CARVE ANSWER (V9) - B0040C3C accepts a name with a `"` OR a `.`; the arg contract, stated¶

Answers "H2 refined: B0040C3C IS LOAD's arg-resolver" (relayed 2026-07-18). Builds on CARVE-ANSWER-LINKER-LOAD-ERROR52-V8.md. Binary: D:\ND\500\nd-linker\linker-b01.dom.asm. ND-500 VAs (hex). Tags: [V] = read from these bytes; [I] = inferred; [OPEN] = not proven.

TL;DR - I under-stated the success condition; there are TWO accept exits¶

B0040C3C succeeds if the parsed argument contains either a " (0x22) OR a . (0x2E) - not only a .. A bare colon B:NRF has neither, so it errors 52. This reframes the whole thread: the arg contract is "quoted string, or a dot-bearing (name.type) spec", and SINTRAN colon form B:NRF satisfies neither at this parser. [V]

1. The two accept exits (byte-exact) [V]¶

After the stage-loop, exits converge at B0040D27; then (B0040D27-B0040D84):

B0040D2F w2 comp r3 ; B0040D31 if > go B0040D44   ; lo>hi -> straight to DEABF
         (lo<=hi:)  QUOTE SCAN:
B0040D33 by1 := $0x22 ('"')
B0040D38 by comp2 @b.0x18+,r1 ; B0040D3C if >< go B0040D40   ; char != '"' -> keep scanning
B0040D3E bi1 clr ; B0040D3F ret                              ; char == '"' -> ACCEPT (exit A)
B0040D40 d loopi b.0x3C,b.0xC0 -> B0040D38                   ; scan [lo..hi] for a '"'
         (no '"' found:) fall to
B0040D44 ... B0040D5C call DEABF (B004D4F4) ... 
B0040D75 (K gate; B0040C44 bookkeeping)
B0040D7D by comp2 b.0x49,$0x4 ; B0040D81 if >< go B0040D85   ; b.0x49 != 4 -> ERROR 52
B0040D83 bi1 clr ; B0040D84 ret                              ; b.0x49 == 4 (a '.') -> ACCEPT (exit B)

- Exit A (B0040D3F): the argument contains a " (0x22) somewhere in [lo..hi]. [V] - Exit B (B0040D84): the stage counter hit 4, which only a . (0x2E) does (;->2,)->3, .->4; : sets flag b.0x4C only). [V, from V5] - Everything else -> B0040D85 error 52. B:NRF (no ", no .) -> error. [V]

2. Who sets the `{0xB0048FEC,5,9}` = "B:NRF" window [V]¶

B003C8F4 (LOAD's arg handler) builds the descriptor b.0x20 as a raw slice of the command line 0xB0048FEC [lo=5, hi=9] = "B:NRF" (indices 5-9 of "LOAD B:NRF"). It is handed down through B003CFDA (recursive spec parser: entry B003CFDA, two B0040C3C call sites at B003D064 and B003D08F->B003D0E2). B003CFDA DOES make the dispatcher calls (B004D4F4 selectors 0x28/0x23 at B003D012/B003D03F), but on the taken path the descriptor reaching B0040C3C is still the raw {0xB0048FEC,5,9} colon window - no dot-form buffer and no quote are ever inserted. [V]

So nothing in the emulator's path puts a " or . into the argument, which is exactly why it errors. The open question is only whether real HW's argument HAS one - and section 3 gives a live test that settles it in one line.

3. The refined arg contract + a decisive one-line live test¶

For B0040C3C to accept LOAD's object argument, the descriptor it parses must contain a " (exit A) or a . (exit B). Candidates for how that happens on real HW, now testable: 1. The argument is QUOTED. If the linker's LOAD expects LOAD "B:NRF" (double-quote 0x22), the quote-scan exit A fires. TEST: type LOAD "B:NRF" (double quotes) in your rig. If it gets past error 52, exit A is the intended contract and the emulator's window-extraction is dropping the quotes. [decisive, [I] until you run it] 2. The argument is dot-form. LOAD B.NRF (or a canonicalizer that emits name.type). TEST: type LOAD B.NRF. If it passes, exit B is the contract and colon->dot canonicalization is missing. [decisive] 3. If BOTH "B:NRF" and B.NRF still error 52, the descriptor window itself is wrong (lo/hi or ptr), independent of content - then dump the descriptor bytes B003C8F4 builds on a known-good HW run. [fallback]

Note SINTRAN's own filespec quoting is the apostrophe ' (0x27), but this parser scans for " (0x22) - so if quoting is the answer it is the linker's string-literal quote, not SINTRAN's. Run test 1 and 2; whichever passes names the contract. [V for the 0x22 vs 0x27 distinction.]

4. Honest status¶

This is round ~6 of the error-52 thread. The banked, committed fixes (i1=0; current-user->SYSTEM fallback) stand. This gate has converged to a single, now-testable question: does LOAD's argument need a " or a . on real HW (exits A/B), which one line at your prompt (LOAD "B:NRF" / LOAD B.NRF) will answer. Recommend pausing further static drilling until that test picks the exit; the answer then tells us exactly what B003C8F4/the input path must produce.

Evidence register¶

Two exits: quote scan B0040D33-B0040D40 (04 CD 22 / 2D E5 18 D0 / C6 04 / 84 / 80 / BF 4F 70 F3); stage gate B0040D7D-B0040D84.
B003CFDA recursive parser entry B003CFDA; B0040C3C call sites B003D064 & B003D0E2; dispatcher B004D4F4 selectors 0x28(B003D012)/0x23(B003D03F).
(From prior rounds) window {0xB0048FEC,5,9}="B:NRF" set by B003C8F4; stage map B0040C74-B0040CC7.

CARVE ANSWER (V9) - B0040C3C accepts a name with a " OR a .; the arg contract, stated¶