Repository navigation
Clear stale checkpoint readiness on managed-run reuse - #80
Merged
Merged
Conversation
Co-Authored-By: GPT-6 <noreply@openai.com>
Treat the checkpoint view as launch-transient while retaining acknowledged and unsynced generations. Add a gated scheduler-start regression for issue 73. Co-Authored-By: GPT-6 <noreply@openai.com>
Report issue 73 regression proof, 99 passing managed-runner tests, full-suite limitations and remaining issue priorities. Co-Authored-By: GPT-6 <noreply@openai.com>
…ction Targeted checkpoint reuse and tunnel-status checks: 14 passed. No additional production edits beyond current main. Co-Authored-By: GPT-6 <noreply@openai.com>
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
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Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be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.
CI update (2026-10-29): merged current main to incorporate the Click test-runner stream capture correction from PR 81. The original Python 3.9 failure matched the pre-existing main failure. Fourteen targeted checkpoint-reuse/tunnel-status checks pass locally; fresh CI is green for Python 3.9, Python 3.11, and docs (run 40990436269).
Reusing a managed run could expose the previous launch's idle checkpoint count before the new scheduler started. Clients comparing counts could then wait indefinitely even though new captures succeeded (issue 73).
Clear
checkpointwith the other launch-transient status fields. Retain acknowledged generations, pending sync records, provenance and transitions. A gated startup regression submits a checkpoint request before the new scheduler starts and verifies publication and subsequent requests. No placement, ownership, sealing or timeout rules change.Validation: both new regressions fail at the stale-checkpoint assertions on baseline and pass on this branch. The full Savio suite reports 2,364 passed, 2 failed. All 99 managed-runner tests passed, including real Codex restore and same-allocation reuse. The two failures concern unchanged code: a test's hard-coded "other node" equals the actual test node (reproduced on baseline), and a dry-run test receives
ProbeTimeout(passes alone on baseline; a baseline probe error demonstrably leaves a 60-second alarm armed, likely cross-test interference). Full details and evidence paths are in the report.The source was self-reviewed; GitNexus analysis was attempted but its native library cannot load
libssl.so.3on Savio. Manual AST comparison limits the production change toStatus.load; Python 3.6 grammar still parses.The included backlog report inventories 13 open issues and examines five. Issues 70, 76 and 79 remain separate; issue 65 has a tested local fix awaiting GitHub integration. Baseline GitHub CI already has a Python 3.9 tunnel-status JSON test failure (run 40875571500).
Sue