Entry 188

"The fields' tables for all four volumes; xfer's layout pass and the misnamed tables it found"

2026-09-26volumes

Filestools/xfer.pytools/field_tables.pycrates/piney-data/src/field/mod.rscrates/piney-data/src/field/mut_.rscrates/piney-data/src/field/out.rscrates/piney-data/src/field/qua.rscrates/piney-data/src/dungeon/out.rscrates/piney-data/src/dungeon/qua.rscrates/piney-data/src/volume.rsdocs/disc/volumes.mddocs/engine/field.mddocs/engine/battle.mdplans/volumes.mdGAPS.md

The field generator was the last of phase 0's generators besides the sound's that ran on Infection only. It failed differently on each volume.

Mutation: a name the carry got wrong

field.py reads BgMatName, which Mutation's .syms did not have. The bytes were there: Infection lays out SmallMeshName, BgMatName, BgMatName2, BGTBL and BGTBL2 0x60 apart, and Mutation keeps that layout. But Mutation's WORLD::DrawBG builds BgMatName2 before BgMatName, and WORLD::Init builds BGTBL2 before BGTBL. The data pass pairs a function's globals by the order its code builds them, so it named BGTBL and BGTBL2 the wrong way round and left BgMatName without a name. The pointer pass then followed the misnamed BGTBL, so every BG_TBL_x sat on BG_TBL_x2. Checked by content: BGTBL's row counts (4, 8, 4, 4, 8, 6, 6, 8, 8, 8, 8) are at 0x00688f70, not 0x00688fd0.

xfer's layout pass (pass 8):

• Two data-pass names of one size near each other swap when each one's words agree better at the other's place.
• Two carried globals that sit the same distance apart in both volumes, and whose words agree, bound a span. The globals between them take their offsets when their words agree: unnamed ones get a name, and a data-pass name moves.

The first try was too trusting in three ways:

• It anchored a span on a wrongly named global, and moved BGTBL2 onto a table Mutation added.
• In .bss, where no words can be compared, it placed main's SkillDamageValueAttributeCritical 8 bytes off. test_battle.py's run of Mutation's ccSkillDamageValue caught that.
• It did not see main's globals that the data pass files under an overlay.

Now:

• Only globals whose words agree anchor a span, so .bss is left alone.
• Every name in main's window counts as inside a span.

The other passes, fixed on the way:

• The pointer pass took a pointer's bytes for text: 0x006f2d98 reads "\x98-o". It then refused Infection's BG_TBL_H because Mutation's pointer does not read as text.
• With that fixed, a table that keeps Infection's layout while rows are added to what it points at voted wrongly. pcMsg4 in Mutation points row 14 at another message than "Whoohoo! ... server is open!". So the content pass now runs first, and what a global holds outranks a pointer to it.
• The content pass named pcMsg7_24 where pcMsg7_04 is, and enemyList46 on enemyList36, because in Infection those pairs hold the same bytes. Two of Infection's globals with one pattern are now both left for the other passes.

What moved. Every move on all three volumes was checked, with Infection's bytes at the new place and not the old:

• BGTBL and BGTBL2, and the BG_TBL_x reached through them;
• BgMatName;
• the Grunties' voice tables, gusoponTbl to aquaTblE, each of which sat on its neighbour;
• tradeMenuStr in Mutation, evs209Tbl, and str1070Tbl.

A few right names were lost where they were only right by luck:

• str0820Tbl: streamTbl and streamTblE point different ways, so the vote is not unanimous.
• Some of Mutation's pcMsg rows, whose text changed.

Counts are in volumes.md. The generated tables did not change:

• area, main data and statics are current;
• dungeon/out.rs and qua.rs only renamed a fog table to dungeonFog_hacked, as Infection names that row.

The tool tests pass: test_battle (Mutation's check included), test_areas, test_evscript, test_save, test_dungeon, test_field and test_font.

Outbreak and Quarantine: the compiler

field_tables.py reads WORLD::Generate's constants from its code, and Outbreak's compiler moved them:

• this is in $s2, not $s0: taken from the prologue's move.
• The objects' percentage is in $f1, not $f20: taken from the first lui/mtc1 pair after the divides.
• A nop follows the SetDungeonEnter call's delay slot, so the no-entrance branches target one word later.
• The hill loops' counters are in $s3 and $s4 the other way round: taken from each loop's decrement. The first loop makes the large hills.

water() finds WORLD.effccs's store by the same this register.

The tables

piney_data::field now has mut_, out and qua and field::tables_of(v). The four are equal except for the square root: Outbreak and Quarantine use Sqrt::Fpu. each_volume_has_its_tables compares the four Debug forms with that one field aside. The run time still uses field::INF: the game refuses other discs until phase 4.

Also

The uncited-functions audit (0186) was run again in full. Of the 484 functions left once the SDK, destructors and later-volume code are filtered out, one was an unread Infection path: ccRegisterRegularGimmick, which registers the gimmick rows each place loads.

• ccRegisterGimmick sets the row's +0x28.
• initEntryCCS gives those rows their file (+0x38).
• A town registers row 16. A field registers its seven rows and its field type's. A dungeon registers its eleven, its dungeon type's, then its attribute's and row 20 for field type 4.

The port reads a gimmick's file when it makes one, which is what the registration gives in the game. This is written up in battle.md. Also from that list:

• BreakMirror is ccBoss02Slave's (Innis, Mutation).
• ccWaterSummonsElement::Level4 is ported; the audit missed it because of its short name.