Skip to content

Frame generation: real-file check tool (execute-only, digests and verdict only) - #178

Merged
phantoslayer21 merged 1 commit into
mainfrom
fg/real-check
Nov 14, 2026
Merged

phantoslayer21 merged 1 commit into
mainfrom
fg/real-check

Conversation

@phantoslayer21

Copy link
Copy Markdown
Owner

Adds a tool for the frame-generation real-file check: does our converter convert a player's own library with the preset-FG map, with every stage digest matching? It records digests and the verdict only (lane h). It stacks on #177 (the map and the converter's offset axes); after #177 merges, this PR's base becomes main.

clean/fg/tools/real_check.py

python -m clean.fg.tools.real_check --exe <wronggpu-convert> --dll <one exact library path> [--map data/fg.v1.map.json] [--work <folder>] [--record <file.json>]
  • Execute-only: the converter reads the library; the tool never opens it. It runs the converter once, with no --raw, as a console-less child at below-normal priority.
  • What it records: the verdict (converted, and the matching stage digests equal the map's stage count), the exit code, the player message, the map's SHA-256 (from the log and from the file), the library's file version and size, the number of matching stage digests, and the SHA-256 and size of the converted model file.
  • What it never records: offsets, values or bytes of the library. The converter's log is not kept.
  • Cleanup: it deletes the converted model file it wrote, file by file (only its own output), and refuses a work folder that already has output.

Refs #96

Tested

  • python -m unittest clean.fg.tests.test_fg_real_check: 6 tests pass locally. They use made-up converter logs (pass, fewer digests, a failure exit, empty output) and a stand-in converter script written by the test: a pass removes the model file, a failure leaves nothing, and a used work folder is refused. The record is checked to hold no offsets.
  • Locally, with a build of Converter: the delivered preset-FG map, storage order and offset axes #177, the tool was run once on a player's own library from a game install. The verdict was pass and the record held digests only. That record stays private and is not part of this PR.
  • The whole clean/fg test set and the line-ending check pass. The public gates pass, with no hits in clean/fg.

Not tested

  • CI runs only the unit tests; it has no library file, by design.

…dict only)

clean/fg/tools/real_check.py runs the converter once on one given
frame-generation library with the preset-FG map and records only the verdict,
the map's digest, the file version and size, the number of matching stage
digests and the converted file's digest; it deletes the converted file. It
never opens the library, and never records offsets, values or bytes of it.

Refs #96

Signed-off-by: phantoslayer21 <156617979+phantoslayer21@users.noreply.github.com>
@phantoslayer21
phantoslayer21 marked this pull request as ready for review November 14, 2026 17:02
Base automatically changed from fg/real-map to main November 14, 2026 17:24
@phantoslayer21
phantoslayer21 changed the base branch from fg/real-map to main November 14, 2026 17:25
@phantoslayer21
phantoslayer21 merged commit 31a88bc into main Nov 14, 2026
9 checks passed
@phantoslayer21
phantoslayer21 deleted the fg/real-check branch November 14, 2026 17:25
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

1 participant