Skip to content

(8/8) chore(setup): share release bounds, atomic writes, capped reads, chatter JSON - #147

Merged
dizys merged 7 commits into
dizys:mainfrom
0libote:chore/shared-helpers-locking
Nov 15, 2026
Merged

dizys merged 7 commits into
dizys:mainfrom
0libote:chore/shared-helpers-locking

Conversation

@0libote

@0libote 0libote commented Nov 14, 2026 •

Copy link
Copy Markdown
Contributor

Release bounds, atomic file writers, capped readers and CLI JSON parsing were duplicated across the runtime, harness and Setup layers. Consolidate their shared contracts while keeping imports within each installed image layer. Document the native-settings queues' single-writer and revision-checking behavior.

Native CI now reads the exact Node version from each upstream release tag's package manifest and passes its official builder image into both architecture jobs. Stable and Nightly can require different compiler versions, as they do for Node 26.12.1 and 26.13.1. Invalid or unavailable metadata stops planning. Preserve upstream's native ARM64 SEA packaging and the existing container runtime Node configuration.

Retain the Files transfer smoke check's stage, HTTP status and assertion output in the existing diagnostics log. Its 20 MiB upload, atomic publication, checksum, Trash and restore assertions remain intact.

Validation:

  • The helper consolidation passed 188 focused tests and the saved combined-source validation.
  • All 31 native build, release-plan, distribution and helper tests passed after the compiler-selection change, including divergent channel toolchains, exact-version requests, invalid manifests and bounded metadata failures.
  • Live manifests select the correct compiler for Stable 0.1.9 and Nightly 0.1.10-nightly.20261114.4709. Both official Node images provide amd64 and arm64 manifests.
  • Workflow YAML and build arguments were parsed and checked; changed JavaScript syntax and whitespace checks passed.
  • Files diagnostics passed the actual extracted smoke function against an isolated HTTP server, including a controlled failure that retained its stage/status without logging its credential. Twelve complete local transfers and 56 focused Files/helper tests passed.
  • Full native and image builds remain gated by CI on both architectures.

Thanks to @0libote for the original helper consolidation. Maintainer review adds release-specific compiler selection and retained Files failure diagnostics.

…ter JSON

Extract single homes for the limits and predicates both release stacks
enforce, one atomic tmp+rename writer for the setup store and instruction
editors, one capped-read module for skill, instruction and settings reads,
and one chatter-tolerant JSON parse per image layer. Document the
single-writer contract on the native-settings queues instead of adding a
cross-process lock no external writer would honor.
@0libote

0libote commented Nov 14, 2026 •

Copy link
Copy Markdown
Contributor Author

Last one (8 of 8). It's the big pure refactor, so it sits on top of everything else. Checked the full order, no conflicts.

@0libote 0libote changed the title chore(setup): share release bounds, atomic writes, capped reads, chatter JSON (8/8) chore(setup): share release bounds, atomic writes, capped reads, chatter JSON Nov 14, 2026
0libote and others added 6 commits November 14, 2026 15:48
Preserve the exact failing transfer stage, HTTP status and assertion output in the existing smoke diagnostics. Keep all upload, checksum and Trash assertions intact.
Read an exact Node engine version from each source tag and pass that official builder image to both architecture jobs. Stable and Nightly can advance independently without breaking ARM64 native SEA packaging. Reject malformed metadata and preserve the bundled runtime Node configuration.
@dizys
dizys merged commit 19b9bb2 into dizys:main Nov 15, 2026
14 checks passed
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.

2 participants