a99f517ec2
tools/run-backend-differential-gate.ps1 is V7's instrument, plus the route it
runs. It executes the SAME connected route twice - ACDREAM_RENDER_BACKEND=gl
then =vulkan - pairs the screenshots by name, compares each pair with
compare-screenshots at the plan's pinned tolerance 2 / fraction 0.001, and
prints one verdict table with a JSON report beside it.
ACDREAM_MSAA_SAMPLES=0 IS FORCED ON BOTH LAUNCHES. Plan section 5.5.16 measured
what leaving it on costs - 8.83% of the frame at 4x, essentially all of it
hugging foliage and silhouette edges, which is two drivers' sample patterns and
not a renderer divergence. Forcing it on one side only would be worse than
either. ACDREAM_DAY_GROUP is pinned on both, and the route additionally pins the
client-only time-of-day override so the Dereth clock cannot drift between two
launches minutes apart.
The desktop witness and its three guards are lifted from the repeat-run gate
rather than reinvented, because that gate learned them the hard way: a blank
frame reads back as a valid PNG (section 5.5.2), so a second instrument that
shares nothing with the renderer decides whether a run is worth comparing at
all; a locked screen or an overlapping window is caught by the grab's size; and
stray input into a foregrounded window moves the camera off the pinned stop, so
the client log is scanned for camera-affecting actions and the run is aborted
rather than reported. Both runs close through WM_CLOSE - a hard kill leaves ACE
holding the session about three minutes and the second launch is the one that
pays for it.
connected-backend-differential.route.txt is written for two-launch determinism,
which is stricter than either existing route needed to be: identity quaternions
on every teleloc so heading is pinned, no movement between arrival and capture,
and no checkpoint verb (the ownership ledger has its own gate). Three stops -
Holtburg town, an interior EnvCell at the Facility Hub, Aerlinthe's island and
water edge. The interior stop is the durable fix section 5.1 names for the
campaign's oldest coverage gap: no connected route has ever visited one, which
is why EnvCellRenderer had no automated coverage on either backend.
ONE SMOKE PAIR was run, on the first stop, and it is reported honestly rather
than tuned. GL versus Vulkan at Holtburg with MSAA off: 170,697 of 921,600
pixels differ, 18.52%, maximum channel delta 255. Attributed with a difference
map:
* 89% of it is in the top 240 rows. Masking the offline gate's top 280 rows
takes the pair to 11,480 of 563,200, 2.04%. That band is the scrolling
cloud sheet, which advances with WALL time and not with the pinned Dereth
clock, plus the treeline behind it.
* Masking the animated portal beside the character as well takes it to 6,190
of 529,450, 1.17%, mean channel delta 0.49. A portal's scrolling texture is
phase, like an emitter's age.
* What remains is thin outlines on silhouette edges throughout the frame,
plus the vitals readouts, whose stamina and mana genuinely advanced between
the two logins.
So the distance to V7's 0.001 is about 12x with the two known phase populations
removed, and V7 owns closing it. -MaskTopPixels exists for that conversation and
DEFAULTS TO 0: the gate is a strict identity check on the whole frame unless
someone deliberately asks otherwise, per section 7.1 rule 2.
Both files are ASCII with CRLF, matching the other tools. That is not cosmetic:
PowerShell 5.1 reads a BOM-less .ps1 as ANSI, so a UTF-8 em dash inside a
double-quoted string is a parse error, which is how the first version failed.
Gates. Release build green; App tests 4,132 / 3 skips. No product code changed,
so commit 1's pixel, connected and validation gates stand unchanged - and the
smoke run is this commit's own evidence, since it drove a complete connected
Vulkan session end to end with a graceful exit.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
77 lines
3.9 KiB
Text
77 lines
3.9 KiB
Text
# Campaign V slice V7: the connected route the GL-versus-Vulkan differential
|
|
# runs TWICE, once per backend, pairing the PNGs by name.
|
|
#
|
|
# Two separate launches have to agree pixel-for-pixel, which is a much stricter
|
|
# requirement than either connected route was written for. Everything below is
|
|
# in service of that:
|
|
#
|
|
# * Every teleloc carries the identity quaternion, so the heading at each stop
|
|
# is the same on both runs rather than whatever the previous heading was.
|
|
# * The world clock is pinned by the client-only time-of-day override
|
|
# (AcdreamCycleTimeOfDay cycles live -> 0.00 -> 0.25 -> 0.50, so three
|
|
# presses land on noon). Without it the sky gradient, the sun angle and
|
|
# every lit surface drift with wall time between the two launches, which is
|
|
# the same effect that makes the offline gate mask its top 280 rows.
|
|
# * Nothing moves the character. The only thing between arrival and capture is
|
|
# a settle window, because a timed turn cannot be relied on to stop at the
|
|
# same angle twice.
|
|
# * There is no `checkpoint` verb here. The differential's verdict is the
|
|
# paired PNG, and the ownership ledger has its own gate
|
|
# (run-connected-world-lifecycle-gate.ps1) that this route does not duplicate.
|
|
#
|
|
# MSAA is forced OFF by the gate script rather than here, on BOTH launches.
|
|
# Sample positions are not specified across implementations and cost 8.8% of the
|
|
# frame at 4x -- measured, plan section 5.5.16 -- so a differential run with MSAA
|
|
# on measures the two drivers' sample patterns and nothing else.
|
|
#
|
|
# Residual divergence this route cannot remove, and which is a finding rather
|
|
# than a failure: PHASE. Two launches cannot agree on an emitter's age, a
|
|
# creature's animation cursor, a portal's scrolling texture, or the cloud
|
|
# sheet's position -- the cloud sheet in particular advances with WALL time and
|
|
# not with the Dereth clock the override above pins, which is why the offline
|
|
# pixel gate masks its own top 280 rows. The V6m smoke run measured the shape
|
|
# of it at stop 1 with MSAA off: 18.52% of the frame differed, 89% of that in
|
|
# the top 240 rows (cloud sheet, treeline, the portal), and rows 300-720 --
|
|
# terrain, blending, roads, buildings, character, entire retained UI --
|
|
# contributed about 9,500 pixels between them, almost all of it thin outlines
|
|
# on silhouette edges. That is why every stop below is chosen for its static
|
|
# content and why the settle windows are long.
|
|
|
|
wait world-ready 90000
|
|
wait world-visible 30000
|
|
sleep 8000
|
|
input press AcdreamCycleTimeOfDay
|
|
input press AcdreamCycleTimeOfDay
|
|
input press AcdreamCycleTimeOfDay
|
|
sleep 5000
|
|
|
|
# 1. Holtburg: the dense outdoor town. Terrain, terrain blending, the road
|
|
# overlay, static world meshes, scenery and the whole retained UI in one
|
|
# frame. This is the SMOKE stop -- the one pair -Smoke compares on its own.
|
|
command /teleloc 0xA9B40019 84.0 7.1 94.005 1 0 0 0
|
|
wait materialized 1 90000
|
|
wait world-visible 30000
|
|
sleep 25000
|
|
screenshot holtburg_town 15000
|
|
|
|
# 2. Facility Hub: an interior EnvCell, and the durable fix for the campaign's
|
|
# oldest coverage gap. Plan section 5.1 records that the offline gate sees no
|
|
# EnvCell at all and that EnvCellRenderer's Vulkan arm is proven by one
|
|
# eyes-on frame; this stop is the first automated instrument that looks at
|
|
# one. Cell 0x164 is >= 0x100, which is what makes it indoor: walls, floor,
|
|
# per-cell lighting and the doorway clip region rather than terrain.
|
|
command /teleloc 0x8A020164 70.35 -40.66 -5.9 1 0 0 0
|
|
wait materialized 2 90000
|
|
wait world-visible 30000
|
|
sleep 25000
|
|
screenshot facility_hub_interior 15000
|
|
|
|
# 3. Aerlinthe: dense island geometry, the water edge and a long view. The
|
|
# stop that exercises fog and the far end of the depth range, where a
|
|
# precision divergence between the two backends would show first.
|
|
command /teleloc 0x09040008 11.4 188.6 87.705 1 0 0 0
|
|
wait materialized 3 90000
|
|
wait world-visible 30000
|
|
sleep 25000
|
|
screenshot aerlinthe_island 15000
|
|
sleep 3000
|