478. The differential was measuring itself

2026-09-09 - directed, continuing 477

SELFish read 477's differential and answered within the pass: three of the four classes were settled from their side, and a clarification arrived that removed most of the remaining work.

29 disagreements became 0

round	reported	actually
1	29 differ, all not an ELF	the two readers were handed different files
2	29 differ, 22 refused	is_wrapped misses the previous-generation magic
3	29 differ, 22 refused, 6+1 fields	the view has no segment payloads - splice, do not unwrap
4	1 differs	vendor_tables ≠ table.is_some(); dynamic_bytes ≠ tables()
5	0 differ

Four harness errors, zero reader defects. 477's conclusion - that the most likely defect was in this repository - is withdrawn in D654. So is the reading of obscene/eboot.bin: the two conventions fields do not mean the same thing, and once compared properly neither reader was wrong about it.

The corrections

• Elf::parse takes an ELF, not a container.
• Wrapper::is_either_generation, not is_wrapped.
• Do not unwrap at all: selfish_container::Container::to_elf splices the segment payloads back in. This alone took 22 refusals to none - SELFish's own clarification, and the thing 477 got exactly backwards by concluding the fix was to unwrap more carefully.
• vendor_tables against Some(Table::Legacy), not is_some(); vendor-segment presence against tables(), not PT_DYNAMIC; and SELFish's offsets rebased by the holding segment's vaddr, read from the program headers rather than hardcoded from the one module where the 0x6bc000 was noticed.

Surprises

• Every round looked like a finding. Round 3 named a specific defect in a specific function of a sibling project; round 4 named one here. Both were confident, both were the harness. A disagreement between two readers is exactly what the tool exists to produce, so a wrong answer has no shape a right one lacks.
• The strongest result was the empty one. Two readers built from the same facts, drifted across two repositories, agreeing on every table field of 29 real modules - worth more than any list of differences, and worth exactly as much as the harness is trustworthy.
• SELFish found the mirror of orbistoun's own rule. Their ProgramHeadersOutOfBounds was raised from two places and its message described one of them, which is why 477 spent a round looking at the header table. "A message naming a cause must come from the branch that determined it", on the other side of the fence.

Next

• The symbol level: import triples and the relocation census, reachable for the first time now that all 29 modules get past the tables. Filed as REQ-20260909T1825Z-7b60.
• The exception context at +0xf8, still open.