night13: #148 tracker correction script
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_018SgmXGNMXavXiafKXf9MDC
This commit is contained in:
Joe DiPrima
co-authored by
Claude Opus 5
Claude Opus 5
parent
f36f0136c8
commit
bb6605d53b
@@ -0,0 +1,84 @@
|
||||
"""Night 13: correct #148's framing with the measured findings. ASCII only."""
|
||||
import sys
|
||||
sys.path.insert(0, r"C:\git\bt411\scratchpad\night7")
|
||||
import gitea
|
||||
|
||||
BODY = """**Investigated, narrowed a long way, NOT fixed. Correcting this ticket's own framing first.**
|
||||
|
||||
## The title/premise was WRONG -- the cadence is authentic
|
||||
|
||||
I filed this as "only 13 update records across a 5-minute run". That is not a defect. Reading the
|
||||
payloads with line numbers:
|
||||
|
||||
```
|
||||
205 RECORD atUpd=0.0436769 rate=0.305433
|
||||
502 RECORD atUpd=2.3558 rate=-0.305433
|
||||
1150 RECORD atUpd=-2.39277 rate=0.305433
|
||||
1698 RECORD atUpd=2.38543 rate=-0.305431
|
||||
2044 RECORD atUpd=-2.39124 rate=0.305427
|
||||
...
|
||||
```
|
||||
|
||||
Those `atUpd` values are the sweep's EXTREMES and `rate` flips sign at each one. **The master sends
|
||||
a torso record on RATE CHANGE**, and the peer dead-reckons `atUpd + rate * elapsed` in between
|
||||
(`ComputeTargetTwist`). 12 records for 12 direction reversals is exactly right -- this is
|
||||
extrapolation-based replication working as designed, not a starved channel.
|
||||
|
||||
Also ruled out: the `ComputeTargetTwist` clamp. The copy's limits load correctly
|
||||
(`limL=2.44346 limR=-2.44346 enab=1 copy=1`), so `Min/Max` is not pinning `targetTwist` to zero.
|
||||
And the extrapolator itself is exact -- the copy reports `cur == target` on every sample.
|
||||
|
||||
## The real defect
|
||||
|
||||
**The peer's copy torso Performance does not run at all until log line ~1006, while its first
|
||||
record arrived at line 205.** ~800 lines of correctly-replicated twist data integrated by nobody.
|
||||
|
||||
This is now directly readable because `[torso-copy]` logs on call #0 (`s_cl++ % 120`), so its first
|
||||
line IS the first `TorsoCopySimulation` call. (That also means my earlier "first copy currentTwist
|
||||
!= 0 at line 1016" was a SAMPLING artifact -- first *sample*, not first non-zero.)
|
||||
|
||||
The first tick coincides with the replicant's MODEL bring-up, not with record arrival:
|
||||
|
||||
```
|
||||
[loadclips] end: fScale=0.8 ... hasGimpClips=1
|
||||
[clipfix] mech 05769358 -> EXTERIOR (lean)
|
||||
[torso] PushTwist COPY node=057A3C68 type=1 twist=-1.52319
|
||||
```
|
||||
|
||||
So the gate is **above the subsystem level**, somewhere in replicant model/clip initialisation.
|
||||
Not yet found.
|
||||
|
||||
## Tried and rejected
|
||||
|
||||
`Entity::Perform` (ENTITY.cpp:733-793, real engine source) picks the executable predicate by
|
||||
instance -- `IsNonReplicantExecutable()` for masters, `IsReplicantExecutable()` for replicants --
|
||||
and they differ exactly on `|| lastUpdate >= lastPerformance`, which is what makes an
|
||||
`ExecuteOnUpdate` subsystem run when a record arrives. Mech's reconstructed tick loop used the
|
||||
NonReplicant predicate for every mech, dropping the branch.
|
||||
|
||||
That is a genuine fidelity gap and it is now **fixed** (`f36f013`) -- but it does **not** move this
|
||||
bug (first copy tick 1014 -> 1006, noise). The torso's own flag was not the gate.
|
||||
|
||||
## Consequence, and what it is NOT
|
||||
|
||||
This is the tail of #141: the peer's missiles launch along the body facing until its torso starts
|
||||
ticking. #141's own fix is complete and verified separately -- every zero-`twistDelta` peer launch
|
||||
now reads `liveTwist=0` on the same log line, and the first launch with `liveTwist=-1.84061` reads
|
||||
`twistDelta=-1.83813`. The launch frame tracks the twist perfectly; there is simply no twist to
|
||||
track until the copy torso wakes up.
|
||||
|
||||
Likely also relevant to **#37** (MadCat torso BACKWARDS) and **#70** (twist stops after respawn) --
|
||||
a peer torso that does not tick would sit at its bind pose, and a respawn re-runs model bring-up.
|
||||
Re-test both once this is found.
|
||||
|
||||
## Diagnostics now in place
|
||||
|
||||
`BT_TORSO_LOG`: `[torso-rec-rx]` (receive + raw payload), `[torso-copy]` (cur/target/atUpd/rate plus
|
||||
`limL/limR/enab`), `[torso] PushTwist master|COPY` (per instance-kind).
|
||||
`BT_PROJ_LOG`: `[launchframe]` now carries the shooter's live torso twist, so `twistDelta` and its
|
||||
driver appear on the same line."""
|
||||
|
||||
gitea.call("/issues/148", method="PATCH", payload={
|
||||
"title": "Peer copy TORSO does not tick until replicant model bring-up -- ~800 lines of replicated twist integrated by nobody"})
|
||||
gitea.comment(148, BODY)
|
||||
print("updated #148")
|
||||
Reference in New Issue
Block a user