[fork] trunk#2
Draft
patroza wants to merge 56 commits into
Draft
Conversation
github-actions
Bot
added
vouch:trusted
omegent-app
Bot
force-pushed
the
fork/changes
branch
from
July 24, 2026 20:32
a872da5 to
a671fa2
Compare
omegent-app
Bot
force-pushed
the
fork/tim
branch
from
July 24, 2026 20:32
17679d2 to
5348a83
Compare
patroza
force-pushed
the
fork/tim
branch
2 times, most recently
from
July 25, 2026 05:23
c6b394a to
83d75d0
Compare
patroza
force-pushed
the
fork/changes
branch
2 times, most recently
from
July 25, 2026 08:30
2a67c78 to
27a72d4
Compare
patroza
added
FORK
and removed
vouch:trusted
* chore: format public fixtures * ci: avoid PR size concurrency self-cancellation * ci: remove PR size workflow
* fix(server): bound provider shutdown drains * fix: repair memory candidate integration checks
* ci(mobile): build labelled PR previews with the release profile The preview workflow built `preview:dev`, a development-client profile, so labelled PR builds ran unminified JavaScript with dev-only assertions. Those builds cannot show how a change behaves for performance or memory. Point the workflow at the plain `preview` profile and give it the two settings it was missing for that role: the fingerprint version policy, so the continuous-deploy-fingerprint action can still reuse compatible builds and publish OTA updates, and an APK Android build type, so the artifact linked from the PR installs directly. `preview:dev` stays for local Metro attachment. Also seed the preview EAS environment with the T3CODE_MOBILE_* variables that only the development environment carried, so remote builds resolve this fork's project instead of the upstream fallback. Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com> * ci(mobile): limit labelled PR previews to iOS Android has no signing keystore on the Expo project, so a `platform: all` preview build fails on credentials before the iOS artifact is published. Restore Android here once a keystore exists. Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com> --------- Co-authored-by: Claude Opus 5 (1M context) <noreply@anthropic.com>
Integration deploys dispatched the production workflow with mode=update, which publishes an OTA update and nothing else. An OTA cannot carry a native runtime change, so any integration that touched native inputs left the installed tester build behind with no signal and no new binary. Add mode=auto, which runs the same fingerprint deploy the development track uses: a JavaScript-only integration publishes to the production channel, and a native change starts a production build that EAS submits to TestFlight through auto-submit-builds. Dispatch that mode from fork CI, restricted to iOS because Android has no signing keystore. Co-authored-by: Claude Opus 5 (1M context) <noreply@anthropic.com>
* fix(discord-bot): make Discord thread title sync durable and race-safe Stop VCS callbacks from re-applying ⏳ after settle using a stale running thread, keep a pending desired title for rate-limit/timeout retries, and retry desynced titles on a periodic tick so idle threads no longer stay wrong. * chore(fork-stack): add update to keep feature PRs mergeable Add `pnpm fork:stack update [--push] [pr]` to rebase or replay feature branches onto fork/changes, retarget wrong bases, and document the agent handoff so PRs are not opened against the upstream main mirror. * fix(fork-stack): parse gh JSON under FORCE_COLOR agent hosts Disable ANSI color in stack subprocess env so `gh --json` stays valid JSON when agents run with FORCE_COLOR set. * fix(fork-stack): force plain gh JSON for agent FORCE_COLOR hosts Pass --color=never and strip residual ANSI so stack helpers can parse `gh --json` when FORCE_COLOR is set by the agent environment. * fix(fork-stack): strip ANSI from gh JSON under agent FORCE_COLOR Avoid --color flags the t3 gh wrapper rejects; force FORCE_COLOR=0 and strip residual SGR sequences so stack update/rebase can parse --json. --------- Co-authored-by: omegent-app[bot] <306514130+omegent-app[bot]@users.noreply.github.com>
) When the stack rewrites fork/changes, also force-with-lease rebase every open PR that targets it (conflicts are skipped and summarized). Add `pnpm fork:stack pull` for local checkouts after remote rewrites: hard-reset when local commits are patch-equivalent to remote, otherwise rebase unique unpushed work. Fix gh --json parsing under FORCE_COLOR. Co-authored-by: omegent-app[bot] <306514130+omegent-app[bot]@users.noreply.github.com>
) Every worktree-backed thread creation surfaced a "Queued message was rejected: Git command failed in GitVcsDriver.createWorktree" toast. The composer persists each outgoing turn to the outbox before sending, and the outbox drain effect fired while the original bootstrap send was still in flight (worktree creation keeps it pending for seconds), so the same turn was delivered twice. The replay re-ran `git worktree add -b` and failed on the branch the first delivery had just created. Client: track in-flight sends and skip them in the outbox drain. Server: when a bootstrap replay reaches prepareWorktree for a thread that already has a recorded worktree, reuse it instead of re-creating it, and skip the duplicate setup-script run. Co-authored-by: Claude Fable 5 <noreply@anthropic.com>
…#52) When fork/changes is rewritten, do not replay the full GitHub PR commit list. Use patch-id uniqueness (git cherry), oldest-first cherry-picks, and auto-skip large conflicting layer commits so multi-generation drift can resync safely. Cascade uses the same fallback after rebase --onto. Co-authored-by: omegent-app[bot] <306514130+omegent-app[bot]@users.noreply.github.com>
Drop the arbitrary file-count skip heuristic. A PR's commits are exactly oldBase..head where oldBase is the newest recorded fork/changes tip still ancestral to the branch. The cascade appends each tip to refs/t3/stack/base-history/fork-changes and rebases with --onto. Co-authored-by: omegent-app[bot] <306514130+omegent-app[bot]@users.noreply.github.com>
* fix(mobile): hand connected follow-ups to server queue * feat(mobile): render optimistic pending messages * feat(mobile): control queued follow-up messages * fix(mobile): make send optimistic on the outbox critical path Update the in-memory outbox before disk I/O, clear the draft immediately, and paint local pending bubbles even while thread detail is still loading so send no longer waits on durability or snapshot hydration. * fix(web): reconcile queued send path after candidate replay
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
1 participant
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.
Purpose
Tracks fork-owned feature layer on top of
fork/candidates.Private operations, host topology, credentials, machine configuration, and deployment automation are deliberately excluded. This is a long-lived stack branch and remains a draft;
fork/integrationis the deployable aggregate.