Conversation
This was referenced Sep 8, 2026
Merged
# Conflicts: # waldo_commander/components/script_execution.py
# Conflicts: # waldo_commander/components/editor.py
start() cleared the tab log and log panel and overwrote last_outcome, last_run_source_digest and the run record before the restart preflight could refuse, so one refusal (a nudged arm, a changed TCP) replaced the "prior run context" the review dialog exists to show. fresh_state also treated a lost queue readback as a non-empty queue and a lost error readback as no fault; the queue must be read, and a controller whose action state is ERROR is a fault whether or not the ERROR query answered. Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
… files The program side (StepIO) connects to a listener the GUI opens per session (multiprocessing.connection: a named pipe on Windows, an owner-only Unix socket file elsewhere). Control states — paused, granted steps, the pause clock — travel down and wake a waiting program at once; command and skill events and the program's waiting state travel up and are queued as they arrive. Nothing is polled from disk: the three JSON files, the 256-event window, events_lost, the per-event rewrite and the Windows file-lock path are gone, and every program client (sync and async) shares the link. Tests drive the real link: a 2 000-event burst arrives complete and in order, a paused program blocks until a grant and is released by play, a pause behind a grant survives, cleanup frees a program to run unmanaged, recorded values only cross the owner's link, and the page-reload case keeps the link open. The stop/blend and pause-budget cases run against the fake-serial controller as before. ty skipped: pre-existing DryRunRobotClient.skill_capabilities diagnostic. Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
A taught setup is programs/setups/<name>.py holding one SetupSnapshot literal, and a library world is programs/worlds/<name>.py holding one ShapeWorld literal, written the way the recorder writes programs so they are copied, diffed and versioned with the programs that use them and can be imported directly. Both stores compile the module from source rather than through the import system, whose bytecode cache is keyed on mtime and size and read back stale values on a same-second re-teach. JSON files from earlier releases, including the old ~/.waldo-commander/setups store, are converted on first use and removed. Live programs get their own directory on sys.path so `from setups.bench import setup` works. SKIP=ty: pre-existing DryRunRobotClient.skill_capabilities diagnostic. Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
…rt shows A function decorated as both a skill and a restart entry binds the module name to a Skill, which is not a function and so is not a restart entry: the dialog offered an entry that failed at launch (or a program that failed at import, with the decorators the other way round). It is refused at discovery, naming the conflict. The restart event recorded its entry under a key the export does not carry, so a shared export of a restart had no entry name in it, and none of the launch state's fields were labelled -- the booleans exported as field_7 and field_13 with no way to tell the fault flag from the empty queue. The dialog's failure label read `error or type(error).__name__`, and an exception is always truthy, so the likeliest failure here -- an argless TimeoutError from the state read -- printed nothing after the colon. SKIP=ty: pre-existing DryRunRobotClient.skill_capabilities diagnostic. Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
The event link is a pipe now, so the conflicts resolve toward it: the queue-based poll_events keeps its no-loss contract and the run-record assertions that described the old 256-event file window stay deleted, along with the hand-kept STEPPABLE_METHODS list the command table replaced. Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
# Conflicts: # waldo_commander/components/script_execution.py # waldo_commander/services/stepping_client.py
# Conflicts: # tests/test_run_records.py
Industrial controllers let the operator move the program pointer to any routine and declare only the pre-start hooks, so the marker is gone: discovery lists each top-level function callable without arguments, skipping private names, generators, skills and decorated functions, and the launch re-checks the signature in the fresh process. Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01N8zt66KjkzgZLtQuSfJd9r
Jepson2k
changed the title
A top-level before_restart() is the program's restart event routine: it runs in the fresh process before whichever function the operator picked, never on an ordinary start, and is never offered as a place to start from. A hook that cannot run without arguments is refused at discovery. Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01N8zt66KjkzgZLtQuSfJd9r
This branch has not been deployed
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
2 participants
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
After interruption, the editor shows prior run context and lets the builder pick any top-level function of the program that can be called without arguments, the way an industrial controller lets the operator move the program pointer to a routine. Private names, generators, skills and decorated functions are not offered, and nothing in the program runs to build the list. A top-level
before_restart()is the program's restart hook, the counterpart of a controller's restart event routine: it runs in the fresh process before the chosen function on every supervised restart, never on ordinary Start, and is never offered as a place to start. Before launch it requires fresh, advancing controller state, an empty stationary queue, valid referencing, and confirmation of the physical setup. A final check detects source, session, world, tool, TCP, or pose changes since that review.The selected function runs in a fresh Python process, which re-checks that it is still a plain function callable without arguments. Discovery does not import or execute the program; entry execution accepts only the documented restricted module initialization. Ordinary programs still start from the beginning. Stopping during launch preflight cancels launch and stops the backend. PAR6's normal idle gravity-compensation/freedrive flag does not itself prohibit restart.
Validation: managed stop/restart/pause/run-record workflows, a populated browser selection/confirmation workflow, and native PAR6 execution with six consecutive restarts passed. Lint and type checks passed. Stacked on run records. The PAR6 dependency still has the three native settling/clearance failures documented in PAR6 #78.
Companion PRs: PAROL6, PAR6. The waldoctl marker PR (#36) was closed without merging; no waldoctl change is needed.
Supervised restart uses a compact dialog with optional details and is accessible through Program → More.
Screenshots — revised interface, PAROL6 and PAR6
Captured in Chromium on September 10 from the combined feature stack; shared panels may include later PRs. PAROL6 uses fake serial; PAR6 uses
par6d --sim.PAROL6 — stopped run, selected restart entry, and current controller readiness.
PAR6 — controller state refreshed after Stop, with a restart entry selected and ready to launch.
PAR6 — initial guard after Stop, before refreshing controller state.