[kiwidream]

	Files	Added	Deleted	Net
Test	1172	$\color{#1a7f37}{\Huge{\mathbf{+}}}$​16642	$\color{#cf222e}{\Huge{\mathbf{−}}}$​18502	$\color{#cf222e}{\Huge{\mathbf{−}}}$​1860
Prod	1828	$\color{#1a7f37}{\Huge{\mathbf{+}}}$​11174	$\color{#cf222e}{\Huge{\mathbf{−}}}$​182622	$\color{#cf222e}{\Huge{\mathbf{−}}}$​171448

ELI5

Upstream Orca moved forward 236 commits since yesterday's sync (#50). This PR merges them into swaplabs/main. 21 files conflicted with fork changes. Most of them (17) come from upstream's big comment cleanup, which shortened comments that sit next to fork lines; the fork's lines are kept unchanged. Three more fork files needed small fixes that the conflict list did not show. This PR replaces #52 (head main), which could not merge.

What Changed

Upstream range 5d32f372e7..30733ac703, 236 commits (compare). One signed merge commit, c361964f6a (Merge upstream main (30733ac703) into swaplabs/main), on sync/upstream-20261024, cut at the integration head 7d33e616ae.

Conflicts, by fork PR (orca-fork-ops conflicts)

Upstream's comment audit (2847bbe9e81, stablyai#32966) shortened the upstream comment beside a fork line. In each of these files the resolution is upstream's new comment plus the fork's lines, unchanged:

Fork PR	Files
#6 (auto-resume)	src/main/ai-vault/cached-session-list.ts, session-list-results.ts, session-scan-issues.ts
#10, #16, #27 (update sources, mac self-update, feed tags)	src/main/updater-prerelease-feed.ts, updater-release-build-cache.ts, updater-release-builds.ts, updater-release-builds.test.ts, updater.build-channel-selection.test.ts, updater-mac-install.ts
#25 (daemon drain)	src/main/daemon/daemon-endpoint-ownership.ts, daemon-out-of-process-launcher.ts, daemon-pty-router.ts, src/main/ipc/pty-management.ts
#30 (host resume)	src/main/agent-hooks/server/server-cleanup.ts, server-hydration.ts, server-types.ts
#30, via the past sync resolution 1a247d1c5c	src/main/agent-hooks/server/server-remote-envelope-normalization.ts

Details worth knowing:

• Where the fork rewrote an upstream comment to describe its own behaviour (server-cleanup.ts's preserveResumeIdentity doc, updater-release-builds.ts's resolveTargetBuild doc), the fork's comment stays in place of upstream's shortened or removed one.
• Where the fork deleted or moved the code a comment sat on, upstream's rewrite of that comment is dropped with the code: updater-prerelease-feed.ts (the fork moved the readiness probes to updater-release-asset-readiness.ts), the protocol-version lookup in pty-management.ts, and the routing comment in daemon-pty-router.ts.
• session-scan-issues.ts: the fork's added sentence (why the bound is shared) now follows upstream's one-line comment.
• updater-release-builds.test.ts: the fork's tests sat between upstream's Windows comment and the test it belongs to. Upstream's rewritten comment now sits above that test again.

The other four conflicts:

Fork PR	File	Upstream change	Resolution
#8 (fork build pipeline)	.gitignore	stablyai#32903 removed the docs/audits allow-list	upstream's removal, plus the fork's two docs/reference/swaplabs-*.md lines
#17 (managed-server update help)	src/renderer/src/i18n/locales/es.json	b9d9dacb008 translated unpackedHelp	upstream's translation, plus the fork's managedByDeploymentHelp and howToUpdate keys
#30 (host resume)	src/main/agent-hooks/manual-compact-status-cleanup.test.ts	stablyai#32991 rewrote the test as one case per cleanup	upstream's test, with the fork's expectation ported in: pane cleanup leaves only the retained resume identity (providerSessionOnly), and the late PostCompact revives nothing
#30 (host resume)	src/renderer/src/runtime/web-session-terminal-handle-events.ts	stablyai#33062 adds incarnationId to the handle update, so a remote pane can re-bind after a host exit	upstream's: it carries the same field the fork added (see below)

web-session-terminal-handle-events.ts: upstream stablyai#33062 now does what the fork's lines in this file did. Fork commit 6713337f23 added incarnationId: string | null to WebSessionTerminalHandleUpdate, set to null when unknown, so announceProcessReplacement can tell a replacement process from a re-minted handle. Upstream's field is optional and is present only when the host stamps one. The fork's consumer in remote-runtime-pty-transport.ts already treats an absent value as null: rebindRemoteTerminalHandle defaults it, and announceProcessReplacement checks truthiness. The behaviour is unchanged, and the fork's remote-runtime-pty-transport-process-replacement.test.ts passes. Fork lines this replaces:

• the type field and both incarnationId: constructions in web-session-terminal-handle-events.ts;
• the incarnationId: null expectations in web-session-terminal-handle-events.test.ts and web-session-tabs-sync-visibility-collision.test.tsx, which are back to upstream's text.

Breaks the conflict list did not show

Fork PR	File	Upstream change	Fix
#6 (auto-resume)	pty-connection/candidate-resume-startup.ts, agent-resume-recovery-install.ts	stablyai#31315 stores each agent's permission; resolveTuiAgentLaunchArgs/Env now take the launch-profile settings and the target platform	follow the new signature: pass state.settings and the target platform, as upstream's own resume path (agent-account-restart-startup.ts) does. A resumed agent gets the stored permission like every other launch, which matches the fork's documented rule that the pane's configured defaults apply
#3 (unreachable-host removal)	src/renderer/src/store/repos/project-removal-outcome.ts	stablyai#33024 moved remote-runtime-client-error-classification into src/shared/remote-runtime/	follow the move (import path only)
#10, #16 (update sources, mac self-update)	electron.vite.config.ts	stablyai#30076 added 4 lines; the clean auto-merge left the file at 302 counted lines (max 300)	the fork's two ...SWAPLABS_UPDATE_DEFINES spreads and the renderer comment move into withSwaplabsUpdateDefines in the fork-owned config/build-plugins/swaplabs-update-defines.ts, applied once to electronViteConfig. The fork now adds one counted line here (the import). Both define blocks carry the same keys as before; checked by loading the config

No other fork lines were moved or dropped. For 74 of the 82 files both sides changed, the fork's added and removed lines against upstream are the same before and after the merge. The other 8 are the comment and port cases above.

Some notable upstream changes in the range

• the comment audit: shorter comments across 2,753 files (2847bbe9e81, [comment audit] Clean up IPC, host lifecycle and native-chat comments stablyai/orca#32966, [comment audit] Continue main-process comment cleanup stablyai/orca#32946 and earlier audit PRs);
• agents store each agent's permission themselves; terminals and new chats read it (Agents: store each agent's permission itself; terminals and new chats read it stablyai/orca#31315);
• remote terminals: a new SSH terminal opened before sign-in waits for it (fix(terminal): a new SSH terminal opened before sign-in waits and starts after sign-in stablyai/orca#33072, fix(terminal): restored SSH terminals recover after password sign-in stablyai/orca#32275); a host-exit-settled pane re-binds on the host's next snapshot (fix(terminal): re-bind a host-exit-settled remote pane on the host's next snapshot stablyai/orca#33062); output and parse barriers survive an xterm resize (fix(terminal): preserve output and parse barriers across xterm resize stablyai/orca#32750);
• worktrees: Orca no longer prunes worktrees the other Git (Windows or WSL) created (fix(worktrees): don't prune worktrees the other Git (Windows or WSL) created stablyai/orca#33103), and reclaims stale dependencies and build output across worktrees (feat: reclaim stale dependencies and build output across worktrees stablyai/orca#32331);
• browser: history controls (feat(browser): history controls: Chrome-style Escape, removable suggestions, history view stablyai/orca#31846), a per-repo or per-workspace profile (feat(browser): per-repo or per-workspace browser profile for new tabs stablyai/orca#32039), links open beside their pane (feat(browser): open links beside their pane, focus link tabs, honor window.close() stablyai/orca#31844), and pages say why they failed to load (feat(browser): say why a page failed to load and reload it when it can stablyai/orca#31847);
• the remote-runtime modules move into src/shared/remote-runtime/ (refactor: colocate remote runtime error classification stablyai/orca#33024, refactor: group remote runtime requests and subscriptions stablyai/orca#32952, refactor: colocate remote runtime passthrough socket stablyai/orca#33009 and others);
• CI fails when a checked-in test file has no runner (ci: fail when a checked-in test file has no runner stablyai/orca#33080); routine checks move to Blacksmith (ci: move routine checks and unsigned builds to Blacksmith stablyai/orca#32913).

Why

Daily upstream sync.

Linked Issue

N/A: daily upstream sync.

Visual Proof

N/A: no UI change of the fork's. The conflicts are in comments, tests, main-process code and the renderer's terminal transport.

Testing

All checks ran on the merge result in the sync worktree, before the commit, except where noted:

• Install: pnpm install --frozen-lockfile (the lockfile and every package.json equal upstream's).
• Typecheck: pnpm tc passes after the two fixes above, and again on the committed tree. Before them it failed on exactly those files. A deliberate type error in a scratch file made it fail, which confirms the gate is live.
• Fork tests: every unit test file the fork adds or changes against upstream, plus orca-runtime.test.ts (it runs the fork's worktree-ps-agent-row-dismissal.spec.ts) and the two handle-event test files: 118 files, 117 pass. The one failure is upstream's repository-project-operations-part-02.spec.ts, because this host's signing wrapper refuses its fixture commit init. With fixture signing off (commit.gpgsign=false), orca-runtime.test.ts passes: 1,497 passed, 3 skipped. The fork's 3 e2e specs were not run. This run came before the electron.vite.config.ts move, so the files that check the build defines (swaplabs-build-workflow-mac-signing, verify-swaplabs-packaging, verify-dev-channel-packaging, release-sources, mac-self-update-public-key; 95 tests) were run again after it, and pass.
• Touched directories: src/main/agent-hooks, src/main/ai-vault, src/main/daemon, src/main/ipc/pty-management*, src/main/updater*, src/renderer/src/runtime/web-session*, src/renderer/src/components/terminal-pane, src/renderer/src/store/repos, plus electron-vite-output-contract.test.ts and main-entry-facade.test.ts: 1,363 files, 13,634 tests passed. Two sets of failures:
  • 3 upstream bash-wrapper files (bash-prompt-command-composition, repro-13767-shell-ready-marker-lost-to-exec, shell-ready-bash-wrapper; 7 tests) fail the same way on pristine upstream 30733ac703 on this host (bash 5.3.9), so they are not the merge's.
  • pty-subprocess-git-credential-guard.test.ts failed only under that run's GIT_CONFIG_* override and passes without it.
• Changed-code gate: check-changed-code-quality.mjs with ORCA_SYNC_PR_DIFF_BASE=30733ac703: 0 findings across 328 fork files.
• Lint and format of everything the merge stages (7,636 code files and 83 others outside cloud/): oxlint passes with both configs, and oxfmt --check reports every matched file formatted. The pre-commit hook then ran normally, and the tree before the commit (3293e86747) equals HEAD^{tree} after it.
• Catalogs and ratchets: the 7 locale catalogs parse with no duplicate keys. verify:localization-runtime-catalog, -catalog, -catalogs, -extraction, -coverage, verify:rpc-params-catalog, check:max-lines-ratchet and upstream's new check:test-file-runners all pass.
• Signature: c361964f6a is G from the sync key BF1DD75F888E96080C2453E2D9E158C5D6874235.

AI Disclosure

Prepared by the "Orca upstream sync" automation (Claude Code, Claude Opus 5.5).

Agent skill upstream boundary

• Not applicable: this PR merges upstream's history; it adds no agent skill.

Notes

orca-fork-ops status for swaplabs/main (7d33e616ae) against upstream 30733ac703, before this merge:

• merged upstream: 0;
• pending upstream: 0;
• local forever: 83 (unchanged since sync: upstream 5d32f372e7 into swaplabs/main #50);
• 30 commits have no LOCAL: prefix and no upstream PR.

Upstream release tags v1.4.238 and v1.4.239 were copied to the fork.

🤖 Generated with Claude Code

---

^{Need help on this PR? Tag @codesmith-bot with what you need.}

[chatgpt-codex-connector]

Codex Review Summary

This comment shows the latest Codex review activity on this pull request.

Review	Status	Commit	Review trigger
📝 Code Review	✅ Completed 2026-10-24T10:28:20.391695Z	c361964	PR opened

ℹ️ About Codex in GitHub

Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you

• Open a pull request for review
• Mark a draft as ready
• Comment "@codex review" or "@codex security review".

Codex reacts with 👀 while any review is running, comments if it has suggestions, and reacts with 👍 once all reviews finish with no findings.

[kiwidream]

Post-merge follow-ups from the sync automation:

• Host resume (#30): upstream PTY pane-copy census flags host-pane-agent-resume-start.ts #54: upstream's PTY pane-copy census (pty-pane-copy-census.test.ts) flags the fork's host-resume code host-pane-agent-resume-start.ts:130. It is the cause of the red test / tests node 24 1/5 shard here. The failure predates this merge: it fails the same way on the pre-merge integration head 7d33e616ae.
• Resume: the fork's automatic resumes relaunch with settings-default flags #40 (comment added): both fork resumes now read the stored agent permission through upstream Agents: store each agent's permission itself; terminals and new chats read it stablyai/orca#31315's launch-profile settings.

Other red checks are upstream's own, with no fork lines involved:

• Mobile verify: mobile-native-chat-merged-snapshot-parity.test.ts timed out after 60 s.
• Cloud Verify test: cloud/apps/relay relay.blackbox.test.ts got device-resume-confirmed where it expected control-error.
• PR checks passed: the aggregate gate over the failures above.