fix: preserve revert consistency - #46974
Open
CasualDeveloper wants to merge 1 commit into
Open
Conversation
Contributor
|
Thanks for your contribution! This PR doesn't have a linked issue. All PRs must reference an existing issue. Please:
See CONTRIBUTING.md for details. |
Contributor
|
The following comment was made by an LLM, it may be inaccurate: The search found PR #42461 "fix: make revert boundaries chronological" which is the related PR mentioned in the current PR's description as being superseded. This is expected since PR #46974 explicitly states it "Supersedes #42461" and folds in that fix. No other duplicate or competing PRs were found that are addressing the same revert consistency and session prompt ordering issue. No duplicate PRs found |
CasualDeveloper
force-pushed
the
fix/37751-revert-consistency
branch
2 times, most recently
from
September 5, 2026 17:32
7c3cb99 to
c0ce754
Compare
CasualDeveloper
force-pushed
the
fix/37751-revert-consistency
branch
from
September 10, 2026 17:06
c0ce754 to
72d9544
Compare
Keep execution parked while a revert is staged and serialize revert mutations with prompt and synthetic admission. Commit replacement prompts and their optional context atomically with the revert, retaining queued input before its chronological boundary. Keep projected reads in SessionStore and scope plugin setup and hook admission bypasses to their active invocation. Serialize client mutations while preserving optimistic display and same-ID retries; resolve undo boundaries by transcript position rather than message ID. Closes anomalyco#37751, anomalyco#44357
CasualDeveloper
force-pushed
the
fix/37751-revert-consistency
branch
from
September 10, 2026 17:11
72d9544 to
f9b46da
Compare
6 tasks
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.
Issue for this PR
Closes #37751, #44357
Supersedes #37752, #42461
Type of change
What does this PR do?
I serialize revert changes and prompt admission per session, and keep model execution parked while a revert is staged. A replacement prompt and its optional synthetic context commit atomically with the revert. Pending input before the chronological boundary is retained.
The clients show input immediately, order agent/model changes with admission, retry admission once under the same message ID, and choose undo boundaries by transcript position. Plugin setup and prompt hooks can admit synthetic input through invocation-scoped capabilities without deadlocking the reservation.
session.promptgains optionalcontext, for steer delivery only. The clients, OpenAPI, and website API documents are regenerated.How did you verify your code works?
Using Bun 1.4.2 on local
v2at573d76933f:v2in the same three test files.Regression coverage includes staged resume/wake, concurrent undo and admission, atomic replacement failure, same-ID retries, plugin-hook reentrancy, pending-input preservation, and paginated undo boundaries. The three new execution tests fail without the drain guard.
Screenshots / recordings
Logic change; existing UI exercised in production-browser tests.
Checklist
Process: human-directed and reviewed, with agentic assistance from Anthropic/Claude and OpenAI/ChatGPT.