2026-09-09 - The lesson put in the tool, and the reversal made airtight

Worklog 049 reversed a confirmation: the container that matched all nine fixed header rows declares itself fake, so the match was a round trip. The rule got written down twice that afternoon and applied once. Writing it down a third time was not going to help, so it went into the tool instead.

The guard

selfish audit now reads ex_info.ptype at the offset the table pins and prints what kind of container it read above the row count:

generation  Current (from the magic)
kind        ptype 0x1 - fake
            this container declares itself FAKE - it agrees with the table because it was
            written from one, so a match is a round trip and says nothing about vendor material

confirmed   9 of 9 fixed header row(s) match the table

Above, not below, because 9 of 9 is the number a reader keeps and a caveat under it is a footnote. A footnote is exactly what failed.

The value is reported, not tested: Declared::Ptype { value, known } carries the raw number and the name the table gives it, if any, and Declared::Unreachable is a separate answer. Using "is this a known ptype" as the test for "did I find the tail" is what produced nine ex_info | not located lines in one sweep - it fails on precisely the material worth having, and then reads as a fact about the format. is_round_trip() is false for an unknown value on purpose: not known to be fake is not known to be genuine. (D093)

What the crate already knew

a_container_this_crate_builds_confirms_every_fixed_row_of_the_table has been in the suite since D084, and its own comment calls the thing it asserts a round trip. The knowledge was written down in a place that runs on every commit and never reaches a person reading output. That is the more useful half of this entry: a fact recorded only in a test comment is not recorded where it is needed.

The reversal, made airtight

Wrapping a module and reading its tail gives, byte for byte:

field	this crate's output	PPSA03416 as measured
paid	0x3100000000000002	0x3100000000000002
ptype	0x1	0x1
app_version	0x0	0x0
fw_version	0x0	0x0
npdrm/type	0x3	0x3

Five of five. The container reported as a retail vendor game has the exact ex_info this crate writes. Whether it came from here specifically or from another OpenOrbis-lineage toolchain is still not knowable from here, and does not matter: it is not vendor material, and D092 stays reversed.

Not done, deliberately

The audit does not refuse a fake container, warn on stderr, or change its exit code. Auditing one is legitimate - it is principle 4's round trip and a test depends on it. The defect was never that the round trip happened, only that the output would not say which one it was. A tool that turned this into an error would be answering a complaint nobody made.