The decomp carries an authentic MobyInstance layout of 0x100 bytes with every
field named. It is now in gc_formats::moby with a 67-entry offset map, attributed
as a transcription rather than presented as our own finding.
Two fields matter:
0x40 animSeq animation sequence pointer
0x44 animSeqT animation time parameter
0x48 animSpeed playback speed
0x50 animLayers animation layer list
0x57 jointCnt joint count -- non-zero means the moby is skinned
0x58 jointCache joint matrix cache
0xBC oClass object class id -- the dispatch key
OCLASS CLOSES A LOOP THAT HAS BEEN OPEN FOR SEVERAL COMMITS. It is the field the
runtime uses to reach the lvl.vtbl handler tables, so an instance and its routine
are now linked by a named offset, rather than by an assumption about which axis
the two tables share. That assumption -- the positional correspondence -- is still
refuted, and still does not need to be true.
jointCnt and jointCache settle something from earlier in this log BY CONTRADICTION
rather than confirmation. A tie vertex's i16[3] was tested as a skin/joint index
and disproved: vertices sharing it were 0.95x FARTHER apart than vertices differing
in it. That stands and is unaffected -- this is the moby struct, and a moby is a
different object from a tie. Recording the distinction rather than letting "joints"
appear in two places look like a confirmation of the earlier hypothesis.
WHAT THIS DOES NOT DO IS FIND moby_anim. It names the instance-side state, so the
asset to locate is whatever those pointers reference, and we still do not know
where that is. The item's unblock condition is now sharper -- it needs the data
behind those pointers, not the struct.
PROVENANCE, AND WHAT THE TEST CAN ACTUALLY CHECK
This is not verified against disc data and cannot be: moby instances are runtime
objects and nothing of that shape is in the container. The test asserts what can be
asserted without the bytes -- offsets ascend, do not overlap, the last field lands
exactly on 0x100, and o_class sits at 0xBC immediately before triggers.
Worth keeping distinct from the collision finding. That project's FUNCTION ADDRESSES
did not transfer: they landed in NOBITS memory. A struct's field order is a
different kind of claim from an address in a program's layout, so one failing does
not impeach the other -- but the distinction is real, and PROVENANCE.md now records
which is which rather than citing the project as a single authority.
268 tests.
A moby class span does not start with its geometry. The first u32 is the
offset of a part table, and the table is 16-byte records of vertex
offset, vertex count, an unestablished second count, the next part offset
and flags.
Parts tile: vertex_offset + 16 * vertex_count == next_part. The stride is
never assumed anywhere -- it falls out of that identity, which is what
pins the record size and the field boundaries. The identity holds for
43,467 of 43,467 part records across every moby class on all 27 levels,
and the parser rejects any span that violates it rather than returning a
plausible wrong answer.
Two terminations are needed. A zero vertex_offset ends the run, but 172
records on the disc have a zero vertex_offset and a zero next_part while
still carrying a vertex count. Stopping only on the first leaves 99.6%
instead of 100% and quietly includes a second table as if it were parts.
1,270,802 vertices over 4,560 class spans; 116 spans carry no geometry.
Vertices are 16 bytes of VIF-packed half-float quads cycling through four
distinct words. Handed back raw -- unpacking them is separate work from
finding them, and second_count is recorded but not guessed at.
ReaperOfVeriod