Skip to content

fix(instances): stage native dev-local builds and report live status#38

Open
vernonstinebaker wants to merge 2 commits intonullclaw:mainfrom
vernonstinebaker:fix/refresh-dev-local-binary-v2
Open

fix(instances): stage native dev-local builds and report live status#38
vernonstinebaker wants to merge 2 commits intonullclaw:mainfrom
vernonstinebaker:fix/refresh-dev-local-binary-v2

Conversation

@vernonstinebaker
Copy link
Copy Markdown
Contributor

@vernonstinebaker vernonstinebaker commented May 5, 2026

Summary

  • stage dev-local binaries at the canonical ~/.nullhub/bin/<component> path and prefer local zig-out/bin/<component>-native builds when available
  • refresh managed dev-local binaries before start so rebuilt local binaries are picked up without manual cache cleanup
  • report imported standalone nullclaw/default status from its live gateway health instead of stale manager state, and ignore zombie PIDs on macOS

Why

  • the managed-instance regression was caused by NullHub launching an outdated staged dev-local binary, not by generated config
  • imported standalone default could also remain falsely running because the old runtime state kept a zombie pid that kill(pid, 0) still treated as alive on macOS
  • together these left NullHub showing or starting the wrong runtime even after the fixed local NullClaw binary was available

Validation

  • zig build
  • zig build test -Dbuild-ui=false -Dembed-ui=false
  • live NullHub smoke against http://127.0.0.1:19800 with default, willy-nilly, and instance-1 all reporting healthy
  • verified managed dev-local staging now matches the rebuilt local NullClaw binary by sha256

Notes

@vernonstinebaker
Copy link
Copy Markdown
Contributor Author

This is the focused follow-up for the managed NullClaw startup regression uncovered during validation of #34. Downstream context: NullClaw already contains the daemon nonblocking-accept fix, and the remaining issue for NullHub-managed instances was stale staged binaries. Related downstream PR context: nullclaw/nullclaw#878.

@vernonstinebaker
Copy link
Copy Markdown
Contributor Author

This is the focused follow-up for the managed NullClaw startup regression uncovered during validation of #34.

Downstream context: NullClaw main already contains the daemon nonblocking-accept fix, and the remaining issue for NullHub-managed instances was stale staged dev-local binaries.

Related downstream PR context: nullclaw/nullclaw#878.

Managed starts were still picking stale local artifacts, and imported standalone instances could stay falsely running behind zombie PIDs on macOS.
@vernonstinebaker vernonstinebaker changed the title fix(instances): refresh staged dev-local binaries on install and start fix(instances): stage native dev-local builds and report live status May 5, 2026
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

1 participant