Skip to content

FRLG AutoStory: merge stage 8 JSON checkpoints into failure capture - #344

Merged
dayfine merged 160 commits into
frlg/autostory-failure-capturefrom
frlg/autostory-failure-capture-json
Nov 10, 2026
Merged

dayfine merged 160 commits into
frlg/autostory-failure-capturefrom
frlg/autostory-failure-capture-json

Conversation

@dayfine

@dayfine dayfine commented Nov 10, 2026

Copy link
Copy Markdown
Owner

Depends on #310

Merges stage 8 (#293, now on the #312 JSON step format) into the failure-capture branch. Stage 8 had dropped the segment policy's stop-at-first-failure and in-place recoveries, so the conflict in run_row resolves to stage 8's loop with the failed hook called right after a failed attempt (before the game-not-running check), and the stats call becomes record_story_counts. The capture tests' char* constants are now string_view for code health. Built locally (exit 0) and ran the unit tests with the isolated unit-tests profile: the FailureCapture and StorySegments suites pass (14), and the remaining failures are the image tests that need the CommandLineTests checkout. Code health is 0 against frlg/stage8-rocktunnel; against this base it reports 8, all in stage 8 files this branch did not have yet.

🤖 Generated with Claude Code

https://claude.ai/code/session_013VUubhHeYSMGSSDXZkKZss

dayfine and others added 30 commits November 9, 2026 10:56
…ock Tunnel

Console B, 2026-11-09: B1F row 37 met a wild battle 10 to 14 tiles into
every try and the walker gave up after 9 stops. Red: both fail.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_013VUubhHeYSMGSSDXZkKZss
…est tiles

Rock Tunnel B1F row 37 (console B, 2026-11-09): a wild battle 10 to 14
tiles into every try of "west 16". The only single hold that found the
player went east to the wall (45, 37), behind the leg's start, and the
way back walked all 32 tiles again, until the route gave up. The
reanchor search now also tries each first hold with the fewest holds
after it, and plan_reanchor() takes the way whose holds and way back
walk the fewest tiles together: there, south to row 38 and west to the
wall (1, 38), then 13 tiles back.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_013VUubhHeYSMGSSDXZkKZss
The walk waited for a warp's fade only once the hold onto the warp had
ended, so a fade that was over by then went unseen and the walk stopped on
"no fade in and out after walking onto the warp" though the player had
warped (00.27, console B, 2026-11-09 01:41, the warp to CeruleanCity: the
watch started about 1.3 s into the hold). The walk and the recovery walk
now show the hold's frames to the same BlackScreenOverWatcher
(Pokemon::WatchAlong) and wait on with it only if it has not fired. The
watcher and its thresholds are unchanged, so a dim textured floor is still
no fade.

The overworld sim's walker watches for the fade from where the console
does (SimWalkerRules::watch_fade_along_hold), and fails on a warp whose fade
it missed, as the console does. OverworldWarpFade walks out of Viridian
Mart with a 200 ms fade that is over before the hold ends: seen when
watched from the start of the hold, missed from its end.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_013VUubhHeYSMGSSDXZkKZss
… player

Rock Tunnel 00.28 with a wild battle every 7 to 25 steps (sim) still lost
the open legs' progress after the first fix:
- Walk on: after an open leg stops, if the route's next holds bring every
  tile the player may be on to one tile (and none of those ways crosses a
  trainer's sight off the planned way), the walker walks on the rest of
  the leg from the middle candidate instead of finding the player at
  walls first. The player still faces the leg's way, so no turn tap.
  A stop on the walked-on leg counts from every tile it may start on.
- A leg onto a warp stopped by a wild battle, where no trainer may see the
  player, holds on to the warp as off the grass: below Rock Tunnel 1F's
  ladder (20, 13) the holds that found the player went down a column
  Hiker Lucas looks over.
- plan_reanchor() weighs open-leg tiles 4x over held ones, and puts ways
  whose holds cross a trainer's sight last.
- The sim looks for trainers before rolling for a wild battle, as the
  game does (pokefirered ProcessPlayerFieldInput()).

OverworldReanchor rules and two OverworldSim rule tests follow the new
choices.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_013VUubhHeYSMGSSDXZkKZss
… walk on through open legs

The 00.28 sim with a wild battle every 7 to 25 steps now passes at all 12
seeds (before these changes: 0 of 12).
- A step into a trainer's line that sets off a wild battle took its
  battle screen for the trainer's, and the battle step after it found no
  message. The step now watches 5 s for a wild battle, runs from one seen
  with no message before it, and steps on (program and sim). The game
  looks for trainers before it rolls for a wild battle, so the trainer
  didn't see the player on that step.
- Walking on after a stop may go through open legs of the plan before the
  hold that finds the player (Rock Tunnel 1F row 31: west, then south 1,
  then the hold west to (12, 32)); a stop on any of them counts from every
  tile it may start on.
- A hold to a wall where wild battles start, stopped right away after
  another interrupt (Cooper, then a wild battle), is walked again.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_013VUubhHeYSMGSSDXZkKZss
SimScript::encounter_rates rolls each step onto a wild tile as pokefirered
TryStandardWildEncounter() does (8 - rate / 10 cooldown steps at 5%, then
rate * 16 / 1600 plus a buff that grows with each roll lost; a battle
clears both), seeded per run. Rock Tunnel 1F/B1F use rate 7, Route 10 21
(wild_encounters.json). SimState counts wild battles and their time.

New test: 00.28 at those odds from 30 seeds must all pass, and logs the
runs' numbers. At the same seeds, before this PR's walker fixes 13/30
passed; now 30/30. Over passing runs, before -> after: time 1037 -> 912 s
mean (p90 1190 -> 1004), tiles 535 -> 406 (p90 679 -> 437), tiles walked
again 158 -> 3 (p90 302 -> 34), wild battles 22.6 -> 17.7, time lost per
battle 4.5 -> 0.3 s (p90 6.6 -> 2.0).

With trainers checked before wild battles in the sim, Cooper is always
fought: the 7-to-25-step test no longer allows him unfought.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_013VUubhHeYSMGSSDXZkKZss
Console B run 2 of 00.28 lost the one-tile step north onto B1F row 34
after a wild battle on row 38; the walk east "to the wall (17, 34)" then
ran along row 35 to the rock (20, 35) and martha_line found no trainer.
below_martha_line now ends at (16, 35) and martha_line walks up column 16
(or 17, a tile short) into her line; below_allen_line starts with the
walk east to the wall (17, 33). Sim test replays the console's battle and
the short step.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_013VUubhHeYSMGSSDXZkKZss
…f missed

Console B lost the one-tile step north onto Rock Tunnel B1F row 34 in
00.28 run 2: the game missed the 40 ms turn tap, the leg's short hold then
only turned the player (pokefirered PlayerTurnInPlace(), 8 frames), and
the leg was taken as walked. OverworldWalkSteps::turn_to() taps, reads the
facing (read_player_facing()) and taps again while it shows another way,
up to 3 taps; an unreadable facing is taken as turned. Both walkers,
step_once() and face_wall() turn with it.

Sim: SimScript drops turn taps by number or at a rate; sim_turn_to()
mirrors turn_to(), switched by SimWalkerRules::confirm_turn. New tests:
a dropped tap before a one-tile leg (red without the rule), and 00.28 at
the game's odds with 1 in 10 turn taps dropped, 30 seeds.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_013VUubhHeYSMGSSDXZkKZss
An UNSETTLED look (a fade with no battle yet) failed any step as
interrupt_unsettled after three looks, including a BATTLE step whose
battle was already opening (BDSP 00.02's Starly battle right after the
starter pick, lab-B run 3). A step that starts on WILD_ENCOUNTER or
TRAINER_SIGHT now starts on UNSETTLED too, and gets it on the first look.
Walk steps keep the check.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_013VUubhHeYSMGSSDXZkKZss
…thod

Code health: OverworldSim kept at 7 public methods (sim_turn_to() reads the
state) and OverworldSim_Tests under 1000 lines (the turn tap cases moved to
OverworldWalk_Tests, as a table).

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_013VUubhHeYSMGSSDXZkKZss
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_013VUubhHeYSMGSSDXZkKZss
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_013VUubhHeYSMGSSDXZkKZss
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_013VUubhHeYSMGSSDXZkKZss
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_013VUubhHeYSMGSSDXZkKZss
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_013VUubhHeYSMGSSDXZkKZss
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_013VUubhHeYSMGSSDXZkKZss
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_013VUubhHeYSMGSSDXZkKZss
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_013VUubhHeYSMGSSDXZkKZss
A lab deployment without PokemonFRLG/PlayerSprites/red_normal.png failed
00.28 with a FileException; no test checked the files the readers load.
autostory_resource_paths() lists the files AutoStory and its readers and
detectors load, and the test fails naming each one missing.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_013VUubhHeYSMGSSDXZkKZss
Move the walker's pace, tap and release times into
PokemonFRLG_MovementMechanics.h with the tap rules measured on console B
(2026-11-09, Route 10 Center 1F): a tap from a standstill only turns a
player facing another way and steps once for one facing its way, a step
tap within 620 ms of the last one's start is ignored, and taps with no
release between them are one hold. OverworldSim now moves its player by
these rules; the random turn-tap drop is gone (dropped_turn_taps stays as
fault injection for the turn check).

Red: sim_step_once() still taps at the turn tap's pace (340 ms), so three
step_once() lose a step under the lockout.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_013VUubhHeYSMGSSDXZkKZss
step_once() tapped 120 ms and released 500 ms, 620 ms from tap to tap:
inside the band where the console's lockout ends (a step tap 566 ms after
the last is ignored, 666 ms works). It now taps STEP_TAP (66 ms) and
releases STEP_TAP_RELEASE (700 ms), 766 ms apart, and sim_step_once()
does the same.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_013VUubhHeYSMGSSDXZkKZss
…s over every row

Each row with a start runs on the overworld sim clean, then with the
classes of failure console B showed injected everywhere they can happen:
a scene that moved the player ending on each open tile next to the
table's pose; wild battles at pokefirered's land encounter rates on every
map the story walks, 30 seeds, logging passed, time, tiles walked again
and time lost per battle; and every warp's fade 1.5 s late. Steps the sim
does not play put the player on their pose_after; scene triggers and
leaders' battles it has no script for pass where the clean run let them.
00.06 is known failing: two seeds give up on Route 1 after 9 stops.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_013VUubhHeYSMGSSDXZkKZss
… 35)

Rock Tunnel without Flash is black but for a lit circle on the player.
The battle text box and black screen checks read it as a battle on its
way, so the check before each step ran a 15 s battle loop, and an open
leg's stop was placed 0 to 3 tiles in when the battle came at its end
(console B, 2026-11-09 13:35). The walker then "found" the player at the
rock (20, 35) while they stood against the wall (25, 35).

- dark_cave_screen(): black corners, lit middle. The observer reports
  nothing on it, and the check before or after a step sees no battle.
- A stop on an open leg that started in a dark cave may be anywhere on
  the leg (whole_leg()), whatever the timing says.
- below_martha_line: west 16, north 2, west 2 to the corner (11, 35),
  east 5 to (16, 35). The corner finds the player from any of
  (11..14, 35) and (11..16, 35) after a battle.
- Sim: SimScript::dark_maps, with walker rules dark_cave_not_battle and
  whole_leg_in_dark; 00.28's game-odds runs walk Rock Tunnel dark.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_013VUubhHeYSMGSSDXZkKZss
The BDSP fault sweep (#226) injects a battle's opening at the start of
each step after a walk: 51 steps of 00.01-01.11 fail as
interrupt_unsettled after three looks, as the opening takes longer than
three looks to show its battle. These cases expect the looks to wait for
the battle and clear it before the walk.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_013VUubhHeYSMGSSDXZkKZss
A walk that ends in grass or in a trainer's sight leaves the next step to
start on the battle's opening: a white flash, or a fade still down, with
no battle yet. Three looks (0.9 s on BDSP) were too few for its battle
to show, and the step failed as interrupt_unsettled (BDSP fault sweep,
#226: 51 steps of 00.01-01.11). An unsettled screen that may be a
battle's opening (a white fade, a fade still down, or a fade after a
dialog box) is now looked at up to max_opening_looks (40) looks, until
its battle shows; the battle is then cleared by its handler (wild: run,
trainer: fight) and the step goes on. A door's fade, back up with no
box before it, keeps the three looks.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_013VUubhHeYSMGSSDXZkKZss
dayfine and others added 27 commits November 10, 2026 07:00
…ction/args steps

The FRLG stack stopped building after #312: the C++ rows set step.target,
legs, direction, fight_wild and may_lose directly, StoryChecks became
StoryCheckLevel, Pokemon_StorySteps.h moved to Pokemon_StoryStepExecutor.h
and the JSON codecs to Data/.

- Each checkpoint is StoryData/frlg-<id>.plan.json (pokesim-authored-
  checkpoint/1), read with read_authored_checkpoint() against FRLG's
  actions: the built-in ones plus ROUTE/LEGS "run". The parser fills in
  the fallbacks (fight_wild, may_lose, yes), so the files give only what
  differs. FRLG's execution object carries the START menu, the supplies,
  the lead after Brock and the notes that were C++ comments.
- The executor, the step-table, sim and route tests read typed args
  (story_arg<T>()); the sim's fake game dispatches by kind in run().
- Check levels: DEFAULT and above check the screen as THOROUGH did;
  FAST and SPEEDRUN trust counts as FAST did.
- AutoStory: StoryRunOptions with the FRLG languages, the check level and
  language into StoryEnv, record_story_counts().

The old C++ rows stay in PokemonFRLG_AutoStoryTableParity_Tests.cpp for
this commit only: each file loads to the same steps, args, poses,
supplies and START menu.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_013VUubhHeYSMGSSDXZkKZss
The previous commit's parity test passed: every StoryData file loads to
its old C++ row. The rows and the test go; StoryData/ is the table.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_013VUubhHeYSMGSSDXZkKZss
…s JSON

Stage 1 moved 00.01-00.10 into StoryData/ on #312's action/args steps.
Stage 2's own rows follow it:

- 00.01 (handoff sid_helper, the new intro why), 00.01b (live start, the
  GAME step close_trainer_card), 00.04 (to the ball, handoff starter_rng)
  and 00.04b are files; FRLG's execution object reads "handoff" and
  "live_start".
- close_trainer_card is an FRLG GAME action with no args; StoryExecutor's
  GAME action dispatches on step.action instead of overriding the old
  per-kind game().
- AutoStory: check level and language into StoryEnv, record_story_counts();
  the run-plan tests convert the start GridPose.

The C++ rows stay in PokemonFRLG_AutoStoryTableParity_Tests.cpp for this
commit only, and every file loads to its row.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_013VUubhHeYSMGSSDXZkKZss
The merge's parity test passed: every StoryData file, 00.01b and the
00.04/00.04b split included, loads to its C++ row. The rows and the test go.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_013VUubhHeYSMGSSDXZkKZss
FRLG AutoStory stage 2: 00.01b and the 00.04 split as JSON
… JSON

Stage 3 ported to #312's action/args steps, its checkpoints moved into
StoryData/ the way #330 and #331 moved stages 1 and 2.

- 00.11, 00.12, 00.12b, 00.12c and 00.13 are files, and 00.10 gains its
  walk to below Pewter's Center. Each carries "metadata": {"phase":
  "testing"}, which FRLG's reader checks and strips before the shared one.
- FRLG's GAME actions declare typed args in place of the variant strings:
  pc (do, slot, box, if_present), capture, fossil, level, mart, teach and
  grind (whose nurse is now its "nurse" arg), plus leave_gift_summary.
  Their specs are in story_action_specs(); the step table binds each to
  the GAME action with its args. VariantWords became GameArgs.
- The execution object's handoff takes gift_rng.

The old C++ rows stay in PokemonFRLG_AutoStoryTableParity_Tests.cpp for
this commit only: every file loads to the same steps, args, poses,
supplies and START menu.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_013VUubhHeYSMGSSDXZkKZss
The parity test passed on the merge before this commit: every StoryData
file through 00.13 loads to the C++ row it replaces.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_013VUubhHeYSMGSSDXZkKZss
FRLG AutoStory stage 3: 00.10-00.13 as JSON, GAME steps with typed args
… as JSON

Stage 4's checkpoints 00.14 to 00.16 (with 00.14b, 00.14c and 00.15b) as
StoryData files in the #312 step format, each marked "phase": "testing".
The STEP "trainer_line" action is a spec, Mart and TEACH steps load as
once per save, and stage 4's tests read steps through their args.
A parity test checks the files load to the C++ rows they replace.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_013VUubhHeYSMGSSDXZkKZss
The parity test passed in the merge before this commit: every story file
loads to the C++ row it replaces. The rows went with the merge's Table.cpp;
this removes the parity test that held them.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_013VUubhHeYSMGSSDXZkKZss
FRLG AutoStory: stage 4's checkpoints as JSON
…ws as JSON

Stage 5's checkpoints 00.17 to 00.20, and 00.14's Mega Kick tutor, as
StoryData files in the #312 step format, each marked "phase": "testing".
TEACH takes the tutor as its item; pacing and the walk back after a
trainer's battle read the GRIND's or capture's args. A parity test checks
the files load to the C++ rows they replace.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_013VUubhHeYSMGSSDXZkKZss
The parity test passed in the merge before this commit: every story file
loads to the C++ row it replaces. This removes it.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_013VUubhHeYSMGSSDXZkKZss
FRLG AutoStory: stage 5's checkpoints as JSON
…s JSON

Stage 6's checkpoints 00.21 to 00.25 as StoryData files in the #312 step
format, each marked "phase": "testing". The row-building TableSteps and
TableVermilionGym files go with the C++ rows. Cut, the trash cans, the
trainer wait and the GRIND check are action specs; a GAME action with
its own row keeps that row's console action. The row sims and stage 6's
tests read step args. A parity test checks the files load to the C++
rows they replace.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_013VUubhHeYSMGSSDXZkKZss
The parity test passed in the merge before this commit: every story file
loads to the C++ row it replaces. This removes it.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_013VUubhHeYSMGSSDXZkKZss
FRLG AutoStory: stage 6's checkpoints as JSON
…ows as JSON

Stage 7's checkpoints 00.26 and 00.27 as StoryData files in the #312 step
format, each marked "phase": "testing". The Celadon and TableSteps rows
go with the C++ rows. Stage 7's own JSON pilot (Data/JsonParser, the
Parcel capture and plan, its README and the Parcel sim test) is dropped
for the one reader in PokemonFRLG_AutoStoryTable; Data/JsonWriter and its
export tests stay. step_onto() is a hop: a STEP with a direction and a
declared end. A parity test checks the files load to the C++ rows they
replace.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_013VUubhHeYSMGSSDXZkKZss
The parity test passed on the merge before this commit: every StoryData
file through 00.27 loads to the C++ row it replaces. The row tests'
registrations move to PokemonFRLG_AutoStoryRows_Tests.h and Viridian's
SHOP step to PokemonFRLG_AutoStoryViridianMart, which keeps the story
table header and the events file within their size limits; the shop is
moved unchanged but for reading its args through story_arg.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_013VUubhHeYSMGSSDXZkKZss
FRLG AutoStory: stage 7's rows as JSON (00.26, 00.27)
…play reads #312 steps

Stage 7 now reads its rows from StoryData in the #312 step format and
has one checkpoint reader, without the legacy export reader or the
Parcel capture the replay read. The replay keeps the capture's exact
bytes as ReplayData/capture.json, which binding.json pins, and reads its
checkpoint in the step format: a route's target and a step's direction
(north where the export left the default out) become args. The executor
reads targets, directions and poses through story_arg and pose_of; its
report gives the engine's pose as a grid pose.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_013VUubhHeYSMGSSDXZkKZss
FRLG story replay: read the capture in #312's step format after stage 7's JSON
…rows as JSON

Stage 8's checkpoints 00.28 and 00.29 as StoryData files in the #312 step
format, each marked "phase": "testing"; the Lavender rows and TableSteps
go with the C++ rows. Stage 8's FaultSweep and Lavender tests read step
args; the row sims keep stage 8's sim_step_once/sim_turn_to and battle
end check with the args. The spin, hole and button tiles go into
Pokemon_MapTiles.h, where the parent moved TileKind. A parity test checks
the files load to the C++ rows they replace.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_013VUubhHeYSMGSSDXZkKZss
The parity test passed on the merge before this commit: every StoryData
file through 00.29 loads to the C++ row it replaces.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_013VUubhHeYSMGSSDXZkKZss
FRLG AutoStory: stage 8's rows as JSON (00.28, 00.29)
Brings the JSON checkpoints (#312 step format) under the failure
capture. Stage 8 dropped the segment policy's stop-at-first-failure and
in-place recoveries, so the failed hook is now told just before the
game-not-running check; the run's stats go through record_story_counts.
The capture tests' char* constants become string_view for code health.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_013VUubhHeYSMGSSDXZkKZss
@dayfine
dayfine marked this pull request as ready for review November 10, 2026 19:54
@dayfine
dayfine merged commit c4c83a4 into frlg/autostory-failure-capture Nov 10, 2026
6 of 11 checks passed
@dayfine
dayfine deleted the frlg/autostory-failure-capture-json branch November 13, 2026 03:26
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