dispatcher: subscribe before the opening rows, and ADR-0058's retired rule (#374 #367) - #429
Merged
Merged
Conversation
The same defect fixed for the scheduler in #373's follow-up, in the one app that still had it. `Dispatcher.__init__` published its standing locks, run, aspects, allocation and disputed rows and subscribed afterwards. `state/run` is what a client waits for to learn the dispatcher is up, so a request, a hold or a placement can arrive the instant after it, and an event is not retained: nothing replays what lands in the gap. Over a broker a publish is asynchronous where a subscribe waits to be acknowledged, so the gap is a round trip wide; in one process it had no width and the order never showed. Safe to subscribe first because `_on_dispatch` already ignores this app's own announcements coming back past it, and nothing on `tc49/layout/#` is published from this constructor. Found by the whole-range review of 1fd42cf..05e22c7, which is the first look anything has taken at those 48 commits together. The layout interface and the simulator had the same order and were fixed while #394 landed; the dispatcher was missed by every per-issue audit and by that fix. `Recording` and the assertion move from `tests/scheduler/test_scheduler.py` to `tests/harness.py`, so the second app to want them does not copy them. Both apps now assert the order on the in-process bus rather than waiting for a slow broker to expose it. ADR-0058 stated the rule #367 retired as current fact, under a heading saying it was settled and untouched, and #368 sends readers there for the fold. Its existing banner covered only `tc49 live --station`. The section now states the rule that holds and carries its own amendment note. Co-Authored-By: Claude Opus 5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01NNNTTEXDua7ZG2M7fvTLjk
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
1 participant
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.
Two live findings from the whole-range review of
1fd42cf..05e22c7— the 48commits of the first two ADR-0059 batches, which merged on per-issue audits
alone because both runs died on a quota limit before their own review ran.
The dispatcher publishes before it subscribes
The same defect that failed CI twice on #425 and was fixed for the scheduler,
in the one app nobody caught.
state/runis what a client waits for to learnthe dispatcher is up, so a gesture can arrive the instant after it, and an
event is not retained.
The layout interface and the simulator had the same order and were fixed while
#394 landed. The dispatcher was missed by every per-issue audit, by that fix,
and by me when I wrote the ruling saying the order applies to every app.
Recordingand its assertion move totests/harness.pyrather than beingcopied into a second suite. The new test fails without the fix.
ADR-0058 states a retired rule as current
#367retired "an address names its system as its first level". ADR-0043 andADR-0022 got amendment banners; ADR-0058 said it under a heading reading "What
ADR-0043 already settles, and this does not touch", and #368 sends readers
there. Now states the rule that holds, with its own note.
Not in this PR
Four findings filed as issues instead — the six-way duplication across the app
command lines,
cold startas vocabulary the glossary tells you to avoid, thestartup order that contradicts its own docstring, and the missing check between
CONTEXT.md and the inventory.
🤖 Generated with Claude Code
https://claude.ai/code/session_01NNNTTEXDua7ZG2M7fvTLjk