The check identifies the update by hash among layout.SUPPORTED and records it in totk-install.json; an install that
records none is 1.2.1. Extraction, lifting, compiling and preparing take the version from there: the patch table and
expected hashes per version, build_game.py --version (rt_game_version), make_romfs_patches.py and prepare_shaders.py
--version. Step keys of a 1.2.1 install stay as they were, so updating such an install does not extract again. Another
update file than the recorded one plans the extraction again (a copy of the same file is recognised by its hash), and
an identified version change re-plans every step after the check. The disk space estimate takes the larger RomFS.
Catalog strings name both versions and take the version as a parameter.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
The jump-table check now walks every way into the dispatch region up to four predecessor blocks and drops an oracle
index only where each way's branch condition, a pure function of the index register evaluated in Unicorn, rules it
out: cmp #k; b.ne with other instructions between, cbnz on the index, and range checks such as sub/cmp/b.hi or
and/cmp/b.ne. Branch targets inside a fall-through region count as entries. capstone lists the first operand of
cmp, cmn and tst as written; the slice reads it from the operands. 1.4.3 drops from 6 unexplained disagreements to 0,
1.2.1 stays at 0.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
The jump-table check now explains oracle targets for case indices peeled off by cmp #k; b.ne before the table.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
The check step now emits a space event with the bytes the install needs and the free bytes. The window logic (event model, form, CLI subprocess, Ctrl+Break) is in installer/guilogic.py and tested without Tk.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>