Skip to content

ci: per-commit SEA build+smoke+boot gate - #10

Merged
schmonz merged 1 commit into
mainfrom
ci/sea-gate
Jul 4, 2026
Merged

schmonz merged 1 commit into
mainfrom
ci/sea-gate

Conversation

@schmonz

@schmonz schmonz commented Jul 4, 2026

Copy link
Copy Markdown
Owner

Adds ci.yml: 3-OS SEA build + offline smoke + REQUIRED real bundle boot on every push/PR. Validates via this PR before landing. Retires the windows-boot spike.

New standalone ci.yml on push(main)+PR+dispatch, matrix ubuntu/macos/windows:
build clode.exe, npm-install the Claude provider, then one node:test run does
build + offline smoke + REQUIRED real bundle boot (CLODE_CLAUDE_BIN set, -p
skipped offline). Provider located via npm root -g (portable) under the
claude-code package (nested platform Bun binary on POSIX, bin/claude.exe on
Windows). Lights up bundle-boot coverage on all 3 platforms. release.yml
untouched. Validated on the real runners via PR (local WSL 9p mount too flaky).

Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01JR9MxqjNcLHxAuZw87asNN
@schmonz
schmonz merged commit f87dc87 into main Jul 4, 2026
3 checks passed
@schmonz
schmonz deleted the ci/sea-gate branch July 4, 2026 21:13
schmonz pushed a commit that referenced this pull request Jul 10, 2026
Dispatch #11 diagnostics were decisive:
- openbsd: the BARE-ENGINE probe fails — tjs.spawn(["/bin/sh","-c",
  "echo ok"]) exits 127 on OpenBSD, so tjs/libuv spawn is broken
  wholesale there; the builder was never the problem. The pinned libuv
  (posix_spawn on every platform, vs upstream's macOS-only) fails
  child-side in OpenBSD's libc posix_spawn. Fixup forces
  posix_spawn_works=0 on OpenBSD, selecting the fork/exec fallback —
  the battle-tested path upstream uses everywhere else. The deeper why
  is an open report candidate for the libuv fork.
- dragonflybsd: sign-conversion demotion proven (lws fully compiles);
  the build marched into txiki's mem.c and hit the OpenBSD wall —
  DragonFly ships no malloc.h either. mem.c guard now excludes both
  (with an upgrade path from the v1 OpenBSD-only guard).

Note: dispatch #10 raced the push and ran the prior sha — dispatches
now get headSha verified against origin/main.

Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_013etF8NEcp8CHMWZxmf1d1h
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