Files
Sylpheed/tools/re-capture/boot_timeline_probe.py
sylph-decoder 14fced07b6 re: I cannot measure this emulator's clock -- and that answers the port's 2%
The port put two of my pages against each other: settle->plate 2.135 s and one
focus-ring revolution 2.177 s, both a declared 120 units during a static hold,
2% apart against a 6 ms run-to-run agreement. Fair challenge.

The resolution is that the question assumes a stable wall clock. Same interval,
same container, same day: 2.138, 2.132, and 2.549 s -- a 19% swing, caused by
adding --log_ui_draws=true. The 2% is a fifth of that. The two pages were never
in conflict about the game; they are three readings of one declared quantity
through a clock that moves. What settles the quantity is the disc.

Wall clock cannot separate the hypotheses, so I tried to measure frames instead.
Both instruments are recorded as failures rather than published as numbers:

  * Canary's own [UI-CAP] counter -- the one that produced the corpus's 28.5 fps
    -- costs a third of the frame rate. 300 frames in 16.567 s = 18.11 fps on a
    screen that gives ~28 without it. That reclassifies 28.5 as a load-dependent
    lower bound; it does not overturn it.
  * A distinct-frame counter over the spinning ring FAILED its decisive control:
    15.88 fps against the game's own 17.59 in the same window, 10% low, so the
    ring does not change on every presented frame. Its static control also read
    2.63 instead of ~0. Dead, not tuneable, per METHOD.md.

The rule that follows, and it applies to everything I hand the port: a measured
interval landing near a round number of declared units almost certainly IS that
number of units. Ship the units.

Also recovered here, because the same question needed it: the static PPC route.
Four tools open /work/xenia-rs/sylpheed.db and nothing in this repository builds
it -- no disassembler, no PPC decoder, and default.xex is encrypted (zero
plaintext "GamePart"). Xenia decompresses the image at load, so dump_image.py
reads it out of guest memory and validates it against the corpus's own landmarks:
the 29-entry GamePart id table at 0x820A1630 and the Xbox 360 D3D runtime
strings. String search and table dumps work again; instruction-level work does
not, and the present interval I wanted is an immediate, not a string.

Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_014voBspJ6kFncNErZJuZcLw
2026-08-29 13:16:02 +00:00

235 lines
9.3 KiB
Python
Executable File

"""Time the whole boot, and measure the PRESENTATION RATE per screen.
Two jobs, one oracle session, because they need each other.
1. **The boot timeline the port asked for.** Every screen-to-screen transition
from launch to the main menu, wall-clock, with the black holds marked. The
port paces its boot off `ScreenView.settle_time()` = a group's `rest.t`, and
`rest.t` is NOT when a screen settles (docs/re/REFUTED.md) — so every screen's
dwell is currently wrong by an unknown amount.
2. **Frames, not seconds.** Two pages of the corpus measure the same declared
120 keyframe units during a static hold and disagree by 2 %: settle→plate is
2.135 s (28.10 fps implied) and one focus-ring revolution is 2.177 s (27.56
implied). Either the presentation rate differed between those sessions, or one
interval is not 120 units. The two were taken on DIFFERENT SCREENS, so this
measures the rate on each — with the game's own frame counter, not a guess
about what changes between grabs.
`--log_ui_draws=true --ui_draw_capture_frames=N` makes Canary log
`[UI-CAP] capture armed` and then `[UI-CAP] done: D draws over F frames`. Timing
between those two lines in its own stdout gives frames/second directly, and it
re-arms (the log index is `{:02d}`), so one session can measure several screens.
⚠️ The instrument can perturb what it measures — writing a draw log costs the
emulator something. Control built in: the ring period is measured both DURING a
capture and OUTSIDE one, and a rate that is an artefact of logging would move it.
boot_timeline_probe.py --control
boot_timeline_probe.py --run SECONDS OUT.tsv CANARY_STDOUT [shots_dir]
"""
import os
import subprocess
import sys
import time
import numpy as np
from PIL import Image
SD = os.path.dirname(os.path.abspath(__file__))
sys.path.insert(0, SD)
import title_timing_probe as T # noqa: E402 (same crop, same ZNCC, same controls)
REPO = os.path.dirname(os.path.dirname(SD))
CAP = os.path.join(REPO, "docs", "re", "captures")
# The boot shows two splashes before the movie; both are committed captures.
T.REFS["splash_pub"] = "title-builds/live-splash-publisher.png"
T.REFS["splash_dev"] = "title-builds/live-splash-developer.png"
T._R.clear()
RATE = 8
CAPTURE_FRAMES = int(os.environ.get("UICAP_FRAMES", "300"))
def control():
"""Every control title_timing_probe has, plus the two splashes."""
T.CONTROLS.extend([
(os.path.join(CAP, "title-builds/live-splash-publisher.png"), "splash_pub"),
(os.path.join(CAP, "title-builds/live-splash-developer.png"), "splash_dev"),
])
return T.control()
def _tail(path, seen):
"""New lines appended to the emulator's stdout since the last call."""
try:
with open(path, "rb") as f:
f.seek(seen)
b = f.read()
return b.decode("utf-8", "replace"), seen + len(b)
except OSError:
return "", seen
def arm_capture(log_path, seen, timeout=90.0):
"""Press F10, then time the emulator's own armed->done lines.
Timed between the two LOG lines, not from the keypress: the arm latency is
then excluded rather than folded into the rate.
"""
win = subprocess.run(["xdotool", "search", "--name", "Xenia-canary"],
capture_output=True, text=True).stdout.split()
if not win:
return None, seen
w = win[-1]
subprocess.run(["xdotool", "windowactivate", w], capture_output=True)
subprocess.run(["xdotool", "key", "--window", w, "F10"], capture_output=True)
subprocess.run(["xdotool", "key", "F10"], capture_output=True)
t_armed = t_done = None
frames = draws = None
deadline = time.time() + timeout
while time.time() < deadline:
chunk, seen = _tail(log_path, seen)
for line in chunk.splitlines():
if "[UI-CAP] capture armed" in line and t_armed is None:
t_armed = time.time()
elif "[UI-CAP] done" in line and t_armed is not None:
t_done = time.time()
# "[UI-CAP] done: 1526 draws over 300 frames"
parts = line.replace(":", " ").split()
try:
draws = int(parts[parts.index("done") + 1])
frames = int(parts[parts.index("over") + 1])
except (ValueError, IndexError):
pass
if t_done:
break
time.sleep(0.02)
if not (t_armed and t_done and frames):
return None, seen
dt = t_done - t_armed
return {"frames": frames, "draws": draws, "seconds": dt, "fps": frames / dt}, seen
def run(limit, out_path, log_path, shots_dir):
os.makedirs(shots_dir, exist_ok=True)
n = T.W * T.H * 3
p = T.open_stream()
t0 = time.time()
seg = t0
frames = 0
seen = 0
ev = []
rates = {}
state = "boot"
last_label = None
last_gray = None
prev_mean = -1.0
fh = open(out_path, "w")
fh.write("#t\tglyph\tmean\tmotion\t" + "\t".join(T.REFS) + "\tlabel\n")
def mark(name, t=None):
t = time.time() - t0 if t is None else t
ev.append((name, t))
print(f"EVENT {name} t={t:.3f}", flush=True)
return t
while time.time() - t0 < limit and state != "done":
now = time.time()
# Restart only while still waiting; never across a measured interval.
if state == "boot" and now - seg > 30:
p.kill(); p = T.open_stream(); seg = now
fh.write(f"#restart\t{now - t0:.3f}\n")
buf = p.stdout.read(n)
if len(buf) < n:
p.kill(); p = T.open_stream(); seg = time.time(); continue
t = time.time() - t0
rgb = np.frombuffer(buf, np.uint8).reshape(T.H, T.W, 3)
g = T.gray_of(rgb)
gl = T.glyph(rgb)
sc = T.scores(g)
lb = T.label(sc)
surf = T.surface(g)
mn = float(surf.mean())
mo = float(np.abs(surf[::8, ::8] - last_gray).mean()) if last_gray is not None else -1.0
last_gray = surf[::8, ::8].copy()
frames += 1
fh.write(f"{t:.3f}\t{gl}\t{mn:.3f}\t{mo:.3f}\t"
+ "\t".join(f"{sc[k]:+.4f}" for k in T.REFS) + f"\t{lb}\n")
# Every label change and every entry/exit from pure black is a boot event.
blk = "black" if mn < 1.0 else lb
if blk != last_label:
mark(f"screen:{blk}", t)
last_label = blk
if blk in ("splash_pub", "splash_dev", "title_noplate", "menu"):
Image.fromarray(rgb).save(os.path.join(shots_dir, f"boot-{blk}.png"))
if state == "boot":
if lb in ("title_noplate", "title_plate") and 0 <= mo < 2.0 and gl >= 100:
mark("title_settled", t); state = "title"
elif state == "title":
if gl >= T.PLATE_GLYPH:
mark("plate", t); state = "plate_hold"; hold_from = t
elif state == "plate_hold":
if t - ev[-1][1] > 3.0:
fh.flush()
r, seen = arm_capture(log_path, seen)
rates["title"] = r
mark(f"rate_title={r and round(r['fps'], 3)}")
state = "press"
elif state == "press":
tp = T.tap("A") - t0
ev.append(("pressA", tp)); print(f"EVENT pressA t={tp:.3f}", flush=True)
state = "toMenu"
elif state == "toMenu":
if lb == "menu":
mark("menu", t); state = "menuSettle"; menu_at = t
elif state == "menuSettle":
if t - ev[-1][1] > 6.0:
fh.write(f"#ring_free_start\t{t:.3f}\n")
state = "ringFree"; ring_from = t
elif state == "ringFree":
# 12 s of ring with NOTHING else running -- the outside-capture control
if t - ring_from > 12.0:
fh.write(f"#ring_free_end\t{t:.3f}\n")
fh.flush()
r, seen = arm_capture(log_path, seen)
rates["menu"] = r
mark(f"rate_menu={r and round(r['fps'], 3)}")
fh.write(f"#ring_capture_end\t{time.time()-t0:.3f}\n")
state = "ringAfter"; after_from = time.time() - t0
elif state == "ringAfter":
if t - after_from > 12.0:
state = "done"
p.kill()
dt = time.time() - t0
fh.write(f"#summary\tframes={frames}\telapsed={dt:.1f}\tfps={frames/dt:.2f}\trequested={RATE}\n")
for k, r in rates.items():
if r:
fh.write(f"#rate\t{k}\tframes={r['frames']}\tdraws={r['draws']}"
f"\tseconds={r['seconds']:.3f}\tfps={r['fps']:.4f}\n")
else:
fh.write(f"#rate\t{k}\tFAILED\n")
for name, t in ev:
fh.write(f"#event\t{name}\t{t:.3f}\n")
fh.close()
print(f"\n{frames} frames in {dt:.1f}s = {frames/dt:.2f} fps (requested {RATE})")
for k, r in rates.items():
print(f" presentation rate on {k}: "
+ (f"{r['fps']:.4f} fps ({r['frames']} frames in {r['seconds']:.3f} s, "
f"{r['draws']} draws)" if r else "FAILED"))
return 0
if __name__ == "__main__":
if len(sys.argv) > 1 and sys.argv[1] == "--control":
sys.exit(control())
if len(sys.argv) > 4 and sys.argv[1] == "--run":
sys.exit(run(float(sys.argv[2]), sys.argv[3], sys.argv[4],
sys.argv[5] if len(sys.argv) > 5 else "/sylph-home/re/shots/boot-timeline"))
print(__doc__)
sys.exit(2)