Repository navigation
docs(sharing): record OpenCode 2.2.5 platform admission - #580
Merged
Merged
Conversation
This was referenced Oct 31, 2026
stamp-build Bot
added a commit
that referenced
this pull request
Oct 31, 2026
…582) <!-- stamp:review:v1 --> <sup><img src="https://cdn.slock.ai/avatars/6aabf32c-20bc-4cba-b55a-2aa72b60999b/8ba4452ceb890f1930886ef4286fdb99.webp" width="16" height="16" align="absmiddle" alt=""> <b>Otter</b> from open-agent-runtime opened this pull request through Stamp</sup><br><sup><a href="https://stamp.build/review?receipt=prc_319c609fc7f214496100898fd3cfd137"><picture><source media="(prefers-color-scheme: dark)" srcset="https://stamp.build/assets/brand/stamp-glyph-dark.d907ffcadb44.svg"><img src="https://stamp.build/assets/brand/stamp-glyph-light.58a2c10ed108.svg" width="13" height="12" align="absmiddle" alt=""></picture> View in Stamp →</a></sup> --- <!-- /stamp:review:v1 --> The successful two-turn orphan cleanup regression could race a Windows worker's first background process-table preparation. Wait for that preparation to succeed before exercising cancellation, then require both fresh cancellation observations and all four original process-cleanup assertions to succeed. Failed assertions include each observation's result, duration and preparation state, plus the final exited response and its reason. Add two actual ACP process controls using the real serial batcher and real process-table reads. Holding a background query beyond the first cancellation's unchanged one-second observation budget leaves the first tool unclaimed, cleans the later tool, and reports `before_cancel ... timeout` in `exited.reason`. Completing that query before cancellation cleans both turns' tools. The test explicitly checks that no second query starts while the first remains blocked. The original PR #580 Windows failure (run 41223079686, job 124601496941) did not retain these diagnostics. The controlled reproduction proves this queue mechanism, not that it caused the original CI failure. Product code, deadlines and behavior-suite entry points are unchanged. Validation: the two focused files pass all 10 tests; full check passes 2,858 tests and 18 mock behaviors. Windows CI will verify the preparation prerequisite and both controls on its native process table. Co-authored-by: Otter <open-agent-runtime+otter@agents.stamp.build>
![https://stamp.build/assets/brand/stamp-glyph-dark.d907ffcadb44.svg"><img](https://stamp.build/assets/brand/stamp-glyph-dark.d907ffcadb44.svg"><img){kind=link}
stamp-build
Bot
force-pushed
the
agent/otter/opencode-2.2.5-evidence
branch
from
October 31, 2026 17:45
9af830a to
f9cf3d7
Compare
stamp-build Bot
added a commit
that referenced
this pull request
Oct 31, 2026
<!-- stamp:review:v1 --> <sup><img src="https://cdn.slock.ai/avatars/6aabf32c-20bc-4cba-b55a-2aa72b60999b/8ba4452ceb890f1930886ef4286fdb99.webp" width="16" height="16" align="absmiddle" alt=""> <b>Otter</b> from open-agent-runtime opened this pull request through Stamp</sup><br><sup><a href="https://stamp.build/review?receipt=prc_b86116cad712e3aaa4b321202d58ee94"><picture><source media="(prefers-color-scheme: dark)" srcset="https://stamp.build/assets/brand/stamp-glyph-dark.d907ffcadb44.svg"><img src="https://stamp.build/assets/brand/stamp-glyph-light.58a2c10ed108.svg" width="13" height="12" align="absmiddle" alt=""></picture> View in Stamp →</a></sup> --- <!-- /stamp:review:v1 --> Windows can finish native exit before asynchronous process cleanup has emitted the final `exited` record. Two tests used Vitest's default one-second poll, and a protocol-close test allowed only five seconds, the same as one Windows cleanup read budget. Give all three polls 15 seconds and attach the records captured when a poll fails, retaining the original assertion as the cause. The shared crash still requires one exit record per session, native exit code 7, and no duplicate after disposal. The isolated ACP tests still require the explicit terminal-cleanup failure or failed turn and rejection after protocol close. Product timeouts, cleanup behavior, and the 30-second test ceiling are unchanged. Validation: - 18 focused tests across the three affected files passed. - `CI=1 pnpm run check`: 2,858 unit tests and 18 mock behavior cases passed. - Audited the remaining `expect.poll` calls: readiness polls and polls after awaited disposal do not wait for this final-record cleanup gap; existing explicitly bounded cancellation checks remain unchanged. Task #140. The triggering failure is the Windows check job [124605383846](https://github.com/botiverse/oar/actions/runs/41223079686/job/124605383846) in PR #580. The log records zero exit responses after the default poll; it does not retain the cleanup phase timings. Co-authored-by: Otter <open-agent-runtime+otter@agents.stamp.build>
stamp-build Bot
added a commit
that referenced
this pull request
Oct 31, 2026
<!-- stamp:review:v1 --> <sup><img src="https://cdn.slock.ai/avatars/6aabf32c-20bc-4cba-b55a-2aa72b60999b/8ba4452ceb890f1930886ef4286fdb99.webp" width="16" height="16" align="absmiddle" alt=""> <b>Otter</b> from open-agent-runtime opened this pull request through Stamp</sup><br><sup><a href="https://stamp.build/review?receipt=prc_e979c0ee5ba7d18ada9f19829b7cf4ea"><picture><source media="(prefers-color-scheme: dark)" srcset="https://stamp.build/assets/brand/stamp-glyph-dark.d907ffcadb44.svg"><img src="https://stamp.build/assets/brand/stamp-glyph-light.58a2c10ed108.svg" width="13" height="12" align="absmiddle" alt=""></picture> View in Stamp →</a></sup> --- <!-- /stamp:review:v1 --> OpenCode 2.2.5 now uses shared placement under `auto` on Linux x64, macOS arm64 and Windows x64, following the independently reviewed native evidence in #580 (run 41221951609). Keep the existing 2.2.3 admission. Installation parsing and the running ACP process both use the same exact-version allowlist; unknown versions, prereleases and unmeasured platforms still fall back or refuse explicit shared placement. Each canonical directory remains reserved for the process lifetime. Reusing it, including after release, falls back under auto and refuses explicit shared placement. The shared and isolated cleanup CI entries now pin 2.2.5; they no longer exercise native 2.2.3 sharing. Native cold-start database contention and cold-resume MCP limitations remain recorded. Release note: `auto` now shares OpenCode 2.2.5 on the three admitted platforms. Hosts that select `isolated` keep independent processes. Validation: full check (2,874 tests plus 18 mock behaviors), build and actionlint passed. Real 2.2.5 shared behavior passed 15 checks with four capability skips; shared and isolated lifecycle controls and the standalone isolated cleanup check passed. Focused tests cover both admitted versions, unsupported version/platform boundaries, native initialization rejection, directory aliases and lifetime reuse. Based on main after #580. The evidence index and admission report now match the adapter: both exact versions are enabled with the lifetime directory restriction. The native-version predicate and platform-refusal function each have their own contract comment. --------- Co-authored-by: Otter <open-agent-runtime+otter@agents.stamp.build>
Merged
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.
OpenCode 2.2.5 now has complete native sharing evidence on Linux x64, macOS arm64 and Windows x64 from run 41221951609. Record admission with the same lifetime directory restriction as 2.2.3: same-directory MCP registrations still cross over and survive native close. Product enablement and CI pins remain a separate change.
The report retains all three failed job conclusions and their exact same-directory failure reasons. Each platform completed seven native observation groups, both project/user configuration comparisons, and both resource cohorts. All 18 post-turn CPU windows are exact; shared memory and CPU are lower. The manifest identifies all three archives and hashes each of the twelve evidence files; the selector rebuilds the selected data without transcripts, temporary paths or native/process ids.
Validation:
CI=1 pnpm run check: 2,858 unit tests and 18 mock behavior cases passed.Task #138. Preserve the existing 2.2.3 admission; no other versions or architectures inherit these results.