Scenarios time out at the first map wait because the teleport right after a save loads doesn't take #112

Open
opened 2026-11-20 11:09:01 +00:00 by klimovich008 · 1 comment
Member

🤖 Many gameplay cells end with a step ran out of time: timeout: not met in time. That is the game bridge's reply when a wait for event = "map" reaches its timeout. The first step of the scenarios involved is a teleport straight after start = "save:kokiri-start", usually to HyruleField_Child, followed by that wait with timeout_s = 60.

What the reports show:

Likely cause (from Longshot's side): the game's teleport (oot_teleport, host/hos_main.c) refuses or drops a request while an event plays or a map loads. It also takes only one request every 8 s. Just after a save loads, the save's own opening event can still be running. A refused teleport is retried, but one that was accepted while an event was about to start seems to be lost silently, and the wait then runs out.

Workaround we're trying on our PRs: wait_s = 10 before the teleport and timeout_s = 120 on the map wait (rounds started 2026-11-20 04:45Z on #1776, #1282, #1574). If those pass, the default plan's hyrule-field could do the same.

Asked:

  1. In the agent, wait until the bridge reports state play with no event playing before the first teleport.
  2. When the map wait times out, send the teleport once more before giving up.
  3. Put the bridge's last cheats: teleport … log lines in the report, so the cause shows without the full log.
🤖 Many gameplay cells end with `a step ran out of time: timeout: not met in time`. That is the game bridge's reply when a `wait` for `event = "map"` reaches its timeout. The first step of the scenarios involved is a `teleport` straight after `start = "save:kokiri-start"`, usually to `HyruleField_Child`, followed by that wait with `timeout_s = 60`. **What the reports show:** - **The teleport didn't happen.** The last-frame picture of the timed-out runs still shows Link in Kokiri Forest. The same happens on main's side, so it isn't the PRs' code. Example: [rp_c92e36a77d344f4c](https://beanstalk.magicbean.dev/reports/rp_c92e36a77d344f4c), RX 7900 XTX on RADV, both cells and both sides. - **Seen on several authors' PRs, Linux and Windows:** - #1488 [rp_0ff356f59680d4d5](https://beanstalk.magicbean.dev/reports/rp_0ff356f59680d4d5) - #1755 [rp_ef5a17016f8f96cf](https://beanstalk.magicbean.dev/reports/rp_ef5a17016f8f96cf), [rp_fbf060c36cfa9fe7](https://beanstalk.magicbean.dev/reports/rp_fbf060c36cfa9fe7) - #1574 [rp_c92e36a77d344f4c](https://beanstalk.magicbean.dev/reports/rp_c92e36a77d344f4c), [rp_6d02c4b07b528a43](https://beanstalk.magicbean.dev/reports/rp_6d02c4b07b528a43), [rp_405be0c8dd6ca57d](https://beanstalk.magicbean.dev/reports/rp_405be0c8dd6ca57d), [rp_e63417936a4c5a35](https://beanstalk.magicbean.dev/reports/rp_e63417936a4c5a35) - #1282 [rp_06680101b0bbd096](https://beanstalk.magicbean.dev/reports/rp_06680101b0bbd096), [rp_f64e1af9b77d2050](https://beanstalk.magicbean.dev/reports/rp_f64e1af9b77d2050) - **The same step can also be refused outright.** On #1702 the teleport was refused with "an event is playing (a cutscene, a map change's)": [rp_01c20b8226d7940d](https://beanstalk.magicbean.dev/reports/rp_01c20b8226d7940d), [rp_2eacdc0f799b5a3f](https://beanstalk.magicbean.dev/reports/rp_2eacdc0f799b5a3f), [rp_4af36d408d9304cf](https://beanstalk.magicbean.dev/reports/rp_4af36d408d9304cf). **Likely cause (from Longshot's side):** the game's teleport (`oot_teleport`, `host/hos_main.c`) refuses or drops a request while an event plays or a map loads. It also takes only one request every 8 s. Just after a save loads, the save's own opening event can still be running. A refused teleport is retried, but one that was accepted while an event was about to start seems to be lost silently, and the wait then runs out. **Workaround we're trying on our PRs:** `wait_s = 10` before the teleport and `timeout_s = 120` on the map wait (rounds started 2026-11-20 04:45Z on #1776, #1282, #1574). If those pass, the default plan's `hyrule-field` could do the same. **Asked:** 1. In the agent, wait until the bridge reports state `play` with no event playing before the first teleport. 2. When the map wait times out, send the teleport once more before giving up. 3. Put the bridge's last `cheats: teleport …` log lines in the report, so the cause shows without the full log.
Member

Possibly related, for context:

  • PR MagicBeanDev/Longshot#1046 WIP: fix(cheats): a debug teleport the game dropped is sent once more, so the control bridge's teleport arrives — Open fix for game-dropped teleports after save load; catches dropped requests, though event-refused requests need agent handling.
  • issue #10 agent: loading a save gives up after one 30 s press while the game is still building its pipelines — Earlier agent issue about startup timing after loading a save; concerned controller-input timeout, not map teleport.
Possibly related, for context: - PR MagicBeanDev/Longshot#1046 WIP: fix(cheats): a debug teleport the game dropped is sent once more, so the control bridge's teleport arrives — Open fix for game-dropped teleports after save load; catches dropped requests, though event-refused requests need agent handling. - issue #10 agent: loading a save gives up after one 30 s press while the game is still building its pipelines — Earlier agent issue about startup timing after loading a save; concerned controller-input timeout, not map teleport.
Sign in to join this conversation.
No milestone
No project
No assignees
2 participants
Notifications
Due date
The due date is invalid or out of range. Please use the format "yyyy-mm-dd".

No due date set.

Dependencies

No dependencies set

Reference
MagicBeanDev/Beanstalk#112
No description provided.