Repository navigation
Upgrade Atomic to the stable 0.9.57 release - #751
Merged
Merged
Conversation
Moves the Atomic harness from 0.9.56 to the stable 0.9.57 release. Versions. catalog:harness pins `@bastani/atomic` 0.9.57. The lockfile resolves `@bastani/atomic`, `@bastani/atomic-natives` with its platform packages, and `@bastani/pi-ai` at 0.9.57, with no older `@bastani` package left. Nothing else in the lockfile moves. What changed in Atomic, and what it means here. 0.9.57 replaces its bundled desktop computer-use skill and guidance with agent-desktop. An atomic Planner runs Atomic in process and doesn't drive desktop apps, so its behavior doesn't change and no code changes. Assistant-model: Claude Opus 5.5 (fast) Assistant-workflow: inline Assistant-verification: bun test unit passed: GIT_CONFIG_GLOBAL=/dev/null bun test src/harness (319 pass, 0 fail) Assistant-verification: typecheck and lint passed: bun run types and bun run ci Co-authored-by: Alex Lavaee <lavaman131@github.com>
lavaman131
added a commit
that referenced
this pull request
Oct 31, 2026
Two connector tests failed on Linux CI for #751, whose diff only moved the Atomic version. One was a product bug; the other raced the connector's own restart. An agent's exit could be reported without its stderr. A failed start's report names how the agent ended and the first line of what it printed. It was built once the child's exit event fired, which Node can emit before the last of its stderr has been read, so the report could say "exit 127" and leave out why. Someone running chopin or reading chopin status lost the one line that explains the failure. The report now waits for the child to close after it exits, still within the one second it already allowed, so a process the agent left holding stderr open cannot hold it up. The exit status still comes from the exit event. A new test has the agent exit at once and a process it left behind print 300 ms later; it failed before this change and passes after it. The long-line stderr test counted a second, legitimate drop. The Planner exits with 127 in that test, and the connector starts it again after a backoff that is uniform from zero to one second. The test stops the Planner once the failure is recorded, polling every 50 ms, so a backoff of a few milliseconds let the second start print the long line again, and its marker was rightly written twice. The product counts one marker per over-long line however it is split across chunks. The test's backend now prints the line only on its first start and idles afterwards. Its assertions are unchanged. With the backoff forced to zero the test failed 4 of 25 runs before and 0 of 25 after. Assistant-model: Claude Opus 5.5 (fast) Assistant-workflow: inline Assistant-verification: bun test unit passed: GIT_CONFIG_GLOBAL=/dev/null bun test src/acp.test.ts 30 runs and src/planner.test.ts 30 runs, 0 failures Assistant-verification: bun test under load passed: taskset -c 0 with two busy loops, the long-line Planner test and the ACP exit tests 30 runs each, 0 failures Assistant-verification: bun test unit passed: GIT_CONFIG_GLOBAL=/dev/null bun test in apps/connector (268 pass, 14 skip, 0 fail) Assistant-verification: typecheck and lint passed: bun run types and bun run ci Assistant-verification: bun test under heavier load failed: taskset -c 0 with three busy loops, src/acp.test.ts timed out 2 of 30 runs in the unrelated initialize name-and-version test at its 5 s default Co-authored-by: Alex Lavaee <lavaman131@github.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.
Moves the Atomic harness from 0.9.56 to the stable 0.9.57 release.
Versions. catalog:harness pins
@bastani/atomic0.9.57. The lockfileresolves
@bastani/atomic,@bastani/atomic-nativeswith its platformpackages, and
@bastani/pi-aiat 0.9.57, with no older@bastanipackageleft. Nothing else in the lockfile moves.
What changed in Atomic, and what it means here. 0.9.57 replaces its bundled
desktop computer-use skill and guidance with agent-desktop. An atomic Planner
runs Atomic in process and doesn't drive desktop apps, so its behavior
doesn't change and no code changes.
Assistant-model: Claude Opus 5.5 (fast)
Assistant-workflow: inline
Assistant-verification: bun test unit passed: GIT_CONFIG_GLOBAL=/dev/null bun test src/harness (319 pass, 0 fail)
Assistant-verification: typecheck and lint passed: bun run types and bun run ci
Co-authored-by: Alex Lavaee lavaman131@github.com