Repository navigation
chore(fork): integrate upstream v0.1.7-nightly.20261101.4016 - #63
Merged
Merged
Conversation
…offs Move an existing Library conversation into a linked worktree by coordinating its binding with the metadata update. Distinguish copy roots from attached worktrees and preserve the committed destination when provider detachment interrupts a handoff. Keep ready starts in the picker through validation and creation. Open options only when needed, retain failed attempts for deliberate recovery, and report whether a no-project launch succeeded. Validation: check:changed passed with 305 tests and all affected typechecks. Isolated Chromium checks covered contextual loading, explicit options, failure recovery, no-project launch, and a narrow viewport.
…ingdotgg#28409) Co-authored-by: Claude Opus 5.5 <noreply@anthropic.com>
…gg#28446) Co-authored-by: Claude Opus 5.5 <noreply@anthropic.com>
…ery (pingdotgg#14519) Co-authored-by: github-actions[bot] <41898282+github-actions[bot]@users.noreply.github.com> Co-authored-by: Claude Opus 5.5 (1M context) <noreply@anthropic.com> Co-authored-by: SAPHID <saphid@users.noreply.github.com>
…ingdotgg#15000) Co-authored-by: github-actions[bot] <41898282+github-actions[bot]@users.noreply.github.com> Co-authored-by: SAPHID <saphid@users.noreply.github.com> Co-authored-by: Claude Opus 5.5 <noreply@anthropic.com> Co-authored-by: maria-rcks <maria@kuuro.net>
…ingdotgg#21527) Co-authored-by: Claude Opus 5.5 <noreply@anthropic.com>
…gdotgg#21913) Co-authored-by: Simone <185146821+Lucenx9@users.noreply.github.com> Co-authored-by: maria-rcks <maria@kuuro.net>
…ntents (pingdotgg#22302) Co-authored-by: PR Batch Tester <agent@local.test> Co-authored-by: Claude Opus 5.5 <noreply@anthropic.com> Co-authored-by: maria-rcks <maria@kuuro.net>
…#23230) Co-authored-by: maria-rcks <maria@kuuro.net>
…gdotgg#23414) Co-authored-by: maria-rcks <maria@kuuro.net>
Co-authored-by: maria-rcks <maria@kuuro.net>
… runs on (pingdotgg#24833) Co-authored-by: Claude Opus 5.5 <noreply@anthropic.com> Co-authored-by: maria-rcks <maria@kuuro.net>
… on (pingdotgg#24898) Co-authored-by: Claude Opus 5.5 <noreply@anthropic.com>
Co-authored-by: github-actions[bot] <41898282+github-actions[bot]@users.noreply.github.com> Co-authored-by: SAPHID <saphid@users.noreply.github.com> Co-authored-by: maria-rcks <maria@kuuro.net>
Co-authored-by: Simone <185146821+Lucenx9@users.noreply.github.com> Co-authored-by: maria-rcks <maria@kuuro.net>
…sts (pingdotgg#25836) Co-authored-by: maria-rcks <maria@kuuro.net>
… Working shelf (pingdotgg#26195) Co-authored-by: SAPHID <saphid@users.noreply.github.com> Co-authored-by: Claude Opus 5.5 <noreply@anthropic.com> Co-authored-by: maria-rcks <maria@kuuro.net>
…ingdotgg#26247) Co-authored-by: PR Batch Tester <agent@local.test> Co-authored-by: Claude Opus 5.5 <noreply@anthropic.com>
…26666) Co-authored-by: maria-rcks <maria@kuuro.net>
…gg#27032) Co-authored-by: maria-rcks <maria@kuuro.net>
…dotgg#27268) Co-authored-by: maria-rcks <maria@kuuro.net>
…dotgg#27310) Co-authored-by: maria-rcks <maria@kuuro.net>
…g#27803) Co-authored-by: maria-rcks <maria@kuuro.net>
…7925) Co-authored-by: maria-rcks <254055478+maria-rcks@users.noreply.github.com>
Co-authored-by: maria-rcks <254055478+maria-rcks@users.noreply.github.com>
… (Running browser tests in apps/web) Failing check: Running browser tests in apps/web "Running browser tests in apps/web" failed in two files. Both were fork code that didn't fit new upstream tests. After the fixes, `DiffPanel.browser.test.tsx` and `ProviderUpdatesAction.browser.test.tsx` pass (3 of 3 tests). Lint, formatting and the apps/web typecheck are clean. **ProviderUpdatesAction.** Upstream's pingdotgg#23183 added a test whose computer fixtures have no `entry` field. The fork's `machine-updates` seam skips switched-off computers by reading `environment.entry.enabled`, so the test crashed with "Cannot read properties of undefined". I changed that seam line to `environment.entry?.enabled === false`. Switched-off computers are still skipped, and the `fork: machine-updates` marker and its seams.tsv count are unchanged. **DiffPanel.** Upstream added a test that replaces several app modules with partial stand-ins. One of them provides only `useThreadRouteRef`, and the one for `~/state/entities` provides only the four functions DiffPanel uses. The fork's diff-header button needed functions those stand-ins don't have, so the test file failed to load. Its import chain went through `toast.tsx` (which needs `useThreadRouteTarget`) and also needed `useServerConfigs` and `readThreadShell` directly. The fix is all in fork files: - The button now lives in a new file, `apps/web/src/fork/code-graph/DiffImpactButton.tsx`. - `diffHeaderAction.tsx` loads it lazily, and only when there is a thread and at least one changed file. - The diff panel's static imports from the fork side are now just three small modules. - The L26 TECHNICAL.md line about this file now explains the lazy load. The one cost is that the impact button loads a small extra chunk the first time it would appear. The seam check also lists missing `brand` rows for `SettingsScheduledTasksRouteScreen.tsx` and `ScheduledTasksSettings.tsx`. That is another check's failure, so I left it alone.
… (Checking that every fork seam survived) Failing check: Checking that every fork seam survived The seam check failed because two files had `fork: brand` markers that had no rows in `docs/fork/seams.tsv`: `apps/web/src/components/settings/ScheduledTasksSettings.tsx` and `apps/mobile/src/features/settings/SettingsScheduledTasksRouteScreen.tsx`. The markers came from an earlier repair in this round (83b6e26, the "new T3 Code product text" fix). Upstream added a new message about one-time scheduled tasks on older servers in both clients, and that repair changed "T3 Code" to "Loom" and marked each line `fork: brand`. It didn't add the matching seam rows, so the seam check flagged them as unregistered markers. I added one row per file to `docs/fork/seams.tsv`, `brand` with a minimum of 1, next to each file's existing rows. No upstream code or other files changed. `check_seams` now passes (exit 0, no output).
… (Checking code formatting) Failing check: Checking code formatting The "Checking code formatting" check failed on one file, `apps/server/src/fork/vault/RecoveryArchive.test.ts`. The rest of the 3778 files were already correctly formatted. The bad formatting came from an earlier repair commit in this merge (609deb6, the round 1 fix for "Running the fork's tests in apps/server"). It left a multi-line `cli(f.root, [ ... ])` call wrapped differently from what the formatter wants. The file is fork-owned (it's under `*/fork/*`), so the rules allow fixing it, and no upstream code is involved. I ran `vp fmt` on that one file. The only change is that the argument array of the `seal` CLI call is now wrapped the formatter's way; the test logic is the same. `vp fmt --check` on the file now reports "All matched files use the correct format."
Both findings were real, and I fixed both. Nothing is committed.
**1. Mobile "Queued messages run after it." on a stopped run.** Confirmed. The merge kept upstream's new rule in `ThreadRouteScreen.tsx`, which passes `onResumeQueue` whenever the run can be resumed, not only when the queue is held. `ThreadRecoveryNotice.tsx` is a fork-only file (upstream doesn't have it), and it still worked out "queue held" from `onResumeQueue !== null`. So an interrupted run with nothing queued showed the queue sentence.
- The notice now takes its own `queueHeld` prop. `ThreadRouteScreen.tsx` passes it `composer.queueHeld`, and `ThreadDetailScreen.tsx` declares the prop and hands it on. Each new line in those two files is marked `fork: failed-run-retry`.
- Typecheck then flagged `NewTaskDraftScreen.tsx`, which also renders `ThreadDetailScreen`, so it now passes `queueHeld={false}` with the same marker.
- `seams.tsv` counts: `ThreadDetailScreen` 5→7, `ThreadRouteScreen` 6→7, `NewTaskDraftScreen` 2→3.
- In `ThreadRecoveryNotice.test.tsx`, the existing tests now set `queueHeld` explicitly where they rely on a held queue. I added one test that the queue sentence only appears while the queue is held; it fails on the old logic and passes now.
On the two Resume buttons: I kept the notice's own Resume next to the upstream composer's and wrote that choice into `FORK.md`'s `failed-run-retry` paragraph. The notice's button stays visible when a draft hides the composer's button, and it continues the run in the thread's saved modes. This feature has no packet folder, so that paragraph is its record. If you'd rather mobile show a single control, the alternative is to drop the notice's Resume for interrupted runs.
**2. Outdated counts and wording in `FORK.md`.** All three held.
- Line 519: `ThreadSettlementService.test.ts` 6→3.
- The `failed-run-retry` paragraph: mobile `ThreadRouteScreen.tsx` and `ThreadDetailScreen.tsx` are now 7 each, after this fix.
- The `ProviderUpdatesAction.tsx` row now reads "Limit provider updates to the chosen computers and skip switched-off ones." That matches the fork's change, which only filters the chosen computers and skips switched-off ones.
I left one extra stale count alone because the review didn't raise it and `main` already had it: that paragraph says `MessagesTimeline.tsx` has 7 markers, while the file and `seams.tsv` have 4.
**Checks run:**
- Mobile tests for the notice, `ThreadRouteScreen` and the recovery presentation: 22 of 22 pass.
- Mobile typecheck is clean.
- `vp lint` and `vp fmt --check` are clean on the changed paths.
- `check_seams` returns 0.
Owner
Author
|
Review of the unattended resolution: fail. Merging upstream v0.1.7-nightly.20261101.4016 conflicted in 32 files; one Claude session took each. Sessions also edited these files to carry a seam to its new home:
After review 1:
Earlier reviews failed; a session fixed their findings:
Verdict: fail. There is one behaviour bug on mobile, plus some outdated counts in Findings1. Mobile: a stopped run wrongly says "Queued messages run after it."
2.
Confirmed sound
VERDICT: fail |
… (Running the fork's tests in apps/mobile) Failing check: Running the fork's tests in apps/mobile I fixed the failing check. One mobile test suite, `apps/mobile/src/features/settings/SettingsServerControlsRouteScreen.test.tsx`, was crashing before any of its tests ran, with `SyntaxError: Unexpected token 'typeof'`. The other 140 files and all 2,471 tests passed. **Why:** the test loads `SettingsProviderAccountsRouteScreen.tsx`, which pulls in the fork's `fork/vault/ProviderCredentials` component through its `vault-provider-credentials` seam. The test mocked every other child component of that screen but not this one. So it loaded the real component, which goes through `components/ui/Sheet` into native React Native code that the test environment can't read. Upstream's code didn't cause this: none of those files changed in the merge, and the suite fails the same way with the pre-merge `apps/mobile` and `packages/client-runtime` put back. The gap appears to date from the fork commit `a0299d5492` (Vault provider credentials), and this merge is the first time the check caught it. Mocking only `ProviderCredentials` is enough to make the suite pass. **What I changed:** - **The test:** I added one mock line for `../../fork/vault/ProviderCredentials` (rendering nothing), marked `// fork: vault-provider-credentials`. It sits next to the existing `github-host-setup` mock, and no assertions changed. - **`docs/fork/seams.tsv`:** I added a row recording that seam in the test file. The suite now passes (4/4), `check_seams` exits 0, and lint on the test file reports no errors. Nothing is committed.
… (Running the fork's tests in apps/web) Failing check: Running the fork's tests in apps/web I fixed the one failing test in "Running the fork's tests in apps/web". The rest of that run passed: 3,203 tests passed and one was skipped. **What failed:** the test "conversation palette ergonomics > keeps the title and source options across settings drill-down and Back until Start" in `apps/web/src/fork/project-library/NewConversationDialog.launch.test.tsx`. It looks for a button called "Conversation title" on the dialog's start page. That button wasn't there, so the test crashed reading `.props` of `undefined` at line 1372. **Why:** upstream didn't cause this. Loom's own commit f6a1df6 ("fix(web): simplify project creation and Computers settings"), which is already on main, changed the dialog in `ConversationPalette.tsx`. It moved Workspace, Model options, Permissions, Mode, Repositories, Conversation title and Context into a new "More options" page and didn't update this test. The upstream merge didn't touch the dialog or the test. In the dialog, both Back and choosing an option go back to the start page, so you now need "More options" each time you want Permissions or Conversation title. **What I changed:** I added `await choose("More options")` to the test before each visit to "Conversation title" and "Permissions", three new lines in total. The test still checks the same things: the title, permissions, mode and model all carry through settings changes and Back until Start. I touched no upstream code or seams; the test file is fork-owned. The whole test file now passes (50/50), and lint on it is clean.
The review's one finding holds, and I've fixed it. Commit `e6153d97b6` changed four upstream test files with no `fork:` markers and no `seams.tsv` rows, so the seam check couldn't see those edits. The merge itself had no other problems. All six steps below changed only markers, one test assertion, `seams.tsv` and FORK.md. **`ThreadMarkdownImage.test.tsx` (mobile):** the mock of `fork/cloudflare-access/mobileAccess` exists because the fork's `assetImageSource.ts` imports that module. I put a `// fork: cloudflare-access` line above the mock, written the same way as the matching mock in `state/assets.test.ts`, and added a `seams.tsv` row. **`ThreadFork.execution.test.ts`:** the fork's `refuseDeleted` guard rejects first with "Thread … is deleted.", so upstream's `"was deleted"` check can't match. The repair had loosened it to just `"deleted"`. I tightened it to `"is deleted"`, which matches the fork's exact message, marked the line `// fork: deleted-history` and added a row. **`externalLauncher.test.ts` and `UsageService.test.ts`:** neither change is caused by fork code. One keeps the Mac's real VS Code installs out of the Insiders test; the other stops the test failing now that today is past its fixture dates. Upstream's own `main` already pins that clock the same way, but only after this nightly. Dropping the changes would leave the date-dependent test failing until that arrives, so I kept them under a new `hermetic-tests` slug: 3 marker lines in `externalLauncher.test.ts` and 2 in `UsageService.test.ts`, each with a `seams.tsv` row. The existing `retention-defaults` row is unchanged. **FORK.md:** I added four rows to the Packet seams table and a short `hermetic-tests` paragraph that says to drop the `UsageService.test.ts` seam once upstream's clock fix lands. The formatter widened the table's File column by one character to fit the longest new path, so every row in that table shows as changed. That's spacing only, in a fork-owned file. **Checks:** - The three server test files pass (82 tests) and the mobile test passes (2 tests). - Lint is clean on the four test files. - FORK.md formatting check passes. - `check_seams` exits 0. Nothing is committed.
Owner
Author
|
Review of the unattended resolution: pass.
After review 1: Earlier reviews failed; a session fixed their findings:
No finding survived verification, so this branch can merge into main without a person. I covered both merges, Upstream changes kept
Loom changes kept or dropped for a valid reason
Unrelated changes: none. Each repair fixes the check it names. The FORK.md table reflow is the formatter widening columns for the new, longer rows. The lazy-loaded Three small points that don't block the merge:
VERDICT: pass |
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
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.
Merges upstream T3 Code v0.1.7-nightly.20261101.4016 into Loom. The seams in
docs/fork/seams.tsvsurvived, the fork's tests and upstream's tests beside the files it changes pass, every package typechecks, and the built web client boots.Land it from Terminal on your Mac (not inside Loom, which the install quits) with
scripts/fork/loom.sh land: it merges with a merge commit, tagsloom-v0.1.7-nightly.20261101.4016, publishes the next Loom release and installs it. Never squash or rebase this pull request; the fork needs upstream's history for the next merge.