Files
Sylpheed/crates/sylpheed-formats/examples/frame_alpha_census.rs
sylph-decoder 088df913b2 re: T8aD +0x04 bit 0x02 predicts the MEASURED blend, 35/35 -- and a prediction to test it
REFUTED.md kills this claim: 'T8aD +0x04 bit 0x02 selects an additive blend ->
mine, and refuted. Blending those sprites additively worsens every measure
against the capture.' That refutation rests entirely on our renderer, which the
corpus's own rule calls a hypothesis under test. The blend is now measured off
the GPU, so the claim can be tested against the oracle.

35 elements over three screens, every label an RB_BLENDCONTROL0 value read from
the command stream: 16 bit-set and additive, 19 bit-clear and alpha-over, zero
false positives, zero false negatives.

The control that makes it a decode rather than a coincidence: of every bit of
the first 12 header words, EXACTLY ONE separates those 35 elements without
error. Nothing ties with it. A perfect partition on a small sample is worthless
if half the header partitions equally well, which is the mistake +0x08 = 0x8050
was.

And the pair no confound survives: ptbtn00 = 0x0110, ptbtn00f = 0x0112 -- the
PRESS (A) plate and its own highlight, same screen, differing in exactly this
bit, drawn alpha-over and additive respectively.

Committed alongside is a PREDICTION for GP_OPTIONS, written before the capture
that tests it: a different archive, a different element set, and a MIXED
prediction -- po_menu_eff01/02/03 additive, 592 elements alpha-over. Falsified
if those three draw alpha-over or anything else draws additive. The developer
splash was considered first and rejected as a test: both its elements predict
alpha-over, so it can fail but cannot discriminate.

Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01Wuu56cE8vJGTBtn1ppsk8v
2026-08-31 07:06:07 +00:00

65 lines
3.0 KiB
Rust

//! The main menu's sprites, by their ALPHA channel — is `ptframe1`/`ptframe2`'s
//! "no fully-opaque pixel" a property of the artwork, and does any alpha value
//! look like a scale the game expands (e.g. 0..128) rather than 0..255?
//!
//! The port measures both frames as rendering too DARK against the capture, with
//! the shortfall correlating with the BACKGROUND. Two different causes predict
//! that: a background-scaling blend selected in code, or an alpha that is too
//! LOW in our decode. This example tests the second, which is on the disc.
//!
//! cargo run -p sylpheed-formats --example frame_alpha_census
use sylpheed_formats::{pak::PakArchive, t8ad, ui_layout};
use std::path::PathBuf;
fn census(name: &str, img: &t8ad::T8adImage) {
let n = (img.width * img.height) as usize;
let mut hist = [0usize; 256];
for p in 0..n {
hist[img.rgba[p * 4 + 3] as usize] += 1;
}
let zero = hist[0];
let full = hist[255];
let max = (0..256).rev().find(|&a| hist[a] > 0).unwrap_or(0);
let nonzero = n - zero;
// the top five alpha values that actually occur, by population
let mut top: Vec<(usize, usize)> = (1..256).map(|a| (hist[a], a)).filter(|&(c, _)| c > 0).collect();
top.sort_unstable_by(|a, b| b.0.cmp(&a.0));
let top5: Vec<String> = top.iter().take(5).map(|&(c, a)| format!("{a}x{c}")).collect();
println!(
"{name:<16} {}x{:<4} px={n:<8} a=0:{:5.1}% a=255:{:5.1}% max={max:<3} \
partial(1..254)/nonzero={:5.1}% top:[{}]",
img.width, img.height,
100.0 * zero as f64 / n as f64,
100.0 * full as f64 / n as f64,
if nonzero > 0 { 100.0 * (nonzero - full) as f64 / nonzero as f64 } else { 0.0 },
top5.join(" ")
);
}
fn main() {
let root = PathBuf::from(std::env::var("SYLPHEED_DISC").expect("SYLPHEED_DISC"));
let argv: Vec<String> = std::env::args().skip(1).collect();
let pak = argv.iter().find(|a| a.parse::<usize>().is_err())
.cloned().unwrap_or_else(|| "GP_TITLE".to_string());
let ar = PakArchive::open(root.join(format!("dat/{pak}.pak"))).expect("pak");
// Builds default to the two the port ships and can be overridden, so the
// same census serves the title (4) and the `PRESS (A)` plate (2).
let args: Vec<usize> = argv.iter().filter_map(|a| a.parse().ok()).collect();
let builds: Vec<usize> = if args.is_empty() { vec![5, 6] } else { args };
for build in builds {
let Ok(by) = ar.read(&ar.entries()[build]) else { continue };
let Some(b) = ui_layout::parse_build(&by) else { continue };
println!("=== {pak} build {build} ===");
let mut names: Vec<&String> = b.sprites.keys().collect();
names.sort();
for n in names {
let (off, size) = b.sprites[n];
let s = &by[off..(off + size).min(by.len())];
match t8ad::parse(s) {
Some(img) => census(n, &img),
None => println!("{n:<16} (not a T8aD / failed to parse)"),
}
}
}
}