORE you pay for a gate (tools/symcheck.py, Phase 29 SESSION-18) _L5405
• §77 — Every extraction tool carries a NARROW hard-coded preamble set; anything outside it silently caps the body's reach. Diff the preamble before you gate. (Phase 29 SESSION-19 — three variants in one session, two different tools) _L6084
• §3-The — CANDIDATE gate and the REAL gate need DIFFERENT preambles — keep the difference out of the bank _L6124
• §79 — For a 0-callee giant, fingerprint by DATA symbols (§71 cannot fire); and the STACK-SLOT ORDER is a declaration-order oracle (Phase 29 SESSION-19, func_8017BF14 4,763 ins, cold start → 45/4763) _L6220
• §82 — Two source-shape oracles from behemoth #6: a duplicated addiu $aN,$sp,K across a jal means the block was INLINED, and scalar-vs-aggregate decides WHEN a stack slot is allocated (Phase 29 SESSION-19, func_8017C730 1,061 ins) _L6370
• §3-2. — Scalar vs aggregate decides when the slot is allocated _L6389
• §3-Why — it survived every candidate gate _L6495
• §3-THE — FIX IS MECHANICAL — the tool already holds the answer _L6510
• §87 — match_one COMPILES but never LINKS, so an unresolvable data symbol reads as MATCH; and stored drafts go STALE against the tree (Phase 29 SESSION-20) _L6633
• §88f — the missing rung: a RELOCATION gate between match_one and the binary _L6718
• §89a — MEASURE the write set; do not assert its tier (tools/blast_radius.py) _L6732
• §89b — the parallel gate farm existed; the family path could not reach it (tools/sweep_parallel.py) _L6745
• §90 — Five tool-integrity laws from one session, each of which changed an answer (Phase 29 SESSION-21, 2026-07-27) _L6766
• §90a — A comparison tool MUST share its reference oracle's index space, exactly _L6772
• §92 — Conforming a DECLARATION: pointer changes are caller-neutral, scalar-WIDTH changes are not (Phase 29 SESSION-21, tools/conform_decls.py) _L6891
• §97 — The gate's own tree hygiene: a refused carve, an unchecked recovery, and a snapshot that captured a dirty tree (Phase 29 SESSION-22) _L7066
• §105 — A gate's revert must survive an EXCEPTION, not just a failure (Phase 29 T53, jtbl_family_bank) _L7431
• §107 — A lever wired into ONE gate path is a lever most families cannot reach (Phase 29 T56, func_80144090 0/136 → 136/136) _L7513
• ��3-Why — it survived so long: a MASKED oracle will MATCH a wrong symbol _L7953
• §122 — GATE RAW BEFORE TRANSFORMING; the undo belongs to the WRITER, as a per-edit journal (P30 T0a, 2026-07-30) _L8084
• §123 — PROPAGATE A FAMILY WITH THE TOOL ITS TIER NEEDS: dedup_propagate is h_exact-only; its refusals are statements about the TOOL (P30 wave 1, 2026-07-30) _L8116
• §126a — a bare except: continue around a coverage-asserting oracle re-creates the silent skip (P30 S28) _L8317
• §130 — An INCREMENTAL build can report BYTE-IDENTICAL for a change the CLEAN build cannot even LINK (P30 S28, the jr pair) _L8486
• §132 — The JR-PAIR-IN-ONE-O0-OBJECT "wall" was TWO instrument defects: a merged-double span the carve could not see, and a truncated object no rule deleted (P30 S29, func_8013B83C + func_8013BD74) _L8570
• §133 — The DEFAULT-FILTER class: three times in one session, a tool silently answered a narrower question than the one asked (P30 S1–S3) _L8685
• §134 — MULTI-LINE BLINDNESS: one root cause, four faces, in family_remap's preamble scanner (P30 S6, 190 zero-crack families) _L8711
• §3-The — splitting/merging rules (each closed a residual, byte-gated) _L8853
• §137a — A gate verdict has a TIMESTAMP; re-check it against the draft's mtime _L9386
• §138 — The propagation lanes: a gate refusal is a DECLARATION, and which lever you owe depends on blast radius _L9415
• §134 — again, in a second tool — and the waiter rule corrected _L9526
• Reconciling — a gate-refused draft: which way you edit depends on WHERE the TU's decl is _L9565
• §139 — A GATE THAT GREPS FOR VERDICTS MUST ASSERT 1:1 ACCOUNTING; and a --src filter must not survive a carve (P30 S38, wave 6: 10 of 16 drafts vanished) _L9594
• §140 — A METRIC IS NOT A MEASUREMENT UNTIL IT IS REPRODUCIBLE FROM THE COMMITTED TREE (P30 S1e: a phantom regression that gated the session's best lever) _L9660
• §142 — An open stub whose h_exact class is MATCHED elsewhere is FREE. Propagate the body; do not gate a draft. (P30 S39, +7,710 ins in two commands) _L9778
• §3-The — measurement (do this before any wave; it is ~20 lines and needs no builds) _L9790
• §155b — check the TYPE your oracle returns before comparing against it (S45 p5) _L10587
• §156 — an ORPHANED reconcile poisons the fleet: dedup_propagate's kept edit (S45 p6/p7) _L10640
• §158 — The RANGE-EXTENDER: a fourth zero-emission asm lever completes the allocno toolkit (P30 S46 tier-3, func_8017CE58, 733 ins) _L10712
• §160 — THE REACH-15 WAVE HARVEST: an align-1 block move, and the instrument that called it a wall (P30 S47) _L10893
• §161 — THE RETRY-WAVE HARVEST: a jump table indexed from zero, and two allocator traps (P30 S47) _L10969
• §162 — S48 WAVE-1 HARVEST (P30, 2026-08-11): the reach-ordered sibling campaign's first 12 targets _L11014
• §163 — S48 WAVES 2-3 HARVEST (P30, 2026-08-11/12): the five that were byte-probed and are actionable _L11760
• §165 — S48 WAVE-4 HARVEST (P30, 2026-08-12): banked the same day the wave landed _L13689
• §166 — THE DESTINATION-TU ORACLE (P30 S48): the seven-attempt bug that was never codegen _L14902
• §167 — S48 WAVE-5/6 HARVEST (P30, 2026-08-12): the saturation point _L14958
• §176e — SYMBOL IDENTITY IS COMPUTABLE OFFLINE (P31 S52): tools/reloc_identity.py _L16842
• §3-A. — The seven under-reporting holes (all in gate_main, all the same shape) _L16976
• §3-C2. — RECONCILE BEFORE THE FIRST GATE — a parked draft gets HARDER to bank, not easier _L17023
• §3-D. — The measured cost shape, and what to build next _L17046
• §176i — WHAT A STATIC PRE-GATE CHECK CAN AND CANNOT PROVE (P31 S52, wave Q) _L17064
• §179 — IDIOMS MINED FROM THE WAVE P/Q JOURNALS (P31 S52, harvest pass) _L17303
• §179-G — 🟡 A PIN CAN CREATE A COMBINE LOG_LINK AND DELETE AN andi (sixth RC-5 channel, n=1) _L17532
• §176c — MAIN (SLUS_007.26) CANNOT BE GATED INCREMENTALLY _L17599
• §180 — THE LEFTOVER-DRAFT HARVEST: RE-VERIFY WHAT YOU ALREADY HAVE BEFORE DRAFTING ANYTHING NEW _L17784
• §180b — WHAT THE PRE-GATE LADDER ACTUALLY FINDS IN A COLD PILE (the shape of integration debt) _L17813
• §181 — WHAT A WAVE'S GATE ACTUALLY REJECTS (P31 S53, measured on wave R's 45-draft main pile) _L17850
• §184 — COMMENT-BLINDNESS IS A DEFECT CLASS, NOT A BUG (P31 S53: three tools, one root cause, one session) _L18003
• §187 — 🔴 "SAME SOURCE" IS NOT "SAME OBJECT": THE SDK BUILD AND THE GAME BUILD DISAGREE ON GTE NOPS _L18109
• §190 — THREE PRESCRIPTIONS FROM THE SAME HARVEST (weaker evidence than §189, honestly labelled) _L18272
• §191 — WHAT THIS HARVEST DID NOT BANK (4 rejected, 4 narrowed) — recorded so it is not re-derived _L18309
• §192 — THE PRE-GATE LADDER WAS MAIN-ONLY, AND NOBODY COULD SEE IT (P31 S54) _L18331
• §193 — THE WAVE-T HARVEST (P31 S54): 71 index_gap reports -> 9 laws, 5 rejected, 61 already-covered _L18396
• §193-A — The wave card ships only OPEN-set pointers and discards the atlas's BANKED one — exemplar/sibs are stubs 100% by construction (tools/atlas.py:96/657), while seed.ref (matched pool, atlas.py:505-536) is dropped at build_wave_atlas.py:143 _L18413
• §193-G — §164-54's "scope to ≥4 arms" bound is byte-wrong — the dispatch-topology oracle goes live at THREE case nodes (balance_case_nodes splits at i > 2), but only for a signed-after-promotion index _L18717
• §193-I — A DECLARED AGGREGATE LOCAL HAS AN 8-BYTE FRAME STRIDE — CEIL(size,8), NOT size. N ARRAYS THEREFORE COST 8·k MORE THAN ONE HAND-LAID STRUCT (k = how many of them have size mod 8 ≠ 0), AND THE TELL IS IDENTICAL INSTRUCTION COUNT WITH EVERY $sp DISPLACEMENT SHIFTED BY THE SAME CONSTANT. _L18818
• §193-REJECTED — what this harvest did NOT bank (recorded so it is not re-derived) _L18877
• §194 — THE WAVE-U HARVEST (P31 S54): 64 index_gap reports -> 14 laws, 5 rejected, 44 already-covered _L18904
• §194-E — The wave card's exemplar is the TARGET ITSELF on 42/73 wave-U and 36/71 wave-T cards — a construction consequence of atlas.py:657 (exemplar = max-nins OPEN member) meeting build_wave_atlas.py:166 (one card per gid); and the shipped §193-A seed_ref is same-binary on 0/51, so no card field can ever name a destination-TU sibling _L19129
• §194-G — Reading the integer half of a 16.16 stack aggregate: REGISTER-LIVENESS, not the C spelling, decides sra vs lhu — and the break is count-neutral _L19239
• §194-K — Blind sched1's alias oracle with a second SET of a pointer pseudo — the first zero-byte, non-volatile, dependence-CREATING lever (corrects §167-05's "volatile is the only door"; fourth consumer of reg_n_sets) _L19415
• §194-REJECTED — what this harvest did NOT bank (recorded so it is not re-derived) _L19656
• §195 — THE WAVE-V HARVEST (P31 S54): 67 index_gap reports -> 14 laws, 9 rejected, 76 already-covered _L19677
• §195-I — §145(b) AMENDED — the pointer-bump addiu is saved by cse, not combine: ANY second SET of a pseudo in the address's equivalence chain (in-place p += K, §145(b)'s p = r; copy, or an asm re-tie) kills the fold; the pass is cse and the gate is invalidate's reg_tick++ _L20125
• §195-M — Frame vars is a SEQUENTIAL bump-allocation, not a flat sum: §193-I's CEIL(aggregate,8) term and §165-03/§167-06's 8×orphan term are the SAME frame_offset walk at two different compiler stages, and each stage re-CEILs frame_offset to 8 before it allocates _L20333
• §195-REJECTED — what this harvest did NOT bank (recorded so it is not re-derived) _L20429
• §197 — THE WAVE-W HARVEST (P31 S54): 68 index_gap reports -> 4 laws, 3 rejected, 41 already-covered _L20530
• §199 — THE WAVE-X HARVEST (P31 S54/S55): 63 index_gap reports -> 7 laws, 2 rejected, 56 already-covered _L20670
• §199-C — A NEGATIVE CONSTANT MULTIPLY ALWAYS TAKES expmed's negate_variant — but whether you ever SEE the neg is decided by COMBINE, and for an EVEN |K| it never disappears _L20775
• §199-D — A narrow UNSIGNED value compared in an ordered if emits sltiu/sltu; the only dial is the WIDTH of a real object, and a widening (s32) cast is inert — AMENDS §35 (whose stated "only CSE-reuse canonicalizes" mechanism is byte-wrong) and does NOT apply inside a switch _L20823
• §199-G — At TWO case nodes the switch-vs-if oracle is not blind — but the tell is ALL tests positive + a trailing j default, NOT the first test's polarity _L20996
• §201 — THE WAVE-Y HARVEST (P31 S55): 67 gap reports -> 5 laws, 8 rejected, 53 already-covered _L21111

process, measurement & doctrine (89)

• §8e — The jtbl ALIGNMENT LAW + the pad-spec filter — multi-table .rodata spans (Phase 29, byte-proven; .run/probe_jtbl/verdict.md) _L530
• §3-The — mechanism: game-code dedup is SOURCE-LEVEL, not an object swap (R-D1, the key lesson) _L926
• Per-binary — toolchain provenance (R24) _L959
• §17a — The TOOLKIT at WAVE scale (Phase-18 Step-3b/Step-1 — measured) + the pipeline-integration gotchas _L1445
• Diagnostic — lesson: a failed in-TU build leaves a STALE .o _L1654
• Phase-20 — RESOLUTION — tools/cast_call_sites.py BUILT + the cap re-diagnosed (R14, byte-proven) _L1701
• §23 — Giant func_80153E00 cracked (scalar-data CAST); BUT most giant drafts are STALE+INCOMPLETE, and a diag MUST remove artifacts (Phase 21 — byte-proven + a self-correction) _L2032
• §3-h — _exact OVER-COUNTS ×134 — verify shareability before crediting a class's "reach-134" (R14) _L2177
• §28c — The close=0 recovery is NOT fully exhausted (§26 corrected, R14); + the dedup_propagate registry-skip recovery (Phase 22 T2) _L2352
• §31-triage — R17 applies to CODEGEN residuals, never to a compile ERROR (Phase 26 session 7, Drew asked) _L2693
• §41a — v3.1: the def-side wall was ~71% TOOL-shaped — the five measured defects + the laws that dissolve them (Phase 25 T6, Fable5, 2026-07-09) _L2754
• §41b — T7 execution: the object-only probe OVER-counts BANKABLE by two link/rodata classes (Phase 25 T7-M1, 2026-07-10) _L2809
• §41b — addendum — M4 "reconcile_decls" tier is ALSO a probe over-count: 0/8 mechanical (Phase 25 T7-M4, 2026-07-10) _L2858
• §53 — SWEEP A FAMILY WITH THE TOOL ITS EXEMPLAR NEEDED: the jr/switch carve, and how omitting it manufactured the "families don't template" doctrine (Phase 28 T1, 2026-07-15) _L4028
• §3-The — meta-lesson (R35, and why this one is expensive) _L4078
• §55 — Core-crack wave levers + the GATE-ORCHESTRATION law (Phase 29 T3, 13-agent ultracode wave, 2026-07-17) _L4093
• §55b — THE GATE-ORCHESTRATION LAW (3 traps, ~3.5h lost; all recovered, 0 data lost) _L4122
• §57a — Two NSD corrections + the SURGICAL-ONLY law + the honest broad-sweep yield (Phase 29, 2026-07-18) _L4303
• §58 — match_one MATCH ≠ BANK: the four blind spots + the crack-wave reconcile-before-bank law (Phase 29 crack-wave, 2026-07-18) _L4343
• §60 — Classify the residual, don't rank it: the deterministic residual→class classifier and what it measured about the backlog (Phase 29 Task-13A, 2026-07-21) _L4407
• What — it measured — the whole open backlog, byte-grounded _L4431
• §3-The — parallel-probe race this surfaced _L4463
• §61b — The jtbl gate stage: built, and the ORDERING law it exposed (Phase 29 Task-14 stage 4, 2026-07-21) _L4658
• §62 — The jtbl RECONCILE must also follow the carve: the post-carve draft reconcile (harvest_verify._jtbl_reconcile, Phase 29 SESSION-11, 2026-07-22) _L4833
• §63 — UPDATE (Phase 29 SESSION-14) — fix_header_decl's "SAFE" verdict is FLEET-BLIND; it MUST be R22-validated _L5000
• §65 — The stranded-draft recovery: BLAST-RADIUS TIERS, and the per-overlay de-macroize that refutes §20's DEF-conflict wall (tools/demacroize.py + tools/blocker_probe.py, Phase 29 SESSION-16, 2026-07-24) _L5015
• §65a — The blast-radius taxonomy (makes §61's law structural instead of remembered) _L5030
• §65d — Existing-ladder baseline, measured (do this before building a recovery stage) _L5091
• §65e — Two oracles, and the disagreement is the finding (R34 in practice) _L5101
• §65f — The de-macroize lever's BOUNDARY, measured: byte-neutral 14 times of 15, and the 15th is why the gate exists _L5118
• §66c — Before a wave, verify the FUEL exists; an "already attempted" set built from the wrong directory lies in BOTH directions _L5217
• §66d-5 — residual_class's "structural ⇒ permuter CPU is waste" is WRONG for schedule permutations (measured, Phase 29 SESSION-18) _L5491
• §3-The — honest fix was source-level and cheap _L5750
• §3-The — residual, and the honest read _L6260
• §83 — The parameterised-repeat law, the spill-area trap, and why a per-case edit cannot move a per-case symptom (Phase 29 SESSION-20, func_80183814 5,122 ins, cold-ish start → 36 structural / 99.3%) _L6412
• Scope — , measured (do not over-generalise — §80) _L6518
• §3-Two — ladder lessons banked with it _L6528
• §3-Do — the WHOLE axis in one edit, then R22 once _L6567
• §3-THE — LAW: all-or-nothing PER FAMILY _L6600
• §88b — the slti literal-position law (extends §78 to comparisons) _L6686
• §88e — a wrong diagnosis, refuted properly (the model for how to treat an inherited lever) _L6708
• §89a — MEASURE the write set; do not assert its tier (tools/blast_radius.py) _L6732
• §90d — Do not measure a live wave's drafts (§87 in real time) _L6810
• §98 — conform_decls had three defects, and only the third needed R22 to find (Phase 29 SESSION-22, func_8014CF04) _L7113
• §106 — Persist the MEASUREMENT, derive the POLICY: a stored route let a stale file out-vote the live table (Phase 29 T54, residual_class._ROUTE) _L7470
• §125 — Split the CARVE from the BODY before calling a jr residue a wall — and measure it by SHA from a CLEAN tree (P30 SESSION-28; this section's first draft was WRONG and the method caught it) _L8206
• §3-The — meta-lesson _L8257
• Wave — economics (measured, for the next batch's sizing) _L8953
• §136b — A prior wave's "genuine byte-DIFF" verdict is NOT reliable evidence (4 of 4 refuted) _L9033
• §136g — When the index points at the WRONG lever: two byte-refuted routings (func_801863B4) _L9211
• §136h — CORRECTION: the zero-crack pool does NOT "refill with cheap work" (my error, byte-measured) _L9240
• §136j — The failure MIX flips with function size (measured across four bands, one session) _L9302
• Rank — the lane by measured concentration, not by class count _L9481
• §140 — A METRIC IS NOT A MEASUREMENT UNTIL IT IS REPRODUCIBLE FROM THE COMMITTED TREE (P30 S1e: a phantom regression that gated the session's best lever) _L9660
• §3-The — measurement (do this before any wave; it is ~20 lines and needs no builds) _L9790
• §3-And — the report-vs-bytes lesson attached to it _L9822
• §146 — RE-MEASURE A WALL BEFORE YOU RESPECT IT. Both "permanent" giants fell to drafts already on disk. (P30 S6, +50,094 ins) _L9977
• §147 — The three-stratum FRAME LAW, and four "stop searching" verdicts (P30 S42, func_8017C294, serial run) _L10036
• §3-C. — Inner-block declaration does NOT delay slot allocation — BYTE-REFUTED _L10068
• §3-D. — "Cheap fuel" that was never probed: 0 of 31 templatable _L10293
• §3-Two — corrections to the record _L10345
• §3-The — two fallouts, and how to close them (both measured, in order) _L10395
• What — does NOT work (14 byte-measured probes) _L10486
• §157 — the cheap-tier size cliff, measured (S45 p6) _L10687
• §162 — The cross-jump "CALL veto" is a COUNT law, not a CALL law (BOUNDS §88a; P30 S48, func_80189540) _L11302
• §3-The — law _L11549
• DIAGNOSTIC — TELL — two faces, one law _L11591
• §163 — S48 WAVES 2-3 HARVEST (P30, 2026-08-11/12): the five that were byte-probed and are actionable _L11760
• §163z — THE UNVETTED REMAINDER (do not cite as law; each needs a dedupe pass) _L11828
• §164z — REFUTED CLAIMS: do NOT re-derive these _L13623
• §165z — REFUTED THIS WAVE: do NOT re-derive _L14858
• §167z — REFUTED IN WAVES 5/6: do NOT re-derive _L16154
• §168 — THE COUSIN TIER (P30 S49, 2026-08-12): h_seq's exact-hash brittleness, measured — and the similarity map above it _L16211
• §176g — SIZE A WAVE BY INSTRUCTIONS, NOT BY CARDS (P31 S52 — the adopted doctrine) _L16930
• §176h — THE BATCH-SUBSTITUTION HAZARD MAP (P31 S52): seven holes, three wrong fixes, one law _L16968
• §3-D. — The measured cost shape, and what to build next _L17046
• §176j-2 — THE REPAIR PASS, MEASURED (do this instead of resuming) _L17122
• §176-C — 🔴 WALL REFUTATION: a hard-register pin CANNOT schedule around a call, because of a genuine gcc-2.7.2 bug _L17680
• §176-E — Two cheap source spellings, both cc1-probed _L17732
• §180 — THE LEFTOVER-DRAFT HARVEST: RE-VERIFY WHAT YOU ALREADY HAVE BEFORE DRAFTING ANYTHING NEW _L17784
• §181 — WHAT A WAVE'S GATE ACTUALLY REJECTS (P31 S53, measured on wave R's 45-draft main pile) _L17850
• §182 — §177's HONEST NEGATIVE: the epilogue lever cracked 4 of 16, and the 800c3 cluster held _L17906
• §183 — THE DECLARATION-RECONCILIATION PLAYBOOK (P31 S53, measured on 20 byte-verified drafts) _L17932
• §185 — EDIT THE SIDE THAT IS CHEAP TO VERIFY, AND CHECK A TU RETYPE AT ITS USE SITES _L18031
• §190 — THREE PRESCRIPTIONS FROM THE SAME HARVEST (weaker evidence than §189, honestly labelled) _L18272
• §194-A — A zero-byte scheduling fence goes AFTER the defining statement to make that computation emit FIRST in its block — and the barrier predicate is volatile-or-colon-less, not the "memory" clobber (COMPLEMENTS §165-40; does not refute it) _L18921
• §194-D — Declared width of a computed-value local is a sched1 dial (count-neutral) — a replication + attribution correction of §165-17, NOT a new argument-register law _L19088
• §195-J — GTE / PsyQ op CALL-vs-INLINE is a PER-SITE SOURCE FACT, not a TU style — both forms coexist in one TU (measured in 2 TUs), and the tell is the target's own opcodes (lwc2/sqr/swc2 vs jal), not the sibling. Costs -8 ins on func_8018505C. Corollary: the game's inline sqr macro emits TWO hazard nops, the SDK's Square0 body emits ONE — so the inline form is provably not Square0 inlined (a second instance of §187's SDK-vs-game GTE nop divergence). _L20190
• §201-E — §194-J-2 — The last_mem_set deletion window is measured in SOURCE/expand order, not in the printed stream; a foreign store moved between the pair in C source is a real lever, and volatile is not always the better one _L21331

(unbucketed — title matched no symptom vocabulary) (198)

• §3-How — to use this _L30
• §1 — Idiom catalog (asm pattern → C that produces it) _L39
• §3-I3 — Division by a constant → magic multiply _L54
• §3-I4 — Runtime (variable) division → divu + zero-check break (NEEDS --expand-div) _L60
• §2 — Writing matching C (what makes gcc emit X) _L69
• §3a — Escalation TIER above the permuter — web-research the compiler internals (HIGH VALUE, proven) _L128
• §3b — §31-directed permuter mutation — bias the search over the class's levers (Phase 24 T5) _L147
• Residual — B — a return <const> materialised late / merged instead of distributed per-site _L867
• §3-The — A→B→C per-overlay workflow _L1091
• Dedup-credit — (§11) for overlays — two shapes _L1108
• §14a — The fleet bulk run (Phase 15 — --auto-from, 947 REAL → 74,527; 3.82% → 22.14% in one pass) _L1169
• §14d — Deterministic recovery beats agent waves on the hard tail (Phase 15, the final-session finding) _L1269
• §14e — Two hard-tail dead-ends (Phase 15 — documented so they aren't re-attempted) _L1298
• §3-The — STEERABLE idioms (byte-confirmed this phase) _L1404
• Strategic — conclusion — the match-% lever post-research _L1435
• §3-NEW — / CONFIRMED residual classes (this session) _L1660
• Operational — gotchas (cost real time) _L1681
• What — WORKED — the Phase-20 reusable wins _L1689
• §21 — wave-distilled idioms (Phase 21) _L1752
• §26 — The cheap close=0 recovery lever is EXHAUSTED; idiom_loop --assess was DOUBLY inflated (Phase 21, cont.7) _L2199
• §3-The — _a close=0 recovery banks 0/20 — same def-side wall as MAIN (0/40) _L2205
• §27 — Giant matching recipe (Phase 21 cont.7b — validated on func_80176D94, 152 ins) _L2259
• §28a — decomp.wiki GCC patterns worth trying on BFM giants (decomp.wiki/compilers/GCC, raw at decompals/decompedia; PS1-applicable subset) _L2345
• §31 — THE gcc-2.7.2 CODEGEN MAP: pass → residual → C-lever catalog (Phase 23; 4 Fable5 agents read the compiler source) _L2402
• §36 — Fable5 giant-crack levers (Phase 24 T7 Fable5 batch; accumulates as each lands — full RTL dumps in .run/t7/fable/) _L2487
• §42d — addendum — wave 4 (rtu_match fan-out over the mapped frontier, 2026-07-10c): 24/26 MATCH, +5 durable levers _L3082
• §3-B. — THE EBB RULE — the general form of §46-L2 _L3468
• §3-E. — Meta _L3518
• §50 — Refinements that BOUND §47/§48 (from the func_80135EB0 wall, 21→6; Phase 26 session 8) _L3576
• §51a — The bug class _L3636
• §51c — THE METHOD (do not audit by reading the regex) _L3662
• §51d — THE LAWS _L3677
• §51e — The false-wall pipeline (why this is not just hygiene) _L3739
• §3-Why — the wall is (probably) intrinsic _L3945
• §3-The — case _L4033
• §3-The — rule _L4062
• §55a — New byte-proven levers (each from a banked or near draft) _L4098
• §60a — What the first DIRECTED grinder run exposed (Phase 29 Task-13B, 2026-07-21) _L4472
• §60b — The plateau autopsy's verdict: a partial drift is a WRONG DRAFT, not a missing transform (Phase 29 Task-13B close, 2026-07-21) _L4510
• §66 — Exercise a banking driver's SUCCESS path before pointing it at a wave: the free re-bank test (Phase 29 SESSION-17, 2026-07-24) _L5159
• §66d-2 — Two operational sharp edges _L5289
• §66d-3 — Read the ILS per-cycle SERIES, not its final best: a repeated score and a still-falling one look identical in a summary line and mean opposite things _L5300
• §67 — The arg-copy PLACEMENT lever: launder a parameter into a fresh pseudo AT the statement where the target's copy lands (Phase 29 SESSION-18, func_8014D820 25 → 16) _L5315
• §66d-4 — "ILS converged" means converged FOR THAT WEIGHT PROFILE, not a floor (amends §66d-3; Phase 29 SESSION-18) _L5438
• §69 — How to attack a behemoth: map it, don't draft it (Phase 29 SESSION-18, func_80183814, 5,122 ins) _L5569
• What — is left, and what is byte-recorded as SPENT _L5766
• §3-The — mechanism, with citations _L6037
• §3-Two — diagnosis traps this function proved _L6064
• Practice — Practice _L6074
• §3-The — drift was an allocation decision, not missing code _L6180
• §3-The — economics _L6211
• §71 — has a blind spot, and this is it _L6226
• §3-The — flagged "#1 move" LOST — and why the failure is informative _L6302
• §78 — 's attribution primitive, run and reproduced _L6312
• Cold-start — economics, now complete _L6318
• Also — reproduced on this function _L6401
• §3-And — the banking footnote (§75a class A, one line) _L6405
• §83a — READ THE HEADLINE NUMBER CORRECTLY: a LENGTH drift makes match_one's count meaningless _L6419
• §83b — THE LEVER: find the parameterised REPEAT before decoding case-by-case _L6427
• §83e — §80 vindicated again, on the same day it was written _L6466
• §3-The — conflict _L6544
• §3-THE — FAILURE MODE — widening only engine_core.h is worse than not starting _L6552
• §3-The — precondition, and how to check it in one grep _L6559
• Reading — , for the next person _L6576
• §3-The — guard refuses a class that largely works _L6591
• §3-The — blindness ladder, now complete — FOUR classes match_one cannot see _L6649
• Consequence — for the backlog ledger _L6659
• §88 — cross_jump will not merge a common suffix containing a CALL; and §78 is scoped to ORDERED comparisons only (Phase 29 SESSION-20, the behemoth close-out) _L6674
• §88a — repeated CALL-shaped blocks are left UNMERGED; call-free tails are merged for you _L6679
• §89 — Two throughput rules the project already had written down and was not following (Phase 29 SESSION-20) _L6726
• §3-The — standing sequence _L6758
• §90b — "Byte-neutral" is not "wanted": undo on the SUCCESS path too _L6789
• §90c — A library-callable function must FAIL CLOSED on an unconfigured module _L6799
• §90e — An agent's CONCLUSION and its EVIDENCE fail independently — re-derive the premise, not the fix _L6818
• §93 — set -o pipefail attributes a pipeline failure to the LAST stage, not the failing one (Phase 29 SESSION-21, func_8014D820) _L6922
• §101 — The STALE DEFAULT class: a guard whose cause was removed is a silent skip wearing a safety label (Phase 29 SESSION-22, three instances in one session) _L7245
• §110 — A unit must define exactly ONE function, and "ends in ;" does not tell you which line defines it (Phase 29 T65) _L7640
• §115 — A func_XXXXXXXX predicate rots by design: the same name-form assumption in THREE places (Phase 29 T78) _L7850
• §3-The — fix moves the DEFINITION, not the stub — and here is why the obvious shortcut fails _L7895
• §117 — Spell the sibling's symbol from the SIBLING's address, not the exemplar's kind (Phase 29 T82) _L7926
• §118 — Ordinal (positional) immediate resolution: compare C tokens to the DIFFERING asm uses (Phase 29 T87) _L7963
• §119 — Two levers on the SAME axis, opposite directions: test the off-diagonal (Phase 29 T89) _L7997
• §3-The — wiring trap that cost two attempts _L8043
• §3-The — method (keep this) _L8213
• §3-The — instrument rules that make its answer trustworthy (this is where I failed) _L8223
• §3-The — corrected results (each SHA-verified, from a clean tree, restore re-verified) _L8238
• §3-The — instrument trap that hid it (and it is §124's shape again) _L8290
• §3-The — mechanics _L8299
• §3-The — idiom they kept re-deriving: the constant-offset fold _L8356
• §127b — the knowledge was in a SOURCE COMMENT, not the cookbook _L8386
• §128 — A raw NUL in C source makes grep SILENTLY SKIP the file (P30 S28, 137 files) _L8393
• §128a — a negative control must corrupt a SCRATCH COPY, never the tracked file _L8425
• §3-The — real blocker underneath, for the record _L8477
• §3-The — diagnostic ladder that finally located it (reusable) _L8523
• Defect — 1 — a pre-§8e MERGED DOUBLE is not a single-table predecessor _L8579
• Defect — 2 — as writes a corpse and nothing deletes it _L8597
• §3-The — fingerprint, and the 30-second ladder that found it _L8607
• §3-The — transferable rule _L8628
• §132a — --like is for a sibling with NO record; against one that HAS a record it over-derives (P30 S29, ov_SC07_010) _L8635
• §135 — Six byte-verified gcc-2.7.2 idioms from the P30 S6f-h waves (and the two-lane wave shape that found them) _L8765
• §3-The — codegen idioms _L8770
• §3-The — wave shape that produced these _L8819
• §136 — The LOCAL-VARIABLE lever: how many C locals, at what scope (P30 wave 4a, 25 byte-verified banks) _L8837
• §136a — Blocker capture: classify on the OUTPUT, never on the exit status _L8976
• §136d — Four gcc-2.7.2 levers the redraft lane found (each closed a residual no other lever moved) _L9094
• §136e — §136c's PRECONDITION, and two more symptom keys (wave 4b batch 3) _L9138
• §136i — The drafter model LADDER: Haiku → Sonnet → Opus → Fable5 (Drew, 2026-08-03) _L9270
• §3-The — triage, cheapest first _L9421
• THREE — carry variants hide in one "CARRY-FIXABLE" bucket — and they need different fixes _L9490
• STEP — 0 of sibling-first: grep src/ for a distinctive LITERAL from the .s _L9546
• §3-The — generalisation — three corollaries worth more than the bug _L9626
• §3-And — the inverse-lookup trap, same session _L9643
• §3-The — three-line proof (do this before diagnosing any metric movement) _L9680
• §3-The — two instrument defects it exposed _L9689
• §3-Two — wrong mechanisms I chased first, and why they were wrong _L9720
• §3-The — trap that hid it — SAME FUNCTION, TWO ROUTES, ONLY ONE IS FREE _L9805
• Route — selection (why --addr sometimes says "nothing changed") _L9814
• §144 — THE LITERAL'S SPELLING PICKS THE IMMEDIATE ENCODING (P30 S40 wave 1, func_801822E0) _L9888
• §3-Why — a correct draft can read as an intrinsic wall _L9988
• §3-The — rule _L10023
• §3-B. — (v < 0x40) ? v : 0x3F is folded to MIN_EXPR and expands to the WRONG SHAPE _L10170
• §3-D. — Reproduce the original's BUGS verbatim _L10194
• §149 — Four instrument defects in one session, and the two questions they were hiding (P30 S43) _L10243
• §3-A. — A prep step that returns its input on failure is indistinguishable from a search that found nothing _L10249
• §3-B. — Same address + same name ≠ same body — and the ledger keys on address _L10267
• §3-C. — make: *** [...] Error N is a summary, never a diagnosis _L10283
• §3-The — fix _L10319
• §3-The — method that found it (this is the transferable part) _L10329
• Diagnostic — order (adopt this) _L10356
• §3-The — mechanism (read from cc1's own -dR trace, not inferred) _L10372
• §3-The — lever — a zero-emission insn that absorbs the blocked tick _L10387
• §3-The — finding _L10423
• §3-The — key _L10432
• §3-Two — cautions that must travel with this technique _L10443
• §3-The — companion defect (open) _L10454
• Symptom — Symptom _L10474
• Mechanism — (gcc source + RTL dumps, not inferred) _L10479
• §3-The — cure — a fresh launder per site, each in its own block _L10492
• Companion — levers from the same function _L10502
• §3-B. — Two static base-derivation methods that must AGREE (use both) _L10536
• §155a — the same failure class, one level up: SHAPE-blind table scanning (S45 p5) _L10567
• §155c — the ZERO-REFERENCE trap: gcc splits a global-array address across the lui and the LOAD (S46) _L10610
• Symptom — Symptom _L10721
• §3-Why — the fork is chained (gcc source, validated insn-by-insn against -dS/-dR dumps) _L10727
• §3-The — method (dump-arithmetic first, then place — no probing) _L10740
• §162a — SHARPENS (sharpens §161a, §131, §8a-pad, §129a) _L11041
• §162b — SHARPENS (sharpens §48-A3, §156, §150, §76) _L11070
• §162d — SHARPENS (sharpens §31, §21, §30, §55a) _L11099
• §162e — NEW _L11141
• §162f — SHARPENS (sharpens §42d, §41d, §73, §10) _L11209
• §162g — NEW _L11264
• §162h — SHARPENS (sharpens §88, §88a, §50-B, §8) _L11300
• §162i — SHARPENS (sharpens §135, §21, §42, §32) _L11368
• §162j — SHARPENS (sharpens §25, §136d-1, §48-B, §46-L2) _L11393
• §162k — SHARPENS (sharpens §1-I2, §12, §160d, §21) _L11422
• §162l — SHARPENS (sharpens §48-B, §48-C1, §20, §21) _L11485
• §162m — SHARPENS (sharpens §36, §158, §148, §153) _L11540
• §158a — THE FIFTH LEVER IS NOT AN ASM: do { } while (0) is a REGION ref-multiplier you MINT (P30 S48, func_8017CBC8, ov_MAIN_012 / jr_801789AC, 188 ins → MATCH) _L11542
• Size — it before you write it (§158 step 1-2, applied) _L11560
• §3-The — wrap BOUNDARY is the dial — and it is indiscriminate _L11581
• §3-Not — a pure dial _L11587
• §162n — NEW _L11613
• §162o — SHARPENS (sharpens §158, §136-1, §136-6, §79) _L11646
• §162p — SHARPENS (sharpens §48-B, §46-L2, §156, §136d-1) _L11705
• §162q — SHARPENS (sharpens §30, §30a, §135-2, §136-13) _L11722
• §164 — S48 §163z SKEPTIC PASS (P30, 2026-08-12): 190 claims vetted, 82 banked _L11848
• §169 — THE MICRO-ADAPT LANE (P30 S49): edit a proven body, don't crack a new one _L16263
• §170 — THE A-PROP WORD-DIFF CARD (P30 S49): the lane that had no owner _L16310
• §171 — THE STALE SEED SYMBOL (P30 S50, 2026-08-13): why §170's 91%→57% was never codegen _L16354
• §172 — THE ORPHAN-SLOT MECHANISM v2 (P30 S50-Max): the complete frame-residue model for gcc-2.7.2 MIPS _L16503
• §172a — TWO DECOMPILATION TELLS FROM THE SAME DIG (P30 S50-Max) _L16557
• §172b — THREE MORE TELLS FROM THE GCC READ (P30 S50-Max, banked on Drew's ask) _L16573
• §174 — THE ADAPT-CARD WAVE RECIPE (P31 waves A/B, 2026-08-14): prevention beats recovery _L16642
• §176a — THE VERIFICATION-LAYER LAWS (P31 overnight, 2026-08-15). What each check can and cannot prove. _L16737
• §176b — BATCH-GATING MECHANICS (P31): what changes when N drafts land in ONE .c _L16777
• §3-C. — The limit that remains (recorded, not solved) _L17016
• §176k — TWO SELECTOR BUGS THAT SILENTLY SHRINK A WAVE _L17142
• §178 — SIX LEVERS MINED FROM THE WAVE-P JOURNALS (P31 S52), each byte-proven and source-cited _L17216
• §3-A. — THE $0-ADD OPAQUE COPY defeats make_regs_eqv (func_80033398, 93 ins) _L17228
• §3-C. — SINGLE-SET TEMPS GET THE BIRTHING BOOST (func_8001D3FC, 196 ins) _L17257
• §3-E. — THE ZERO-OFFSET ALIAS HOLE (func_80037028, 71 ins) _L17272
• §179-B — 🔴 THE HAND-WRITTEN-ASM TRANSCRIPTION CHECKLIST (maspsx + masked_diff; 10 byte-proofs) _L17362
• §179-C — 🔴 A FUNCTION WITH NO EPILOGUE (falls into a sibling's shared tail) MUST BE FILE-SCOPE __asm__ _L17412
• §179-D — gte_stflg MUST CLOBBER "$12" OR THE WHOLE TEMP FILE ROTATES BY ONE _L17463
• §179-H — A MID-BODY .global LABEL PAIR SLICES A BYTE-COMPARABLE FRAGMENT OUT OF A LARGER ROUTINE _L17559
• Considered — and NOT banked _L17582
• §176-F — Misdiagnosis triage: four residual verdicts that were lying _L17754
• What — is NOT banked here _L17771
• §180c — WHEN A BINARY'S MASS BAND IS SPENT, THE FLEET-WIDE DRAW IS STRICTLY BETTER _L17832
• Only — ONE of 27 blocked drafts was wrong. The other 26 were correct and unbankable. _L17851
• §180d — THE pgrep BRACKET TRICK PROTECTS THE PATTERN, NOT THE COMMAND LINE _L17920
• Three — defects in one call path; the overlay slates that carry most of the wave work were being waved through _L18332
• §194-F — if ((*p = v = f()) == 0) is an expand-time pseudo SPLITTER (store_expr's want_value && MEM path), not a fold — it partitions one value between local_alloc and global_alloc. BOUNDS §21's L1872 bullet, whose stated direction is byte-wrong on 3 of 4 instances, and CLOSES the control §167-37 asked for. _L19166
• §194-L — §88b and §189-E are BOTH half-wrong, but not the way the candidate says: the compare-constant shape is a 2-D lookup (cmp_info ROW × constant-in-window), and naming matters on OPPOSITE sides of the window for the GE/LT rows vs the GT/LE rows _L19493
• §197-REJECTED — §197-REJECTED _L20655
• §199-REJECTED — §199-REJECTED _L21051
• §201-B — In a narrowed PLUS/MINUS/AND/IOR/XOR expression the destination pointee is INERT — the sign of the materialized constant is decided by an OR over the UNWIDENED operands (convert.c trunc1), which bounds §1841 to direct constant stores _L21153
• §201-REJECTED — eight, the session's highest _L21387

All sections, in order

• §3-How — to use this _L30
• §1 — Idiom catalog (asm pattern → C that produces it) _L39
• §3-I1 — Unsigned range check: (x - lo) < (hi-lo) → addiu+sltiu _L41
• §3-I2 — Byte mask forces andi even after lbu _L47
• §3-I3 — Division by a constant → magic multiply _L54
• §3-I4 — Runtime (variable) division → divu + zero-check break (NEEDS --expand-div) _L60
• §2 — Writing matching C (what makes gcc emit X) _L69
• §3-T1 — Loop pointer: top-of-body for addu induction, not constant-folded addiu _L71
• §3-T2 — Source statement order drives instruction scheduling _L78
• §3-T3 — Types _L85
• §3-T4 — Branch polarity: invert the source condition to flip gcc's chosen branch _L90
• §3 — When a diff is pure scheduling → decomp-permuter (harness built, Phase 6) _L107
• §3a — Escalation TIER above the permuter — web-research the compiler internals (HIGH VALUE, proven) _L128
• §3b — §31-directed permuter mutation — bias the search over the class's levers (Phase 24 T5) _L147
• §4 — Flag/toolchain gotchas _L190
• §5 — Known hard-residual classes (instruction-identical, one byte-exact blocker) _L199
• §5a — Cross-jump tail-merge — gcc collapses two byte-identical blocks the original kept separate (FIX FOUND) _L211
• §6 — Per-module optimization mixing — the -O0 boot module (Phase 7) _L246
• Detecting — the opt level (do this first) _L254
• Build — mechanism — per-file opt override (splat resegmentation) _L288
• §7 — PsyQ SDK types & symbols (the library-call prerequisite) _L303
• §8 — rodata island (compiler jump tables) — the .data→.rodata→.data sandwich (Phase 7) _L320
• §8a — rodata island in a flat OVERLAY — the tail sandwich, per matched jr-function (Phase 26 — PoC PROVEN) _L342
• §8a-pad — a trailing .word 0x00000000 under a jtbl dlabel is .align PAD, not an entry (Phase 26 session 6, byte-proven) _L380
• §8b — MULTI-jtbl per overlay — the ld_interleave --order sandwich + the same-subseg cases (Phase 26 session 4) _L404
• §8c — Splitting a TU means rebuilding its DECLARATION ENVIRONMENT, not moving text (Phase 26 session 6) _L437
• §8d — Templating a body INTO a TU must not CHANGE its declaration environment — demote the carried data externs (Phase 26 session 8, byte-proven on func_8015AE2C ×133) _L483
• §8e — The jtbl ALIGNMENT LAW + the pad-spec filter — multi-table .rodata spans (Phase 29, byte-proven; .run/probe_jtbl/verdict.md) _L530
• §9 — Link real PsyQ library objects byte-exact (Phase 7 — GO proven) _L613
• §9.1 — Generalised per-object linker — tools/psyq_link.py (+ psyq_link_lib.py), 18/18 libcd byte-exact _L642
• §9.2 — Wire a library region into the build with NOLOAD — no data carving (tools/psyq_link_region.py) _L666
• §9.3 — Make it the build: resegment + swap + resolve (tools/psyq_integrate.py, libcd DONE) _L685
• §9.4 — Integrating a SECOND library (libgs block 6 after libcd) — multi-library gotchas _L712
• §9.5 — Integrating a WHOLE multi-block library in one call (full libgs — Phase 7 session G) _L740
• §9.6 — Scaling library linking to the whole EXE (Phase 8 — 8 libs linked, 20%→50% byte-identical) _L766
• §9.7 — Binary-agnostic toolchain refactor (Phase 9) — the reusable pattern for Gen2 _L802
• §10 — Closing the regalloc/scheduling hard tail by hand (LZSS, Phase 7 session F — the full close) _L835
• §3-The — clean object-level metric (use this, not the permuter score, for jtbl/rodata functions) _L844
• Residual — A — commutative |/&/+ result lands in the wrong source-operand register _L856
• Residual — B — a return <const> materialised late / merged instead of distributed per-site _L867
• §11 — Cross-binary dedup & code-sharing (Phase 11 — "one match unlocks many") _L908
• §3-The — mechanism: game-code dedup is SOURCE-LEVEL, not an object swap (R-D1, the key lesson) _L926
• Per-binary — toolchain provenance (R24) _L959
• §12 — Ultracode harvest — parallel-draft + byte-gate at scale (Phase 12, resident: 1.4%→71.7% in one session) _L964
• §13 — Add a location overlay — the canonical runbook (Phase 13; the Phase-15 fleet recipe) _L1056
• One-command — onboarding — tools/new_overlay.sh <SCxx> <FILE_nnn> _L1062
• §3-The — flat-blob overlay config (what the template encodes) _L1071
• §3-THE — NON-4-ALIGNED-OVERLAY GOTCHA (≈75% of the fleet; fixed in the template + Makefile, automatic) _L1079
• §3-The — A→B→C per-overlay workflow _L1091
• Dedup-credit — (§11) for overlays — two shapes _L1108
• Fleet — build — make build-all / make check-all _L1120
• §14 — Propagate a matched function across the fleet — tools/dedup_propagate.py (Phase 15) _L1126
• §14a — The fleet bulk run (Phase 15 — --auto-from, 947 REAL → 74,527; 3.82% → 22.14% in one pass) _L1169
• §14b — Harvesting the UNMATCHED shared core — the match_one wall (Phase 15) _L1192
• §14c — Callee-signature-aware harvest — breaking the extern-type-conflict wall (Phase 15, T6) _L1213
• §14d — Deterministic recovery beats agent waves on the hard tail (Phase 15, the final-session finding) _L1269
• §14e — Two hard-tail dead-ends (Phase 15 — documented so they aren't re-attempted) _L1298
• §15 — Struct-heavy shared-core pipeline (Phase 16) — empirical determinations (S0) _L1313
• §16 — Guided hand-matching the struct-heavy core (Phase 17 — beats the §15 brute-force) _L1335
• §17 — The compiler-quirk wall — the matching TOOLKIT (Phase 18; gcc-2.7.2 source + byte-gated) _L1376
• Register-allocation — ORDER (call-crossing $s0/$s1 swap) → FORCE it with register pins (byte-proven) _L1384
• §3-The — STEERABLE idioms (byte-confirmed this phase) _L1404
• §3-The — pipeline gotcha — match_one ≠ the gate; sig_unify is MANDATORY _L1421
• §3-The — LOOSE-TYPING wall is real for narrow params (Phase 16, reconfirmed) _L1428
• Strategic — conclusion — the match-% lever post-research _L1435
• §17a — The TOOLKIT at WAVE scale (Phase-18 Step-3b/Step-1 — measured) + the pipeline-integration gotchas _L1445
• §18 — Per-file -O0 split inside an overlay/blob (Phase 19 T1) _L1519
• §19 — Scaling the toolkit waves — the recovery PIPELINE + the propagation CAP (Phase 19 T3) _L1597
• §20 — The wave-at-scale GATE CAP + residual-class verdicts (Phase 20) _L1632
• §3-THE — CAP: at this tail the match_one→gate gap is the LOOSE-TYPING CALL-GRAPH wall (not cheap plumbing) _L1637
• Diagnostic — lesson: a failed in-TU build leaves a STALE .o _L1654
• §3-NEW — / CONFIRMED residual classes (this session) _L1660
• Operational — gotchas (cost real time) _L1681
• What — WORKED — the Phase-20 reusable wins _L1689
• Phase-20 — RESOLUTION — tools/cast_call_sites.py BUILT + the cap re-diagnosed (R14, byte-proven) _L1701
• §21 — wave-distilled idioms (Phase 21) _L1752
• §22 — DEF-side loose-typing recovery + grinder blacklist (Phase 21) _L2005
• §23 — Giant func_80153E00 cracked (scalar-data CAST); BUT most giant drafts are STALE+INCOMPLETE, and a diag MUST remove artifacts (Phase 21 — byte-proven + a self-correction) _L2032
• §24 — The ov_SC01_077_a.c split-file vein: split-aware propagation, but loose-typing-gated (Phase 21) _L2085
• §25 — The "schedule" class is mostly COALESCING (pin-crackable), not scheduling; + the gate two-stage + h_exact over-counts ×134 (Phase 21, cont.6) _L2132
• §3-The — crack: the residual was gcc-2.7.2 COPY-COALESCING, not the instruction scheduler _L2144
• §3-The — genuine scheduler — rank_for_schedule (sched.c), for when it IS scheduling _L2160
• §3-The — genuine schedule WALLS (do NOT re-grind — stub) _L2169
• §3-h — _exact OVER-COUNTS ×134 — verify shareability before crediting a class's "reach-134" (R14) _L2177
• §3-The — gate two-stage — sig_unify is a FALLBACK, not unconditional (gate_stage.py, §19 folded in) _L2189
• §26 — The cheap close=0 recovery lever is EXHAUSTED; idiom_loop --assess was DOUBLY inflated (Phase 21, cont.7) _L2199
• §3-The — _a close=0 recovery banks 0/20 — same def-side wall as MAIN (0/40) _L2205
• Where — this leaves the reach-134 tail (cont.6 option-3, now CONFIRMED byte-backed) _L2233
• §27 — Giant matching recipe (Phase 21 cont.7b — validated on func_80176D94, 152 ins) _L2259
• §28 — Banking a "close=0 gate-rejected" giant: the canonical-extern recovery (Phase 22 T1, byte-proven on func_8015126C ×134) _L2325
• §28a — decomp.wiki GCC patterns worth trying on BFM giants (decomp.wiki/compilers/GCC, raw at decompals/decompedia; PS1-applicable subset) _L2345
• §28c — The close=0 recovery is NOT fully exhausted (§26 corrected, R14); + the dedup_propagate registry-skip recovery (Phase 22 T2) _L2352
• §28b — The struct-walled close=0 giant (§28 case #3) IS bankable: the engine_types.h type-lift (Phase 22 T2, byte-proven on func_80156B74 ×134) _L2357
• §28d — The "macro-extern-injection" lever: freeing reach-134 inline matches dedup_propagate skips as "not self-contained" (Phase 23, tools/inject_capped_externs.py) _L2366
• §29 — Reasoning-model (GLM5.2) DEF-side reconciliation idioms + the wall's hard limit (Phase 23 T10.7, tools/glm_reconcile.py) _L2374
• §30 — Fable5Max cracks §20/§10 "unsteerable" from the gcc SOURCE: store-vs-load is a /s aliasing flag, the def-side wall has a macro escape, + the birthing-boost (Phase 23, byte-proven on giant func_8014EE14 248 ins ×134) _L2382
• §30a — §30 generalizes to the FULL near-miss backlog (via STANDARD Opus agents, not just Fable5) + 2 more steer levers + the mechanical-integration throughput unlock (Phase 23 (a)) _L2394
• §31 — THE gcc-2.7.2 CODEGEN MAP: pass → residual → C-lever catalog (Phase 23; 4 Fable5 agents read the compiler source) _L2402
• §32 — The region-a CAMERA-GIANT idiom set: struct-base hoisting + 4 sibling levers (Phase 24 T7, Fable5-cracked on func_80129CF8 191 ins, match_one MATCH; transferable to the 6 sibling giants) _L2422
• §33 — Automating the giant decl-reconcile: tools/reconcile_decls.py (the DATA analog of §20's cast_call_sites) + a fleet-majority type oracle (Phase 24 T7b, byte-proven on func_80129CF8) _L2434
• §34 — The func_80138ED0 giant crack: gcc-2.7.2's 3-qty sort bug + the zero-byte asm allocation toolkit + the giv-init fence (Phase 24 T5; Opus→close=21, Fable5→MATCH ×134) _L2454
• §35 — The region-a sibling-giant harvest: difficulty ≠ $s-reg count (it's global-array hoisting) + the banking recipe + new loop idioms (Phase 24 T7, 5 parallel Opus-Max agents) _L2471
• §36 — Fable5 giant-crack levers (Phase 24 T7 Fable5 batch; accumulates as each lands — full RTL dumps in .run/t7/fable/) _L2487
• §37 — The T7 §G giant endgame: 6/8 cracked, meta-laws + transferable levers (Phase 24, 2026-07-07; FULL byte-verified detail + gcc-2.7.2 line cites in docs/gcc-2.7.2-map/t7g-giant-harvest.md) _L2506
• §38 — The WHALE func_80144B9C (770 ins): the -O0 struct-assign memcpy idiom + the -O0 reach-134 ×134 rollout (Phase 24 T7 §G, cheap Opus — no Fable5, no calls.c) _L2527
• §39 — The ×1→×134 giant-endgame: propagate a matched -O2 giant via the NATIVE DEFINE-macro path (Phase 24 T7 §G close, 2026-07-08) _L2537
• §40 — Structural families: the MECHANICAL symbol-remap (crack one exemplar → remap the rest, ~0 tokens) (Phase 25 T3/T7, 2026-07-08) _L2553
• §40a — The DECL-RECONCILE pass (type-lift so remapped drafts compile in the sibling TU) (Phase 25 T7.2, 2026-07-08) _L2601
• §40b — The reloc-tracker blind spot: the indexed-global idiom that hid the "reach-1 tail" (Phase 26 Task 1, 2026-07-11, byte-verified V0/V1) _L2634
• §40c — The h_seq per-sibling reconcile: templating a reconcile-class crack ×134 (Phase 26 Task 8, 2026-07-12, byte-proven) _L2676
• §31-triage — R17 applies to CODEGEN residuals, never to a compile ERROR (Phase 26 session 7, Drew asked) _L2693
• §41 — The DEF-SIDE canonical-sig wall: mechanically banking a drafted giant past conflicting types (Phase 25 T5b batch-2, 2026-07-09; tools/canon_sig_reconcile.py, byte-proven on func_8013B274) _L2710
• §41a — v3.1: the def-side wall was ~71% TOOL-shaped — the five measured defects + the laws that dissolve them (Phase 25 T6, Fable5, 2026-07-09) _L2754
• §41b — T7 execution: the object-only probe OVER-counts BANKABLE by two link/rodata classes (Phase 25 T7-M1, 2026-07-10) _L2809
• §41c — T7-M2: the ×134 def-side-wall sweep via per-sibling RE-reconcile (Phase 25, 2026-07-10; 4,389 banks, ~0 agent tokens) _L2839
• §41b — addendum — M4 "reconcile_decls" tier is ALSO a probe over-count: 0/8 mechanical (Phase 25 T7-M4, 2026-07-10) _L2858
• §41d — void→s32 is NOT always byte-neutral: gate the RAW draft FIRST (Phase 26 session 6, byte-proven) _L2878
• §42 — The F-band ≤28 regalloc crack wave: register-pin/DENSITY levers beat the permuter (Phase 25 T7 F-band, 2026-07-10; Ultracode 9-worker wave, 4/9 banked byte-identical, 266 swept ×134) _L2900
• §42a — addendum — wave 2 (residuals + 29-100 band): iso-MATCH ≠ real-TU bank, the memcpy→struct-assign fix, +5 levers (2026-07-10b) _L2953
• §42b — addendum — wave 3 (Max, 2026-07-10c): THE STALE-OBJECT GATE TRAP + the read-global &-cast drift + fix _L2994
• §42c — addendum — wave 3 (Max orchestrator + CORRECTED Ultracode fan-out, 2026-07-10c): the real-TU-faithful parallel harness (rtu_match) + 7/9 crack, ZERO iso-drift _L3033
• §42d — addendum — wave 4 (rtu_match fan-out over the mapped frontier, 2026-07-10c): 24/26 MATCH, +5 durable levers _L3082
• §42e — propagating a CRACK ×134: the def-finder bug + the byte-drift residual (the "remap-fail" misdiagnosis) _L3121
• §43 — The K&R s16-param definition DISSOLVES the "narrow-param wall" for by-value register args (Phase 25 task A, Fable5 crack of the 369-ins giant func_80166994 ×134, 2026-07-11) _L3195
• §44 — The Phase-25 cheap-Opus giant batch: 5 structural levers + the §43 extension (2026-07-11, 6 crackers over the frontier giants) _L3242
• §45 — The flagship func_80133CD4 crack (399 ins ×134): the merged-variable permutation-breaker + the 1-death local-alloc gate (Phase 25 task A giant escalation, Fable5 gdb-on-cc1, 2026-07-11) _L3298
• §46 — The func_80178D40 crack (890 ins ×134, the heaviest core in the game): four LOOP-STRUCTURE levers cheap-Opus found by reading loop.c/jump.c/cse.c (Phase 26 session 8, 2026-07-13) _L3313
• §47 — The live-length SLIDER: splitting a global.c allocno-priority TIE with one zero-byte asm (Phase 26 session 8, Fable5 Max, byte-proven on func_8017BEBC 952 ins ×113) _L3365
• §48 — The 12-core jr crack wave: the ALLOCNO-PRICING dials and the EBB rule (Phase 26 session 8, Ultracode, 9/12 MATCH first pass) _L3405
• §3-A. — ALLOCNO-PRICING DIALS — move a value into the register you want, byte-neutrally _L3413
• §3-A4 — SINK THE CONSUMER CALL INTO THE ARMS (the inverse of A1; func_8016AB6C, byte-proven) _L3444
• §3-B. — THE EBB RULE — the general form of §46-L2 _L3468
• §3-C. — TYPE- AND SHAPE-DRIVEN CODEGEN (the C type literally selects the addressing mode) _L3484
• §3-D. — THE CROSS-JUMP RATCHET (the sharpest new trap — func_80131340 L-C) _L3508
• §3-E. — Meta _L3518
• §49 — The LUID DIAL: a zero-byte SCHEDULING dial (the sched.c analogue of §47) — func_8017A4AC (536 ins ×134), Phase 26 session 8 _L3527
• §50 — Refinements that BOUND §47/§48 (from the func_80135EB0 wall, 21→6; Phase 26 session 8) _L3576
• §51 — TOOLING INTEGRITY: the silent skip, and how to hunt it _L3630
• §51a — The bug class _L3636
• §51b — Why the byte-gate cannot save you _L3652
• §51c — THE METHOD (do not audit by reading the regex) _L3662
• §51d — THE LAWS _L3677
• §51e — The false-wall pipeline (why this is not just hygiene) _L3739
• §51f — Checklist for any new corpus-scanning tool _L3755
• §51g — When the thing you are scanning has a GRAMMAR, parse the grammar (tools/cdecl.py) _L3769
• §52 — The WALKER-FAMILY skeleton: 6 regalloc-order levers + a deeply-characterized intrinsic wall (func_80178004, 165 ins ×134; Phase 26, Fable5, 2026-07-15) _L3913
• §3-The — 6 levers (the "walker-family skeleton" — apply to the 12 siblings; levers 1-5 retire ~half the residual) _L3924
• §3-Why — the wall is (probably) intrinsic _L3945
• §52a — The regalloc sibling wave: new levers + two new wall classes (cheap-Opus applying §52, 2026-07-15) _L3959
• §52b — Sibling wave 2: more de-pin levers, a third wall class, and the match_one→whole-binary gap at scale (2026-07-15) _L3997
• §53 — SWEEP A FAMILY WITH THE TOOL ITS EXEMPLAR NEEDED: the jr/switch carve, and how omitting it manufactured the "families don't template" doctrine (Phase 28 T1, 2026-07-15) _L4028
• §3-The — case _L4033
• §3-Why — 0/8 was structural, and predictable from two words _L4045
• §3-The — rule _L4062
• §3-The — meta-lesson (R35, and why this one is expensive) _L4078
• §55 — Core-crack wave levers + the GATE-ORCHESTRATION law (Phase 29 T3, 13-agent ultracode wave, 2026-07-17) _L4093
• §55a — New byte-proven levers (each from a banked or near draft) _L4098
• §55b — THE GATE-ORCHESTRATION LAW (3 traps, ~3.5h lost; all recovered, 0 data lost) _L4122
• §55c — Sizing the propagate: a TARGETED propagate is ~4 min/core, and "it's slow" was a BROKEN-TREE ARTIFACT _L4141
• §54 — --fix-def-sig: the member's CANONICAL DECLARATION is a build step too (tiny-IMM mega-pools, +4,801, Phase 29 T6, 2026-07-16) _L4163
• §56 — Banking a hand-drafted GIANT into its exemplar TU: self-contained draft vs live-TU decls, 4 reconciliation tactics (func_8013FAF8, 312 ins ×138, Phase 29 T4, 2026-07-17) _L4187
• §56b — PROPAGATING an h_seq giant: the exemplar's externs MUST be fleet-canonical, not the draft's types _L4232
• §57 — The SELF-decl normalize: the sibling's OWN caller declares the templated fn divergently (the third §17a-1 direction; tools/normalize_self_decls.py, func_801670E4 ×137, Phase 29, 2026-07-18) _L4256
• §57a — Two NSD corrections + the SURGICAL-ONLY law + the honest broad-sweep yield (Phase 29, 2026-07-18) _L4303
• §58 — match_one MATCH ≠ BANK: the four blind spots + the crack-wave reconcile-before-bank law (Phase 29 crack-wave, 2026-07-18) _L4343
• §59 — Three h_seq sweep-residual classes match_one/the-exemplar-bank don't reveal (Phase 29 crack-wave close, 2026-07-18) _L4377
• §60 — Classify the residual, don't rank it: the deterministic residual→class classifier and what it measured about the backlog (Phase 29 Task-13A, 2026-07-21) _L4407
• What — it measured — the whole open backlog, byte-grounded _L4431
• §3-Two — corollaries worth remembering _L4450
• §3-The — parallel-probe race this surfaced _L4463
• §60a — What the first DIRECTED grinder run exposed (Phase 29 Task-13B, 2026-07-21) _L4472
• §60b — The plateau autopsy's verdict: a partial drift is a WRONG DRAFT, not a missing transform (Phase 29 Task-13B close, 2026-07-21) _L4510
• §61 — Task 14: the gate ladder's missing stage is the ARITY pre-pass, and it is TU-side not draft-side (Phase 29, 2026-07-21) _L4545
• §61a — The Task-5 wave: 11/12 MATCH, 0 banked — three DISTINCT integration walls, each now named (Phase 29, 2026-07-21) _L4606
• §61b — The jtbl gate stage: built, and the ORDERING law it exposed (Phase 29 Task-14 stage 4, 2026-07-21) _L4658
• §61c — The jtbl bank is INCREMENTALLY valid and CLEAN-INVALID (Phase 29, 2026-07-21) — the blocking finding _L4713
• §61d — The undo was eating the tree: two tools, one defect, invisible to the byte-gate (Phase 29, 2026-07-22) _L4779
• §62 — The jtbl RECONCILE must also follow the carve: the post-carve draft reconcile (harvest_verify._jtbl_reconcile, Phase 29 SESSION-11, 2026-07-22) _L4833
• §63 — The fresh-138 DEF-SIDE blocker: fix the HEADER decl, not the draft (fix_header_decl.py, Phase 29 SESSION-13, 2026-07-23) _L4877
• §64 — The §20 type-lift's three laws: fold the tagged typedef, check VISIBILITY, and strip only what is TEXTUALLY IDENTICAL (lift_types.py, Phase 29 SESSION-14, 2026-07-23) _L4908
• §64a — VARIANT types: UNIQUIFY the camps, do not reconcile them (uniquify_type.py, Phase 29 SESSION-14) _L4957
• §63 — UPDATE (Phase 29 SESSION-14) — fix_header_decl's "SAFE" verdict is FLEET-BLIND; it MUST be R22-validated _L5000
• §65 — The stranded-draft recovery: BLAST-RADIUS TIERS, and the per-overlay de-macroize that refutes §20's DEF-conflict wall (tools/demacroize.py + tools/blocker_probe.py, Phase 29 SESSION-16, 2026-07-24) _L5015
• §65a — The blast-radius taxonomy (makes §61's law structural instead of remembered) _L5030
• §65b — The escape: de-macroize the instantiation, don't touch the shared header _L5045
• §65c — rtu_match MATCH → bank held 13/13 on self-decl, but broke on the FIRST callee-decl case _L5081
• §65d — Existing-ladder baseline, measured (do this before building a recovery stage) _L5091
• §65e — Two oracles, and the disagreement is the finding (R34 in practice) _L5101
• §65f — The de-macroize lever's BOUNDARY, measured: byte-neutral 14 times of 15, and the 15th is why the gate exists _L5118
• §65g — Where the cheap levers STOP: the local-type and in-TU-self-decl classes did not yield _L5139
• §66 — Exercise a banking driver's SUCCESS path before pointing it at a wave: the free re-bank test (Phase 29 SESSION-17, 2026-07-24) _L5159
• §66a — The widest write in a pipeline is the one most likely to be UNDECLARED _L5181
• §66b — A metric parsed out of another tool's prose goes NULL silently when the label changes _L5200
• §66c — Before a wave, verify the FUEL exists; an "already attempted" set built from the wrong directory lies in BOTH directions _L5217
• §66d — The permuter⇄reader loop: alternate a random search with a byte-verified idiom, and let residual_class decide whose turn it is (Phase 29 SESSION-17) _L5245
• §66d-1 — What transfers between giants is the LOOP, not the PIN _L5280
• §66d-2 — Two operational sharp edges _L5289
• §66d-3 — Read the ILS per-cycle SERIES, not its final best: a repeated score and a still-falling one look identical in a summary line and mean opposite things _L5300
• §67 — The arg-copy PLACEMENT lever: launder a parameter into a fresh pseudo AT the statement where the target's copy lands (Phase 29 SESSION-18, func_8014D820 25 → 16) _L5315
• §67a — Run the symbol-set guard BEFORE you pay for a gate (tools/symcheck.py, Phase 29 SESSION-18) _L5405
• §66d-4 — "ILS converged" means converged FOR THAT WEIGHT PROFILE, not a floor (amends §66d-3; Phase 29 SESSION-18) _L5438
• §66d-5 — residual_class's "structural ⇒ permuter CPU is waste" is WRONG for schedule permutations (measured, Phase 29 SESSION-18) _L5491
• §68 — A comment-only line halted the extern scan, and the skip label blamed the type cap (Phase 29 SESSION-18) _L5524
• §69 — How to attack a behemoth: map it, don't draft it (Phase 29 SESSION-18, func_80183814, 5,122 ins) _L5569
• §70 — The giv-init base register: walk the PARAMETER, not a copy of it (Phase 29 SESSION-18, func_801777BC) _L5619
• §71 — Before mapping a giant, look for an already-matched SIBLING beside it (Phase 29 SESSION-18, func_8017D960, 3,338 ins) _L5660
• §72 — A register __asm__ pin is a PREFERENCE, not a reservation (Phase 29 SESSION-18, func_8017F510) _L5726
• §3-The — honest fix was source-level and cheap _L5750
• §3-Giv — record order (the §70 family) _L5760
• What — is left, and what is byte-recorded as SPENT _L5766
• §73 — A def-side self-decl conflict has TWO axes: RETURN (fleet widen, T2) and PARAMS (casts at each use, T0). Diagnose which before paying for the expensive one (Phase 29 SESSION-19, func_8014F3E8 + func_8014D4C0) _L5783
• §74 — Auditing a pinned draft: the §72 hazard is CALLER-SAVED pins spanning a call, and only the disassembly can tell you (Phase 29 SESSION-19, func_8017D960 b2, 5 pins) _L5822
• §75 — A propagation cap is usually a MINORITY-SPELLING SOURCE OVERLAY, not a wall: census the carried extern before believing the exclusion message (Phase 29 SESSION-19, func_8014F3E8 ×4 → ×138) _L5866
• §75a — The exclusion classes, enumerated with named causes (Phase 29 SESSION-19, the 134-binary dedup_extend sweep) _L5925
• §75b — A body's preamble can carry #defines, not just externs; extraction lifts only the externs (Phase 29 SESSION-19, func_80165CA0 ×3 → fleet) _L5948
• §75c — Class-B's remedy is the FULL §17a-1 PAIR (decl and call-site cast); a decl-only fix moves the error and looks like a new wall (Phase 29 SESSION-19, func_8012F14C) _L5997
• §76 — The allocno CLASS (local vs global) is the dominant regalloc lever, and C reaches it ONLY through declaration scope and variable reuse (Phase 29 SESSION-19, behemoth #3 func_8017F510 1,511 ins, 97 → MATCH, pin-free) _L6028
• §3-The — mechanism, with citations _L6037
• §3-The — attribution primitive (use this before calling anything a scheduling residual) _L6057
• §3-Two — diagnosis traps this function proved _L6064
• Practice — Practice _L6074
• §77 — Every extraction tool carries a NARROW hard-coded preamble set; anything outside it silently caps the body's reach. Diff the preamble before you gate. (Phase 29 SESSION-19 — three variants in one session, two different tools) _L6084
• §3-The — CANDIDATE gate and the REAL gate need DIFFERENT preambles — keep the difference out of the bank _L6124
• §78 — A LENGTH drift can be a register grant in disguise; and fold never leaves a literal first in an | chain (Phase 29 SESSION-19, behemoth #2 func_8017D960 3,338 ins, 1806 → 0, pin-free) _L6174
• §3-The — drift was an allocation decision, not missing code _L6180
• §3-The — economics _L6211
• §79 — For a 0-callee giant, fingerprint by DATA symbols (§71 cannot fire); and the STACK-SLOT ORDER is a declaration-order oracle (Phase 29 SESSION-19, func_8017BF14 4,763 ins, cold start → 45/4763) _L6220
• §71 — has a blind spot, and this is it _L6226
• §3-NEW — LEVER — the frame layout reads back the original declaration order _L6240
• §76 — confirmed at scale, and a pin nuance _L6251
• §3-The — residual, and the honest read _L6260
• §80 — A do-not-re-buy entry is scoped to its BASE, not to the function; and the pin's hidden cost is an unconditional qty_phys_sugg (Phase 29 SESSION-19, func_8017BF14 45 → 0) _L6270
• §3-The — pin's hidden cost, with the citation _L6291
• §3-The — flagged "#1 move" LOST — and why the failure is informative _L6302
• §78 — 's attribution primitive, run and reproduced _L6312
• Cold-start — economics, now complete _L6318
• §81 — Banking a jr (jump-table) function: the 3-step carve chain, and why match_one cannot see the problem (Phase 29 SESSION-19, func_8017C954) _L6323
• §3-The — defect this exposed: a shared type that is present but invisible _L6354
• §82 — Two source-shape oracles from behemoth #6: a duplicated addiu $aN,$sp,K across a jal means the block was INLINED, and scalar-vs-aggregate decides WHEN a stack slot is allocated (Phase 29 SESSION-19, func_8017C730 1,061 ins) _L6370
• §3-1. — The inlined-helper signature _L6375
• §3-2. — Scalar vs aggregate decides when the slot is allocated _L6389
• Also — reproduced on this function _L6401
• §3-And — the banking footnote (§75a class A, one line) _L6405
• §83 — The parameterised-repeat law, the spill-area trap, and why a per-case edit cannot move a per-case symptom (Phase 29 SESSION-20, func_80183814 5,122 ins, cold-ish start → 36 structural / 99.3%) _L6412
• §83a — READ THE HEADLINE NUMBER CORRECTLY: a LENGTH drift makes match_one's count meaningless _L6419
• §83b — THE LEVER: find the parameterised REPEAT before decoding case-by-case _L6427
• §83c — TRAP: a "dead local" in a prior draft may be gcc's OWN spill area _L6445
• §83d — CSE's quantity budget is WHOLE-FUNCTION, so a local rewrite cannot fix a local symptom _L6452
• §83e — §80 vindicated again, on the same day it was written _L6466
• §84 — The DERIVED-OFFSET remap bug: a hand-computed literal that encodes the DISTANCE between two per-overlay symbols, and why match_one is structurally blind to it (Phase 29 SESSION-20, func_8013D53C) _L6472
• §3-The — construct _L6477
• §3-Why — it survived every candidate gate _L6495
• §3-THE — FIX IS MECHANICAL — the tool already holds the answer _L6510
• Scope — , measured (do not over-generalise — §80) _L6518
• §3-Two — ladder lessons banked with it _L6528
• §85 — The RETURN-axis fleet widen is ALL-OR-NOTHING: widening the shared header alone guarantees a conflict in the source overlay (Phase 29 SESSION-20, func_8012CC88 / func_8014D12C) _L6539
• §3-The — conflict _L6544
• §3-THE — FAILURE MODE — widening only engine_core.h is worse than not starting _L6552
• §3-The — precondition, and how to check it in one grep _L6559
• §3-Do — the WHOLE axis in one edit, then R22 once _L6567
• Reading — , for the next person _L6576
• §86 — Pinned-exemplar templatability is a PER-FAMILY property, not a per-member rate; and the §42e pin guard is now over-conservative (Phase 29 SESSION-20) _L6589
• §3-The — guard refuses a class that largely works _L6591
• §3-THE — LAW: all-or-nothing PER FAMILY _L6600
• §3-Why — the two live families differ from the three dead ones — the open question _L6627
• §87 — match_one COMPILES but never LINKS, so an unresolvable data symbol reads as MATCH; and stored drafts go STALE against the tree (Phase 29 SESSION-20) _L6633
• §3-The — blindness ladder, now complete — FOUR classes match_one cannot see _L6649
• Consequence — for the backlog ledger _L6659
• §3-The — cheap discriminator, before spending a sweep _L6666
• §88 — cross_jump will not merge a common suffix containing a CALL; and §78 is scoped to ORDERED comparisons only (Phase 29 SESSION-20, the behemoth close-out) _L6674
• §88a — repeated CALL-shaped blocks are left UNMERGED; call-free tails are merged for you _L6679
• §88b — the slti literal-position law (extends §78 to comparisons) _L6686
• §88d — BANKING ORDER: run the §81 carve chain BEFORE banking, never after _L6701
• §88e — a wrong diagnosis, refuted properly (the model for how to treat an inherited lever) _L6708
• §88f — the missing rung: a RELOCATION gate between match_one and the binary _L6718
• §89 — Two throughput rules the project already had written down and was not following (Phase 29 SESSION-20) _L6726
• §89a — MEASURE the write set; do not assert its tier (tools/blast_radius.py) _L6732
• §89b — the parallel gate farm existed; the family path could not reach it (tools/sweep_parallel.py) _L6745
• §3-The — standing sequence _L6758
• §90 — Five tool-integrity laws from one session, each of which changed an answer (Phase 29 SESSION-21, 2026-07-27) _L6766
• §90a — A comparison tool MUST share its reference oracle's index space, exactly _L6772
• §90b — "Byte-neutral" is not "wanted": undo on the SUCCESS path too _L6789
• §90c — A library-callable function must FAIL CLOSED on an unconfigured module _L6799
• §90d — Do not measure a live wave's drafts (§87 in real time) _L6810
• §90e — An agent's CONCLUSION and its EVIDENCE fail independently — re-derive the premise, not the fix _L6818
• §91 — A structure-TRANSFER is only valid where the structure corresponds: the --like role trap (Phase 29 SESSION-21, func_8012AAAC ×137) _L6842
• §3-The — three-hypothesis trail, because two of them were wrong and the wrongness is instructive _L6871
• §92 — Conforming a DECLARATION: pointer changes are caller-neutral, scalar-WIDTH changes are not (Phase 29 SESSION-21, tools/conform_decls.py) _L6891
• §93 — set -o pipefail attributes a pipeline failure to the LAST stage, not the failing one (Phase 29 SESSION-21, func_8014D820) _L6922
• §94 — A family sweep's 0/N is a TYPE-CARRY failure until proven otherwise: lift the exemplar's local types, ALL of them, transitively (Phase 29 SESSION-21, func_8016B6BC 0/137 → 137/137) _L6947
• §95 — reconcile_tu dropped the SIBLING declarators of a multi-symbol extern line (Phase 29 SESSION-21, func_80176218) _L6983
• §96 — The same rewrite, one shape down: reconcile_tu matched statements to lines by TEXT, so every COMMENTED declaration was silently skipped (Phase 29 SESSION-22, func_80176218 banked) _L7018
• §97 — The gate's own tree hygiene: a refused carve, an unchecked recovery, and a snapshot that captured a dirty tree (Phase 29 SESSION-22) _L7066
• §98 — conform_decls had three defects, and only the third needed R22 to find (Phase 29 SESSION-22, func_8014CF04) _L7113
• §99 — The narrow-param wall is a DEF-side problem with a ZERO-blast-radius fix: convert the definition to K&R (Phase 29 SESSION-22, func_80175AB8 + func_80175DA8) _L7167
• §3-Two — reconcile_tu bugs found underneath, one introduced while fixing the other _L7190
• §100 — Prefer the DRAFT-LOCAL fix: a type only one function uses belongs in its BODY, not in a shared header (Phase 29 SESSION-22, func_80175DA8 0/137 → 137/137) _L7216
• §101 — The STALE DEFAULT class: a guard whose cause was removed is a silent skip wearing a safety label (Phase 29 SESSION-22, three instances in one session) _L7245
• §102 — A PLUMBING verdict can MASK a DIFF; and K&R is not always a codegen change (Phase 29 SESSION-22, func_8016EC0C) _L7272
• §103 — A FILE-scope extern in a shared overlay TU is a GLOBAL constraint on every LATER function; move the DECL, not the draft (Phase 29 T48/T51, func_80135260 — the fleet-wide half) _L7307
• §104 — Two silent-skip defects in one scan: match on MASKED text, emit from the ORIGINAL (Phase 29 T53, gather_externs) _L7395
• §105 — A gate's revert must survive an EXCEPTION, not just a failure (Phase 29 T53, jtbl_family_bank) _L7431
• §106 — Persist the MEASUREMENT, derive the POLICY: a stored route let a stale file out-vote the live table (Phase 29 T54, residual_class._ROUTE) _L7470
• §107 — A lever wired into ONE gate path is a lever most families cannot reach (Phase 29 T56, func_80144090 0/136 → 136/136) _L7513
• §108 — Diagnosing a family 0/N: the four causes, and the third opt-in lever (Phase 29 T59) _L7551
• §109 — Conforming a definition to a shared header: fix the NAMES, then check the RETURN precondition (Phase 29 T60) _L7600
• §110 — A unit must define exactly ONE function, and "ends in ;" does not tell you which line defines it (Phase 29 T65) _L7640
• §111 — The distinct-code metric is not noisy: a family pays it only if its members are byte-VARIANT (Phase 29 T66) _L7681
• §112 — A macro-scoped declaration only collides where the macro is INSTANTIATED (Phase 29 T67/T69, audit_header_sigs.py) _L7730
• §113 — An ARITY blocker only exists if the macro CALLS the function; an address-taken use has no call site (Phase 29 T72) _L7776
• §114 — The THIRD decl axis: a CALLEE the draft declares differently from the target TU (Phase 29 T76/T77) _L7808
• §115 — A func_XXXXXXXX predicate rots by design: the same name-form assumption in THREE places (Phase 29 T78) _L7850
• §116 — Optimization level is a property of the FILE, not the function: read a family 0/N against the member's stub HOME (Phase 29 T79) _L7877
• §3-The — fix moves the DEFINITION, not the stub — and here is why the obvious shortcut fails _L7895
• §117 — Spell the sibling's symbol from the SIBLING's address, not the exemplar's kind (Phase 29 T82) _L7926
• §3-Why — it survived so long: a MASKED oracle will MATCH a wrong symbol _L7953
• §118 — Ordinal (positional) immediate resolution: compare C tokens to the DIFFERING asm uses (Phase 29 T87) _L7963
• §119 — Two levers on the SAME axis, opposite directions: test the off-diagonal (Phase 29 T89) _L7997
• §120 — Uniquify draft-defined TYPE names; and check which of N staging sites you actually patched (Phase 29 T93) _L8026
• §3-Do — NOT "strip the duplicate typedef" — it breaks the extern that uses it _L8037
• §3-The — wiring trap that cost two attempts _L8043
• §121 — Synthesise externs for macro-DEFINED callees from the macro's own definition head (Phase 29 T95) _L8061
• §122 — GATE RAW BEFORE TRANSFORMING; the undo belongs to the WRITER, as a per-edit journal (P30 T0a, 2026-07-30) _L8084
• §123 — PROPAGATE A FAMILY WITH THE TOOL ITS TIER NEEDS: dedup_propagate is h_exact-only; its refusals are statements about the TOOL (P30 wave 1, 2026-07-30) _L8116
• §124 — A "not matched" verdict can mean the definition is there under a DIFFERENT C NAME: the asm-label alias blind spot (P30 SESSION-28, func_8016191C ×137) _L8153
• §124a — a family sweep's 0 matched-exemplar families may be a FILTER, not a wall _L8198
• §125 — Split the CARVE from the BODY before calling a jr residue a wall — and measure it by SHA from a CLEAN tree (P30 SESSION-28; this section's first draft was WRONG and the method caught it) _L8206
• §3-The — method (keep this) _L8213
• §3-The — instrument rules that make its answer trustworthy (this is where I failed) _L8223
• §3-The — corrected results (each SHA-verified, from a clean tree, restore re-verified) _L8238
• §3-Two — further notes worth keeping _L8248
• §3-The — meta-lesson _L8257
• §126 — The carve-within-a-carve: an ADDRESS RANGE is not an OPTIMIZATION REGION (P30 T2, byte-proven end-to-end) _L8264
• §3-The — finding: opt level is per FILE, so the file's contents must be opt-HOMOGENEOUS _L8279
• §3-The — instrument trap that hid it (and it is §124's shape again) _L8290
• §3-The — mechanics _L8299
• §126a — a bare except: continue around a coverage-asserting oracle re-creates the silent skip (P30 S28) _L8317
• §127 — The -O0 regime: the CONSTANT-OFFSET FOLD, and why -O0 needs its own idiom set (P30 T3 wave, 15 targets) _L8349
• §3-The — idiom they kept re-deriving: the constant-offset fold _L8356
• §3-The — rest of the -O0 regime (write PLAIN C, and mean it) _L8367
• §127a — §71 (sibling-first) is the strongest -O0 lever, and it beats the index _L8377
• §127b — the knowledge was in a SOURCE COMMENT, not the cookbook _L8386
• §128 — A raw NUL in C source makes grep SILENTLY SKIP the file (P30 S28, 137 files) _L8393
• §128a — a negative control must corrupt a SCRATCH COPY, never the tracked file _L8425
• §129 — Post-carve, rtu_match/match_one COUNT THE JUMP TABLE AS INSTRUCTIONS; and a carve must never be committed without its owner (P30 S28, func_8013BD74) _L8437
• §129a — the target instruction count is INFLATED after a carve _L8441
• §129b — never commit a carve whose owner is still a stub (it strands the carve) _L8462
• §3-The — real blocker underneath, for the record _L8477
• §130 — An INCREMENTAL build can report BYTE-IDENTICAL for a change the CLEAN build cannot even LINK (P30 S28, the jr pair) _L8486
• §3-The — diagnostic ladder that finally located it (reusable) _L8523
• §131 — The jtbl OVER-SPAN: sltiu N is ground truth in BOTH directions, and the zero-word rule only guards one (P30 S28, func_80191C50) _L8533
• §132 — The JR-PAIR-IN-ONE-O0-OBJECT "wall" was TWO instrument defects: a merged-double span the carve could not see, and a truncated object no rule deleted (P30 S29, func_8013B83C + func_8013BD74) _L8570
• Defect — 1 — a pre-§8e MERGED DOUBLE is not a single-table predecessor _L8579
• Defect — 2 — as writes a corpse and nothing deletes it _L8597
• §3-The — fingerprint, and the 30-second ladder that found it _L8607
• §3-The — transferable rule _L8628
• §132a — --like is for a sibling with NO record; against one that HAS a record it over-derives (P30 S29, ov_SC07_010) _L8635
• §132b — When the span's already-matched owner is ITSELF multi-switch: --span-rel (P30 S1, func_8014032C) _L8660
• §133 — The DEFAULT-FILTER class: three times in one session, a tool silently answered a narrower question than the one asked (P30 S1–S3) _L8685
• §134 — MULTI-LINE BLINDNESS: one root cause, four faces, in family_remap's preamble scanner (P30 S6, 190 zero-crack families) _L8711
• §135 — Six byte-verified gcc-2.7.2 idioms from the P30 S6f-h waves (and the two-lane wave shape that found them) _L8765
• §3-The — codegen idioms _L8770
• §3-The — integration idioms (these decide whether a byte-correct draft BANKS) _L8804
• §3-The — wave shape that produced these _L8819
• §136 — The LOCAL-VARIABLE lever: how many C locals, at what scope (P30 wave 4a, 25 byte-verified banks) _L8837
• §3-The — splitting/merging rules (each closed a residual, byte-gated) _L8853
• §3-The — type-form rules _L8879
• §3-The — scheduling rules (refining §135-2 and §135-4) _L8912
• §3-The — declaration surface (integration, not codegen) _L8941
• Wave — economics (measured, for the next batch's sizing) _L8953
• §136a — Blocker capture: classify on the OUTPUT, never on the exit status _L8976
• §136b — A prior wave's "genuine byte-DIFF" verdict is NOT reliable evidence (4 of 4 refuted) _L9033
• §136c — SIBLING-FIRST is a DERIVATION shortcut, not just a conflict fix (the fastest route in a family wave) _L9069
• §136d — Four gcc-2.7.2 levers the redraft lane found (each closed a residual no other lever moved) _L9094
• §136e — §136c's PRECONDITION, and two more symptom keys (wave 4b batch 3) _L9138
• §136f — Two declaration sub-cases the reconcile lane surfaced (lane now 15/15 lifetime) _L9186
• §136g — When the index points at the WRONG lever: two byte-refuted routings (func_801863B4) _L9211
• §136h — CORRECTION: the zero-crack pool does NOT "refill with cheap work" (my error, byte-measured) _L9240
• §136i — The drafter model LADDER: Haiku → Sonnet → Opus → Fable5 (Drew, 2026-08-03) _L9270
• §136j — The failure MIX flips with function size (measured across four bands, one session) _L9302
• §137 — REGALLOC-PERM is a TWO-COMPILE ARITHMETIC PROBLEM, not a permuter job _L9344
• §137a — A gate verdict has a TIMESTAMP; re-check it against the draft's mtime _L9386
• §138 — The propagation lanes: a gate refusal is a DECLARATION, and which lever you owe depends on blast radius _L9415
• §3-The — triage, cheapest first _L9421
• §3-The — DEFINITION-side alias is the only escape when the fleet canon disagrees on a promoting param _L9473
• Rank — the lane by measured concentration, not by class count _L9481
• THREE — carry variants hide in one "CARRY-FIXABLE" bucket — and they need different fixes _L9490
• §134 — again, in a second tool — and the waiter rule corrected _L9526
• STEP — 0 of sibling-first: grep src/ for a distinctive LITERAL from the .s _L9546
• Reconciling — a gate-refused draft: which way you edit depends on WHERE the TU's decl is _L9565
• §139 — A GATE THAT GREPS FOR VERDICTS MUST ASSERT 1:1 ACCOUNTING; and a --src filter must not survive a carve (P30 S38, wave 6: 10 of 16 drafts vanished) _L9594
• §3-The — generalisation — three corollaries worth more than the bug _L9626
• §3-And — the inverse-lookup trap, same session _L9643
• §140 — A METRIC IS NOT A MEASUREMENT UNTIL IT IS REPRODUCIBLE FROM THE COMMITTED TREE (P30 S1e: a phantom regression that gated the session's best lever) _L9660
• §3-The — three-line proof (do this before diagnosing any metric movement) _L9680
• §3-The — two instrument defects it exposed _L9689
• §3-The — same swallow, twice more, in the integration spine _L9709
• §3-Two — wrong mechanisms I chased first, and why they were wrong _L9720
• §141 — The §134 class is CLOSED: every line-shape decision now routes through cdecl._mask (P30 S39) _L9740
• §142 — An open stub whose h_exact class is MATCHED elsewhere is FREE. Propagate the body; do not gate a draft. (P30 S39, +7,710 ins in two commands) _L9778
• §3-The — measurement (do this before any wave; it is ~20 lines and needs no builds) _L9790
• §3-The — trap that hid it — SAME FUNCTION, TWO ROUTES, ONLY ONE IS FREE _L9805
• Route — selection (why --addr sometimes says "nothing changed") _L9814
• §3-And — the report-vs-bytes lesson attached to it _L9822
• §143 — cast_call_sites read a RETURN STATEMENT as a prototype and deleted it. A 0/39 sweep became 18/39. (P30 S40) _L9834
• §144 — THE LITERAL'S SPELLING PICKS THE IMMEDIATE ENCODING (P30 S40 wave 1, func_801822E0) _L9888
• §145 — Three loop/combine levers from the S40 wave-2 drafters (16/16 match_one) _L9927
• §146 — RE-MEASURE A WALL BEFORE YOU RESPECT IT. Both "permanent" giants fell to drafts already on disk. (P30 S6, +50,094 ins) _L9977
• §3-Why — a correct draft can read as an intrinsic wall _L9988
• Then — propagation returned 0/137 TWICE — both times a missing TYPE _L10000
• §3-Two — errors of mine, both instructive _L10009
• §3-The — rule _L10023
• §147 — The three-stratum FRAME LAW, and four "stop searching" verdicts (P30 S42, func_8017C294, serial run) _L10036
• §3-A. — The frame has THREE strata, and stratum 3 is unreachable from C _L10041
• §3-B. — A ?: on MEMORY operands costs ~16 bytes of invisible frame; on REGISTER operands, zero _L10057
• §3-C. — Inner-block declaration does NOT delay slot allocation — BYTE-REFUTED _L10068
• §3-D. — A lone $t8/$t9 in the target is RELOAD SCRATCH — reproduce the spill, don't pin the register _L10073
• §3-E. — A qty_compare TIE is not spelling-reachable — recognise it and stop _L10083
• Consequence — for the family (a real scheduling decision) _L10095
• §148 — The loop.c hoisting THRESHOLD is arithmetic you can compute, and the ?: clamp that folds to MIN_EXPR (P30 S42, func_8017C6F4, 947 ins) _L10145
• §3-A. — move_movables hoists iff threshold × savings × lifetime ≥ insn_count — and you can read it _L10151
• §3-B. — (v < 0x40) ? v : 0x3F is folded to MIN_EXPR and expands to the WRONG SHAPE _L10170
• §3-C. — A zero-byte ALLOCNO-PRIORITY slider _L10184
• §3-D. — Reproduce the original's BUGS verbatim _L10194
• §149 — Four instrument defects in one session, and the two questions they were hiding (P30 S43) _L10243
• §3-A. — A prep step that returns its input on failure is indistinguishable from a search that found nothing _L10249
• §3-B. — Same address + same name ≠ same body — and the ledger keys on address _L10267
• §3-C. — make: *** [...] Error N is a summary, never a diagnosis _L10283
• §3-D. — "Cheap fuel" that was never probed: 0 of 31 templatable _L10293
• §150 — A register ROTATION across symmetric blocks is VARIABLE-IDENTITY evidence, not an allocator tie (P30 S43, func_8017C6F4, 947 ins ×4) _L10312
• §3-The — fix _L10319
• §3-The — method that found it (this is the transferable part) _L10329
• §3-Two — corrections to the record _L10345
• Diagnostic — order (adopt this) _L10356
• §151 — THE GHOST WEDGE: when a load-before-store transposition is unreachable by ANY statement order (P30 S43, func_8017EF68, 969 ins) _L10367
• §3-The — mechanism (read from cc1's own -dR trace, not inferred) _L10372
• §3-The — lever — a zero-emission insn that absorbs the blocked tick _L10387
• §3-The — two fallouts, and how to close them (both measured, in order) _L10395
• When — to reach for it _L10407
• §152 — BYTE SIZE is the family key that name- and h_seq-grouping both miss (P30 S43, the 0xECC family: 1 crack → 12 overlays → 11,364 ins) _L10418
• §3-The — finding _L10423
• §3-The — key _L10432
• §3-Two — cautions that must travel with this technique _L10443
• §3-The — companion defect (open) _L10454
• §153 — THE ADDRESS-REMATERIALISATION LAUNDER: a third zero-emission asm lever (P30 S43, func_8018D98C, 710 ins) _L10466
• Symptom — Symptom _L10474
• Mechanism — (gcc source + RTL dumps, not inferred) _L10479
• What — does NOT work (14 byte-measured probes) _L10486
• §3-The — cure — a fresh launder per site, each in its own block _L10492
• Companion — levers from the same function _L10502
• §154 — Reading a disc payload: the module-id word, static base derivation, and "type 1 = uncompressed overlay" (P30 S44) _L10522
• §3-A. — Payload word0 is a global MODULE ID; code starts after the header _L10527
• §3-B. — Two static base-derivation methods that must AGREE (use both) _L10536
• §3-C. — PAC type 1 = the same payload class as type 4, just NOT compressed _L10548
• §155 — hi/lo literal scanning MUST track base registers (S45) _L10558
• §155a — the same failure class, one level up: SHAPE-blind table scanning (S45 p5) _L10567
• §155b — check the TYPE your oracle returns before comparing against it (S45 p5) _L10587
• §155c — the ZERO-REFERENCE trap: gcc splits a global-array address across the lui and the LOAD (S46) _L10610
• §156 — an ORPHANED reconcile poisons the fleet: dedup_propagate's kept edit (S45 p6/p7) _L10640
• §157 — the cheap-tier size cliff, measured (S45 p6) _L10687
• §158 — The RANGE-EXTENDER: a fourth zero-emission asm lever completes the allocno toolkit (P30 S46 tier-3, func_8017CE58, 733 ins) _L10712
• Symptom — Symptom _L10721
• §3-Why — the fork is chained (gcc source, validated insn-by-insn against -dS/-dR dumps) _L10727
• §3-The — method (dump-arithmetic first, then place — no probing) _L10740
• Bonus — facts worth keeping _L10757
• §156 — THE PREFERENCE-DONOR MERGE: cross-region variable reuse is what fills a0-a3, and a call-arg use in ONE region steers the fill in ALL of them (P30 S46 tier-3, func_80186E24, 611 ins: 236-off "S11 regalloc-order" → MATCH, zero new pins) _L10772
• §159 — THE DECLARATION AXIS: conform to byte-truth, and make every guard state its COVERAGE (P30 S47; ~10,930 sites across 8 axes, fleet byte-identical) _L10828
• §160 — THE REACH-15 WAVE HARVEST: an align-1 block move, and the instrument that called it a wall (P30 S47) _L10893
• §161 — THE RETRY-WAVE HARVEST: a jump table indexed from zero, and two allocator traps (P30 S47) _L10969
• §162 — S48 WAVE-1 HARVEST (P30, 2026-08-11): the reach-ordered sibling campaign's first 12 targets _L11014
• §162a — SHARPENS (sharpens §161a, §131, §8a-pad, §129a) _L11041
• §162b — SHARPENS (sharpens §48-A3, §156, §150, §76) _L11070
• §162d — SHARPENS (sharpens §31, §21, §30, §55a) _L11099
• §162e — NEW _L11141
• §162 — THE LICM PAIR: what makes an address a movable AT ALL, and why the preheader order is the body order (P30 S47, ov_MAIN_012) _L11143
• §162f — SHARPENS (sharpens §42d, §41d, §73, §10) _L11209
• §162g — NEW _L11264
• §162 — CROSS-JUMP DIRECTION: the surviving copy is always the LATER one, so a BACKWARD j into a sibling arm is a source goto (P30 S48) _L11266
• §162h — SHARPENS (sharpens §88, §88a, §50-B, §8) _L11300
• §162 — The cross-jump "CALL veto" is a COUNT law, not a CALL law (BOUNDS §88a; P30 S48, func_80189540) _L11302
• §162i — SHARPENS (sharpens §135, §21, §42, §32) _L11368
• §162j — SHARPENS (sharpens §25, §136d-1, §48-B, §46-L2) _L11393
• §162k — SHARPENS (sharpens §1-I2, §12, §160d, §21) _L11422
• §162l — SHARPENS (sharpens §48-B, §48-C1, §20, §21) _L11485
• §162m — SHARPENS (sharpens §36, §158, §148, §153) _L11540
• §158a — THE FIFTH LEVER IS NOT AN ASM: do { } while (0) is a REGION ref-multiplier you MINT (P30 S48, func_8017CBC8, ov_MAIN_012 / jr_801789AC, 188 ins → MATCH) _L11542
• §3-The — law _L11549
• Size — it before you write it (§158 step 1-2, applied) _L11560
• §3-The — wrap BOUNDARY is the dial — and it is indiscriminate _L11581
• §3-Not — a pure dial _L11587
• DIAGNOSTIC — TELL — two faces, one law _L11591
• §162n — NEW _L11613
• §162o — SHARPENS (sharpens §158, §136-1, §136-6, §79) _L11646
• §162p — SHARPENS (sharpens §48-B, §46-L2, §156, §136d-1) _L11705
• §162q — SHARPENS (sharpens §30, §30a, §135-2, §136-13) _L11722
• §163 — S48 WAVES 2-3 HARVEST (P30, 2026-08-11/12): the five that were byte-probed and are actionable _L11760
• §163z — THE UNVETTED REMAINDER (do not cite as law; each needs a dedupe pass) _L11828
• §164 — S48 §163z SKEPTIC PASS (P30, 2026-08-12): 190 claims vetted, 82 banked _L11848
• §16Xy — SHARPENS (sharpens §136d-3, §37 /s-DEP LATTICE, §135-2, §136-13, §162q) _L12300
• §3-The — /s drop clause is in ALL THREE dependence predicates, so the FIXED-ADDRESS STORE is what floats (P30 S48, func_8017CA18, ov_MAIN_012) _L12302
• §164z — REFUTED CLAIMS: do NOT re-derive these _L13623
• §165 — S48 WAVE-4 HARVEST (P30, 2026-08-12): banked the same day the wave landed _L13689
• §16Z — SHARPENS (sharpens §37 "the /s-DEP LATTICE", §136-13, §136-14, §16Xy, §162q, gcc-2.7.2-map/sched.md §64) _L14340
• §3-The — ADDRESS-CLASS TABLE: which load/store pairs even REACH the /s clause (P30 S48 wave 4, func_80185B44, ov_SC03_014) _L14342
• §165z — REFUTED THIS WAVE: do NOT re-derive _L14858
• §166 — THE DESTINATION-TU ORACLE (P30 S48): the seven-attempt bug that was never codegen _L14902
• §167 — S48 WAVE-5/6 HARVEST (P30, 2026-08-12): the saturation point _L14958
• §167z — REFUTED IN WAVES 5/6: do NOT re-derive _L16154
• §168 — THE COUSIN TIER (P30 S49, 2026-08-12): h_seq's exact-hash brittleness, measured — and the similarity map above it _L16211
• §169 — THE MICRO-ADAPT LANE (P30 S49): edit a proven body, don't crack a new one _L16263
• §170 — THE A-PROP WORD-DIFF CARD (P30 S49): the lane that had no owner _L16310
• §171 — THE STALE SEED SYMBOL (P30 S50, 2026-08-13): why §170's 91%→57% was never codegen _L16354
• §171a — THE MECHANICAL A-PROP DRAFT (P30 S50): 256 members banked with no agent in the loop _L16420
• §171b — THREE CARRIES THE MECHANICAL DRAFT NEEDS (P30 S50, banking the top-reach families) _L16470
• §172 — THE ORPHAN-SLOT MECHANISM v2 (P30 S50-Max): the complete frame-residue model for gcc-2.7.2 MIPS _L16503
• §172a — TWO DECOMPILATION TELLS FROM THE SAME DIG (P30 S50-Max) _L16557
• §172b — THREE MORE TELLS FROM THE GCC READ (P30 S50-Max, banked on Drew's ask) _L16573
• §173 — THE STORED-PLUMBING RECOVERY RECIPE (P31 T6): symfix-first, per-group isolation, and where the verdicts have no drafts _L16606
• §174 — THE ADAPT-CARD WAVE RECIPE (P31 waves A/B, 2026-08-14): prevention beats recovery _L16642
• §175 — A CALLER-SAVED REGISTER PIN CAN BE A CORRECTNESS BUG, NOT JUST A SCHEDULING CHOICE (P31 wave H, 2026-08-15) _L16712
• §176a — THE VERIFICATION-LAYER LAWS (P31 overnight, 2026-08-15). What each check can and cannot prove. _L16737
• §176b — BATCH-GATING MECHANICS (P31): what changes when N drafts land in ONE .c _L16777
• §176d — THE CONFLICT TABLE MUST BE SEEDED FROM THE TU, AND KEYED PER FILE (P31 S52, 2026-08-15) _L16803
• §176e — SYMBOL IDENTITY IS COMPUTABLE OFFLINE (P31 S52): tools/reloc_identity.py _L16842
• §176f — THE DECLARATION FORM IS A MATCHING LEVER, SO RECONCILE TOWARD THE FORM THE MATCH NEEDS (P31 S52) _L16899
• §176g — SIZE A WAVE BY INSTRUCTIONS, NOT BY CARDS (P31 S52 — the adopted doctrine) _L16930
• §176h — THE BATCH-SUBSTITUTION HAZARD MAP (P31 S52): seven holes, three wrong fixes, one law _L16968
• §3-A. — The seven under-reporting holes (all in gate_main, all the same shape) _L16976
• §3-B. — Typedef handling — the only strategy that survives contact _L16994
• §3-C. — The limit that remains (recorded, not solved) _L17016
• §3-C2. — RECONCILE BEFORE THE FIRST GATE — a parked draft gets HARDER to bank, not easier _L17023
• §3-D. — The measured cost shape, and what to build next _L17046
• §176i — WHAT A STATIC PRE-GATE CHECK CAN AND CANNOT PROVE (P31 S52, wave Q) _L17064
• §176j — STOPPING A WAVE MID-FLIGHT COSTS THE IN-FLIGHT TAIL (and how much is recoverable) _L17096
• §176j-2 — THE REPAIR PASS, MEASURED (do this instead of resuming) _L17122
• §176k — TWO SELECTOR BUGS THAT SILENTLY SHRINK A WAVE _L17142
• §177 — 🔴 THE EPILOGUE RETURN-DELAY SLOT IS DECIDED BY YOUR SAVED-REGISTER SET, NOT BY SCHEDULING _L17159
• §178 — SIX LEVERS MINED FROM THE WAVE-P JOURNALS (P31 S52), each byte-proven and source-cited _L17216
• §3-A. — THE $0-ADD OPAQUE COPY defeats make_regs_eqv (func_80033398, 93 ins) _L17228
• §3-B. — A return <const> IS A PRIORITY-1 HARD-REG SET THE SCHEDULER PLACES FIRST (func_8001BE30, 92 ins) _L17244
• §3-C. — SINGLE-SET TEMPS GET THE BIRTHING BOOST (func_8001D3FC, 196 ins) _L17257
• §3-D. — A NARROW TYPE BLOCKS COPY ELISION (func_8001D3FC — new idiom) _L17265
• §3-E. — THE ZERO-OFFSET ALIAS HOLE (func_80037028, 71 ins) _L17272
• §3-F. — MEM_IN_STRUCT_P ASYMMETRY IN true_dependence (func_80037144, 124 ins) _L17281
• §3-G. — TWO MODELLING TRAPS THAT COST THESE AGENTS SWEEPS OF HUNDREDS OF COMPILES _L17287
• §179 — IDIOMS MINED FROM THE WAVE P/Q JOURNALS (P31 S52, harvest pass) _L17303
• §179-A — 🔴 A LOOP-WALKED POINTER PARAMETER HANDS ITS ARGUMENT REGISTER TO THE GIV (9 byte-proofs) _L17313
• §179-B — 🔴 THE HAND-WRITTEN-ASM TRANSCRIPTION CHECKLIST (maspsx + masked_diff; 10 byte-proofs) _L17362
• §179-C — 🔴 A FUNCTION WITH NO EPILOGUE (falls into a sibling's shared tail) MUST BE FILE-SCOPE __asm__ _L17412
• §179-D — gte_stflg MUST CLOBBER "$12" OR THE WHOLE TEMP FILE ROTATES BY ONE _L17463
• §179-E — A >2*MAX_MOVE_BYTES BLOCK COPY IS A STRUCT ASSIGNMENT, NOT A HAND LOOP _L17483
• §179-F — PINNING A LOOP-WALKED POINTER IS A TOTAL OFF-SWITCH FOR STRENGTH REDUCTION _L17509
• §179-G — 🟡 A PIN CAN CREATE A COMBINE LOG_LINK AND DELETE AN andi (sixth RC-5 channel, n=1) _L17532
• §179-H — A MID-BODY .global LABEL PAIR SLICES A BYTE-COMPARABLE FRAGMENT OUT OF A LARGER ROUTINE _L17559
• Considered — and NOT banked _L17582
• §176c — MAIN (SLUS_007.26) CANNOT BE GATED INCREMENTALLY _L17599
• §176 — SEVEN LEVERS FROM THE P31 OVERNIGHT WAVES (2026-08-15): statement order, false regalloc, and the pin that fights back _L17610
• §176-A — "SCHEDULE / DELAY-SLOT / LENGTH-DRIFT ±1" ⇒ check STATEMENT ORDER around the call first _L17616
• §176-B — "REGALLOC-PERM, 1-4 instructions off" ⇒ it is usually NOT register allocation _L17640
• §176-C — 🔴 WALL REFUTATION: a hard-register pin CANNOT schedule around a call, because of a genuine gcc-2.7.2 bug _L17680
• §176-D — CSE-class levers used in reverse (two sharpenings of §153 and cse_expr §2) _L17706
• §176-E — Two cheap source spellings, both cc1-probed _L17732
• §176-F — Misdiagnosis triage: four residual verdicts that were lying _L17754
• What — is NOT banked here _L17771
• §180 — THE LEFTOVER-DRAFT HARVEST: RE-VERIFY WHAT YOU ALREADY HAVE BEFORE DRAFTING ANYTHING NEW _L17784
• §180b — WHAT THE PRE-GATE LADDER ACTUALLY FINDS IN A COLD PILE (the shape of integration debt) _L17813
• §180c — WHEN A BINARY'S MASS BAND IS SPENT, THE FLEET-WIDE DRAW IS STRICTLY BETTER _L17832
• §181 — WHAT A WAVE'S GATE ACTUALLY REJECTS (P31 S53, measured on wave R's 45-draft main pile) _L17850
• Only — ONE of 27 blocked drafts was wrong. The other 26 were correct and unbankable. _L17851
• §182 — §177's HONEST NEGATIVE: the epilogue lever cracked 4 of 16, and the 800c3 cluster held _L17906
• §180d — THE pgrep BRACKET TRICK PROTECTS THE PATTERN, NOT THE COMMAND LINE _L17920
• §183 — THE DECLARATION-RECONCILIATION PLAYBOOK (P31 S53, measured on 20 byte-verified drafts) _L17932
• §3-18 — of 20 reconciled while keeping the match. The two that did not are mechanism, not effort. _L17933
• §184 — COMMENT-BLINDNESS IS A DEFECT CLASS, NOT A BUG (P31 S53: three tools, one root cause, one session) _L18003
• §185 — EDIT THE SIDE THAT IS CHEAP TO VERIFY, AND CHECK A TU RETYPE AT ITS USE SITES _L18031
• §186 — CROSS-JUMPING RUNS AFTER SCHEDULING, SO NO C-LEVEL BARRIER CAN STEER IT _L18063
• §186b — A NO-SAVE 16-BYTE FRAME IN A LEAF FUNCTION MEANS s16 LOCALS, NOT A HIDDEN CALL _L18084
• §186c — WHERE A VALUE IS LOADED DECIDES WHICH ALLOCATOR OWNS IT, AND THEREFORE ITS REGISTER _L18092
• §187 — 🔴 "SAME SOURCE" IS NOT "SAME OBJECT": THE SDK BUILD AND THE GAME BUILD DISAGREE ON GTE NOPS _L18109
• §188 — 🔴 THE jr $ra + addiu $sp TAIL IS AN ASSEMBLER ARTIFACT, NOT A FRAME SHAPE _L18147
• §189 — FIVE COMPILER LAWS MINED FROM THE WAVE R/S JOURNALS (P31 S53), each source-cited and re-derived by a second agent _L18197
• §190 — THREE PRESCRIPTIONS FROM THE SAME HARVEST (weaker evidence than §189, honestly labelled) _L18272
• §191 — WHAT THIS HARVEST DID NOT BANK (4 rejected, 4 narrowed) — recorded so it is not re-derived _L18309
• §192 — THE PRE-GATE LADDER WAS MAIN-ONLY, AND NOBODY COULD SEE IT (P31 S54) _L18331
• Three — defects in one call path; the overlay slates that carry most of the wave work were being waved through _L18332
• §193 — THE WAVE-T HARVEST (P31 S54): 71 index_gap reports -> 9 laws, 5 rejected, 61 already-covered _L18396
• §193-A — The wave card ships only OPEN-set pointers and discards the atlas's BANKED one — exemplar/sibs are stubs 100% by construction (tools/atlas.py:96/657), while seed.ref (matched pool, atlas.py:505-536) is dropped at build_wave_atlas.py:143 _L18413
• §193-B — A NARROW CAST OF A WIDE PARAMETER COSTS ONE EXTRA INSTRUCTION WHEN ITS STATEMENT SITS AFTER AN INTERVENING jal — AND THE DECIDER IS combine.c:929's CROSS-CALL GUARD, NOT REGISTER ALLOCATION _L18480
• §193-C — gcc-2.7.2 cross_jump merges the SCHEDULED common SUFFIX only — there is no prefix/head merge, so §8/§48-A1's "duplicate into both arms and cross_jump refunds it" is a TAIL-only lever _L18518
• §193-D — A COMPILER-GENERATED ARGUMENT COPY IS A optimize_reg_copy_1 TRIGGER: when the pointer's LAST use in a block precedes a call that takes the pointer as a register, sched1 hoists the implicit move $aN,$sN to the block top and local-alloc re-bases the WHOLE block's memory operands onto $aN. The only C dial is a label (the §165-24 goto-join) between the block and the call. _L18552
• §193-E — A varying-address (pointer) load is re-emitted once per CSE-LIVE INTERVAL, and naming it in a C local is the only C-level lever over that count — no store SPELLING has any reach _L18623
• §193-F — §148-A2 — The threshold -= 3 STAIRCASE: move_movables admits a COUNT of invariants, not a boolean — two identical merged constants split hoisted/not-hoisted by LIST ORDER, and insn_count picks the rank cutoff _L18653
• §193-G — §164-54's "scope to ≥4 arms" bound is byte-wrong — the dispatch-topology oracle goes live at THREE case nodes (balance_case_nodes splits at i > 2), but only for a signed-after-promotion index _L18717
• §193-H — A pointer-derived base load *(s32*)(p+K) is uncacheable across ANY memory write or call (cse's third kill disjunct) — so the target emits one lw per store-separated RUN of spellings, and a run of stores sharing one lw is a C local you must declare _L18778
• §193-I — A DECLARED AGGREGATE LOCAL HAS AN 8-BYTE FRAME STRIDE — CEIL(size,8), NOT size. N ARRAYS THEREFORE COST 8·k MORE THAN ONE HAND-LAID STRUCT (k = how many of them have size mod 8 ≠ 0), AND THE TELL IS IDENTICAL INSTRUCTION COUNT WITH EVERY $sp DISPLACEMENT SHIFTED BY THE SAME CONSTANT. _L18818
• §193-REJECTED — what this harvest did NOT bank (recorded so it is not re-derived) _L18877
• §194 — THE WAVE-U HARVEST (P31 S54): 64 index_gap reports -> 14 laws, 5 rejected, 44 already-covered _L18904
• §194-A — A zero-byte scheduling fence goes AFTER the defining statement to make that computation emit FIRST in its block — and the barrier predicate is volatile-or-colon-less, not the "memory" clobber (COMPLEMENTS §165-40; does not refute it) _L18921
• §194-B — A addu $rA,$rB,$zero copy feeding ≥2 sh stores is a SECOND, 16-BIT-DECLARED local — the width, not the clamp or the join liveness, is what keeps the copy (and func_80183094's pin + "memory" clobber are both provably inert) _L18987
• §194-C — A CALLER-SAVED loop counter proves its live range crosses ZERO calls — so it cannot share a pseudo with any value LIVE ACROSS a call (but it may freely share one with values that merely sit between calls) _L19025
• §194-D — Declared width of a computed-value local is a sched1 dial (count-neutral) — a replication + attribution correction of §165-17, NOT a new argument-register law _L19088
• §194-E — The wave card's exemplar is the TARGET ITSELF on 42/73 wave-U and 36/71 wave-T cards — a construction consequence of atlas.py:657 (exemplar = max-nins OPEN member) meeting build_wave_atlas.py:166 (one card per gid); and the shipped §193-A seed_ref is same-binary on 0/51, so no card field can ever name a destination-TU sibling _L19129
• §194-F — if ((*p = v = f()) == 0) is an expand-time pseudo SPLITTER (store_expr's want_value && MEM path), not a fold — it partitions one value between local_alloc and global_alloc. BOUNDS §21's L1872 bullet, whose stated direction is byte-wrong on 3 of 4 instances, and CLOSES the control §167-37 asked for. _L19166
• §194-G — Reading the integer half of a 16.16 stack aggregate: REGISTER-LIVENESS, not the C spelling, decides sra vs lhu — and the break is count-neutral _L19239
• §194-H — §164-29's WAR fence runs again at SCHED2 ON HARD REGISTERS: an in-place v &= K before a branch fences the store that reads v, and when TWO independent stores compete for the one delay slot the fence picks the winner at ZERO length drift (a WIDTH/sh!=sw swap, not a nop). Corrects §167-42's reconstruction (pseudo arm → hard-reg arm) and supplies its first re-runnable A/B. _L19282
• §194-I — §16N+2's magic-per-odd-part ladder has exactly one broken row — read the divisor arithmetically instead: d = round(2^(32 + post_shift) / magic_read_as_unsigned) _L19320
• §194-J — Back-to-back identical stores: flow.c's last_mem_set deletes the first, and only volatile saves it _L19363
• §194-K — Blind sched1's alias oracle with a second SET of a pointer pseudo — the first zero-byte, non-volatile, dependence-CREATING lever (corrects §167-05's "volatile is the only door"; fourth consumer of reg_n_sets) _L19415
• §194-L — §88b and §189-E are BOTH half-wrong, but not the way the candidate says: the compare-constant shape is a 2-D lookup (cmp_info ROW × constant-in-window), and naming matters on OPPOSITE sides of the window for the GE/LT rows vs the GT/LE rows _L19493
• §194-M — A STORE in a CONDITIONAL branch's delay slot proves its C statement DOMINATES the branch — reorg can never pull a store out of either thread (gcc-2.7.2, -mips1) _L19541
• §194-N — §193-D's C dial is misstated: the lever is a SURVIVING CODE_LABEL (a label with a real incoming edge), not "a label between the block and the call" — a bare label, or a goto L; L: pair whose target is the next active insn, is deleted by jump1 (jump.c:663-669 → delete_insn → jump.c:3458-3461, and jump.c:243 for the bare case) long before sched1/local-alloc, and costs exactly zero bytes _L19596
• §194-REJECTED — what this harvest did NOT bank (recorded so it is not re-derived) _L19656
• §195 — THE WAVE-V HARVEST (P31 S54): 67 index_gap reports -> 14 laws, 9 rejected, 76 already-covered _L19677
• §195-A — §167-08's "an $aN READ before the jal is scratch" has a byte-proven FALSE-NEGATIVE class: an argument that DIES at the call is allocated straight into $aN, so its only def is a plain load far above the jal and every intervening use reads $aN — there is no positive tell in either direction, only the two-arity A/B _L19691
• §195-B — A CALL_INSN does not start a basic block in gcc-2.7.2 — so a call-crossing temp can be a LOCAL-alloc quantity (the missing precondition under §48-A2 / §52 / regalloc.md K8) _L19762
• §195-C — A call-argument %hi/%lo pair sitting at the block head, far above its jal, is a load-delay-gap filler chosen by SOURCE STATEMENT ORDER — swap the two independent statements nearest the call; a single-use void *p = &SYM; call-arg temp is OUTPUT-inert against it (but NOT expand-stream-inert) _L19804
• §195-D — §195 — masked_diff.mask_for returns 0 for EVERY j/jal word, so an internal j destination is invisible to match_one, the permuter scorer AND every similarity tier: a control-flow semantic error (which calls execute) surfaces as a 1-instruction residual mislabelled DELAY-SLOT / profile=schedule, or as an outright false MATCH _L19852
• §195-E — A 0/1 materialised at a JOIN immediately before the controlling beqz/bnez proves the source NAMED the condition — a truth expression in an if's controlling position reaches do_jump, which has no value path _L19920
• §195-F — fold-const.c:4825 canonicalises a ?: whose THEN arm is zero (or constant against a non-constant ELSE) by inverting the condition and swapping the arms — so a ternary's written arm order is byte-inert there, and c ? 0 : X is unspellable: it always compiles as !c ? X : 0 _L19978
• §195-G — §NEW — TWO ARMS CALLING THE SAME CALLEE MERGE INTO ONE jal UNLESS EACH ARM'S OWN CODE CONSUMES THE RESULT: the C dial that keeps two call sites apart is WHERE the consumer test lives, and the branch SENSE you spell it with is byte-inert (jump.c:1737 canonicalises it) _L20030
• §195-H — §165-27g — ON A FIXED-SYMBOL GLOBAL, extern T D[]; + D[0] IS THE CSE RELOAD DIAL, AT ZERO ADDRESSING COST, AND THE ELEMENT COUNT IS INERT (§165-27's T v[2] CAVEAT IS A LOCAL-FRAME FACT AND DOES NOT TRANSFER) _L20078
• §195-I — §145(b) AMENDED — the pointer-bump addiu is saved by cse, not combine: ANY second SET of a pseudo in the address's equivalence chain (in-place p += K, §145(b)'s p = r; copy, or an asm re-tie) kills the fold; the pass is cse and the gate is invalidate's reg_tick++ _L20125
• §195-J — GTE / PsyQ op CALL-vs-INLINE is a PER-SITE SOURCE FACT, not a TU style — both forms coexist in one TU (measured in 2 TUs), and the tell is the target's own opcodes (lwc2/sqr/swc2 vs jal), not the sibling. Costs -8 ins on func_8018505C. Corollary: the game's inline sqr macro emits TWO hazard nops, the SDK's Square0 body emits ONE — so the inline form is provably not Square0 inlined (a second instance of §187's SDK-vs-game GTE nop divergence). _L20190
• §195-K — At -O2 the §18 array-of-struct lever is a FRAME lever, not a length lever — but only when the N reads carry DISTINCT index expressions; with a SHARED index §18's +2-instruction residual is still alive at -O2 (the submitted "memory-loaded narrow index" precondition and unconditional length-neutrality are both falsified) _L20234
• §195-L — The cse store-re-seed does not cross a JOIN LABEL: per-arm stores + a join read keep the reload that one join store deletes (bounds §193-E BOUND 1/BOUND 3 with §48-B's EBB boundary) _L20286
• §195-M — Frame vars is a SEQUENTIAL bump-allocation, not a flat sum: §193-I's CEIL(aggregate,8) term and §165-03/§167-06's 8×orphan term are the SAME frame_offset walk at two different compiler stages, and each stage re-CEILs frame_offset to 8 before it allocates _L20333
• §195-N — In a call-bearing chain of N≥2 if (f(...)) return 1; tests closed by return 0;, the LAST test must stay in STATEMENT form — the value form (return f() != 0; / ? 1 : 0 / !!f()) costs +1 j and empties the other N−1 delay slots. The cause is REORG block placement, not jump.c's delete_jump. _L20382
• §195-REJECTED — what this harvest did NOT bank (recorded so it is not re-derived) _L20429
• §196 — PUT ON THE CARD WHAT THE TREE ALREADY KNOWS: the fleet's declaration consensus (P31 S54) _L20475
• §197 — THE WAVE-W HARVEST (P31 S54): 68 index_gap reports -> 4 laws, 3 rejected, 41 already-covered _L20530
• §197-A — A NARROW SIGNED MEMORY READ FEEDING A CONSTANT >> LOSES ITS lh, AND THE CURE IS AN ASM RE-TIE (attribution CONTESTED: cse vs combine) _L20544
• §197-B — A REPEATED COMPARE OF ONE VALUE AGAINST ONE CONSTANT IS DELETED BY cse's qty_comparison_code CHANNEL (the non-EQ complement of §165-03) — and a front-end-opaque mask on EITHER compare is a pure-C dial that keeps the target's second branch _L20582
• §197-C — Fix A1 (operand order) cannot move a commutative destination whose .greg conflict set already contains BOTH operand hard registers — split the accumulate so the destination IS the load's pseudo _L20610
• §197-REJECTED — §197-REJECTED _L20655
• §199 — THE WAVE-X HARVEST (P31 S54/S55): 63 index_gap reports -> 7 laws, 2 rejected, 56 already-covered _L20670
• §199-A — §189-A's asm→source inference is byte-FALSE: an interloper between a split constant's lui/ori is a SCHEDULE fact, not a source fact — and the separator is the BIRTHING BOOST, not a "priority floor" (§189-A's split-timing half survives; its "no statement order / no pin" absolute and the candidate's own forward-scheduler narrative both fall) _L20681
• §199-B — A permutation sweep that holds ANY statement fixed is not a sweep: the statement an agent pins as "obviously load-bearing" is the one carrying the signal, and the partial sweep returns a FLAT residual that reads as proof of order-invariance _L20730
• §199-C — A NEGATIVE CONSTANT MULTIPLY ALWAYS TAKES expmed's negate_variant — but whether you ever SEE the neg is decided by COMBINE, and for an EVEN |K| it never disappears _L20775
• §199-D — A narrow UNSIGNED value compared in an ordered if emits sltiu/sltu; the only dial is the WIDTH of a real object, and a widening (s32) cast is inert — AMENDS §35 (whose stated "only CSE-reuse canonicalizes" mechanism is byte-wrong) and does NOT apply inside a switch _L20823
• §199-E — §189-A BOUNDED AND CORRECTED — the discriminator is INSN_PRIORITY, not "is the interloper a constant": an insn between a lui/ori pair proves NOTHING about the source spelling unless it TIES the ori on priority, and on the pinned -mcpu=3000 triple a dependent load never does _L20883
• §199-F — §164-36b — THE TARGET-HEAD FENCE IS A DELAY-SLOT THREAD SELECTOR, AND IT ONLY FIRES WHEN mostly_true_jump > 0 (amendment to §164-36; its "−1 instruction" tell is falsified) _L20942
• §199-G — At TWO case nodes the switch-vs-if oracle is not blind — but the tell is ALL tests positive + a trailing j default, NOT the first test's polarity _L20996
• §199-REJECTED — §199-REJECTED _L21051
• §200 — THE ALIAS IS THE UNIVERSAL DECLARATION ESCAPE: stop negotiating with the TU's spelling (P31 S55) _L21062
• §201 — THE WAVE-Y HARVEST (P31 S55): 67 gap reports -> 5 laws, 8 rejected, 53 already-covered _L21111
• §201-A — §150-B applies to decl_prior's DEF row: for an overlay-window symbol the banked "definition" is usually another overlay's function, and the card ranks it ABOVE the destination TU _L21121
• §201-B — In a narrowed PLUS/MINUS/AND/IOR/XOR expression the destination pointee is INERT — the sign of the materialized constant is decided by an OR over the UNWIDENED operands (convert.c trunc1), which bounds §1841 to direct constant stores _L21153
• §201-C — §X — A CALL'S OWN DELAY SLOT AND THE UPSTREAM CONDITIONAL BRANCH'S SLOT COMPETE FOR ONE INSN (the call's argument copy), AND ONE STATEMENT'S POSITION RELATIVE TO THE CALL DECIDES BOTH — the residual is visible at the BRANCH, not at the call _L21233
• §201-D — THE SIGNEDNESS OF A div/mod MAGIC IS DECIDED BY THE STATIC TYPE OF THE DIVIDEND TREE AFTER get_narrower STRIPS WIDENING CONVERSIONS — NEVER BY A PROVABLE RANGE. AN & 0xFF IS NOT A CONVERSION, SO IT NEVER FLIPS THE MAGIC; A DECLARED-UNSIGNED LOCAL OR A NARROWING CAST WRITTEN AT THE DIVIDE BOTH DO. _L21284
• §201-E — §194-J-2 — The last_mem_set deletion window is measured in SOURCE/expand order, not in the printed stream; a foreign store moved between the pair in C source is a real lever, and volatile is not always the better one _L21331
• §201-REJECTED — eight, the session's highest _L21387