Entry 202

"The carry made reliable and every reader through it: Mutation boots from its title into the desktop's setup"

2026-09-26volumes

Filestools/main_data.pycrates/piney-data/src/exe.rscrates/piney-data/src/main_datacrates/piney-battle/src/tables.rscrates/piney-battle/src/party_chat.rscrates/piney-stream/src/table.rscrates/piney-stream/src/effect.rscrates/piney-stream/src/ending.rscrates/piney-effect/src/particle/tables.rscrates/piney-effect/src/spell.rscrates/piney-effect/src/files.rscrates/piney-demo/src/names.rsplans/volumes.md

After 0201 Mutation's title worked. Its New Game led into streams and the desktop, which stopped at readers still using Infection's addresses directly. This entry covers four things: the carry, the readers, the streams and titles, and a check of the whole source.

The carry

Four changes to tools/main_data.py's carry, each forced by a wrong or missing address:

• Votes at a global's start come first. Mutation's skillTbl (Infection 0x0061f4a0) had two votes each for 0x0064e2f0 and 0x0064f4a0. Code that indexes into a table votes for "its start", computed from Infection's offset, and the table grew in Mutation. The carry now counts only the votes at a global's own start when there are any. 0x0064e2f0 is right: its rows read "NOTHING", "ATTACK", "Data Drain".
• The code outranks a name. Where a unanimous vote disagrees with the name xfer.py carried, the vote wins (96 cases on Mutation). Most are a run of $gp variables off by 4, after a variable Mutation inserted; one is skillTbl, which the plan had flagged. Each case is printed.
• An overlay's copy of a main global. Outbreak keeps a tpcTradeList in GCMN.PRG as well as main's. GCMN's votes for it now go under the overlay's key, and Image::carried tries the loaded overlay's key first.
• Content, and neighbours. Globals the other passes miss, of 8 bytes or more with no pointer in them (Infection's relocations say which), are carried where their bytes occur exactly once in the volume's section. In carried, an address between two carried rows that lie the same distance apart in both executables maps by its offset from the lower. That covers a run of local arrays, such as the tornado's @1868 .. @1872.

The readers

• The first sweep (0200) wrapped every direct image.read(CONST ...).
• This one did the indirect uses by hand:
  • constants passed to helpers (the battle tables' rows, shorts, ints, item_lists) or chosen first (if english { STREAM_TBL_E } else { STREAM_TBL });
  • the chat tables, keyed by Infection's address but read at the volume's.
• Helpers that only ever get an Infection address now carry inside: read_f32s, read_i16s, read_rows. So do GenParam::read and FfParam::read, which keep Infection's address as the row's identity (p.va - GENERATOR_TBL).
• Reads of literal addresses (exe.u32(0x0037_7f10)) are wrapped the same way.

Streams and the title

• The stream archive table has six types: CMN, STR1, STR2, STRSUB, STR3, STR4, the order the later ccCdInits list them.
• Each disc's opening movie is its own: opening.pss, opening2.pss, opening3.pss, opening4.pss.
• eventObjTbl_0580 and _0581 are carried separately. The port had read them as one run.

Checking the whole source

A script checks every Infection data address the crates name (constants and literals, 867 once other volumes' own values are left out) against each volume's carry, using the same lookup Image::carried does. It misses:

Mutation's 8 are each referenced by one function Mutation changed, where no vote can align:

• ccGame::ChangeScene's @1489;
• ccEvent::Execute's @5293-@5295;
• InitSpcParam's default item lists;
• ConvergenceSystem's @3328;
• the name entry's STR_INFO_MSG_0.

Mutation now boots through its logos, its title (with its own opening stream 21), New Game, its opening stream 2, and into the desktop's setup. The desktop's events then stay in their first phase. Infection's tests all pass unchanged.