Repository navigation
chore: sync upstream through e3949714e3 - #298
Conversation
Co-authored-by: Julius Marminge <julius0216@outlook.com>
… keep uploading (pingdotgg#37681) Co-authored-by: Claude Opus 5.5 <noreply@anthropic.com>
…gdotgg#37703) Co-authored-by: Julius Marminge <julius0216@outlook.com>
Co-authored-by: screen <screen@users.noreply.git.cafe> Co-authored-by: Devin <158243242+devin-ai-integration[bot]@users.noreply.github.com> Co-authored-by: Julius Marminge <julius0216@outlook.com>
…ead (pingdotgg#37735) Co-authored-by: Julius Marminge <julius0216@outlook.com>
…s as a silent success (pingdotgg#37738)
…ingdotgg#37733) Co-authored-by: Claude Opus 5.5 <noreply@anthropic.com>
…ingdotgg#37741) Co-authored-by: Usama Khan <usama@Usamas-MacBook-Pro.local> Co-authored-by: Julius Marminge <julius0216@outlook.com>
Co-authored-by: Claude Fable 5.2 <noreply@anthropic.com>
…heir project (pingdotgg#37682) Co-authored-by: Claude Opus 5.5 <noreply@anthropic.com>
…hread (pingdotgg#37725) Co-authored-by: Claude Opus 5.5 <noreply@anthropic.com>
… with web (pingdotgg#37743) Co-authored-by: Claude Opus 5.5 <noreply@anthropic.com>
…er area (pingdotgg#37508) Co-authored-by: Claude Opus 5.5 <noreply@anthropic.com>
…lds (pingdotgg#37755) Co-authored-by: Claude Opus 5.5 <noreply@anthropic.com>
…tes instead of running forever (pingdotgg#37728)
…s an error (pingdotgg#37747) Co-authored-by: Martin Bergo <14320143+mInrOz@users.noreply.github.com> Co-authored-by: Julius Marminge <julius0216@outlook.com>
…ingdotgg#37745) Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com> Co-authored-by: Julius Marminge <julius0216@outlook.com>
…ingdotgg#38216) Co-authored-by: Hannes Lenz <hlenz@esic-sarida.cloud> Co-authored-by: Julius Marminge <julius0216@outlook.com>
…gdotgg#38235) Co-authored-by: Grok 4.7 <noreply@x.ai> Co-authored-by: Julius Marminge <julius0216@outlook.com>
…swers (pingdotgg#38237) Co-authored-by: Grok 4.7 <noreply@x.ai> Co-authored-by: Julius Marminge <julius0216@outlook.com>
…#38238) Co-authored-by: Simone <185146821+Lucenx9@users.noreply.github.com> Co-authored-by: Julius Marminge <julius0216@outlook.com>
…me (pingdotgg#38249) Co-authored-by: Grok 4.7 <noreply@x.ai> Co-authored-by: Julius Marminge <julius0216@outlook.com>
pingdotgg#38250) Co-authored-by: Julius Marminge <julius0216@outlook.com>
…ingdotgg#38276) Co-authored-by: Grok 4.7 <noreply@x.ai> Co-authored-by: Julius Marminge <julius0216@outlook.com>
Co-authored-by: Grok 4.7 <noreply@x.ai> Co-authored-by: Julius Marminge <julius0216@outlook.com>
…n recovery (pingdotgg#38241) Co-authored-by: Grok 4.7 <noreply@x.ai> Co-authored-by: Julius Marminge <julius0216@outlook.com>
Co-authored-by: Julius Marminge <julius0216@outlook.com>
Merge upstream pingdotgg/t3code through e394971 (225 commits since fecd713). Conflicts resolved keeping fork behavior: CoreWsRpcGroup split with issue RPCs, readThreadPage takes runCount, t3_thread_launch keeps viaFleet and Home logic with upstream's launch changes, mobile cloud-wake status and box-aware environment pick, outbox box demand in upstream's fiber rewrite, box-wake states set refusedRouteIds. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
|
This PR changes the test suite by +300 test cases (browser +32, unit +176, server +92). New browser test files: Keep tests that would fail for a real reason, and test logic below the UI where it can be. See Testing. |
Upstream's new i18n catalog check (pingdotgg#37954) requires every source string in en.json and each locale. Home names the fork's feature, so it stays Home. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Thread transfer impact✅ Thread transfer remains within every enforced ceiling.
Other scenariosAll 64 scenario measurements match All 64 scenario measurements
Baseline: Scenario and decoded snapshot size10 historical turns, 5 command tools per turn, 878.9 KiB retained MCP result per historical turn, and a 1.05 MiB retained result in the measured turn.
Updated in place by a trusted workflow. PR artifacts are strictly validated and never executed. |
… sends Upstream now stamps runId on a thread's background tasks (pingdotgg#38035) and keeps a failed turn's tasks on a live process until their result wakes the agent (pingdotgg#31840). The fork test expects the runId, and the shell's end now sends the task notification and the recorded wake turn with the wake worker running, as production does. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Why
Syncs the fork with upstream pingdotgg/t3code through e394971, upstream's newest main with green CI as of 2026-11-17 20:17Z. That is 225 commits on top of the last sync (#294, fecd713). Merge commit; land with a merge, never squash.
What changed
Notable upstream changes:
t3_thread_launchwith no workspace follows the worktree default (fix(server): t3_thread_launch without a workspace follows the worktree default pingdotgg/t3code#38261). Claude: a running thread forks natively (fix(server): fork a running Claude thread natively pingdotgg/t3code#37955), and a provider switch sent mid-turn is queued (fix(server): queue a provider switch sent while a Claude turn runs pingdotgg/t3code#38216).Conflict resolutions (18 files):
CoreWsRpcGroup/CoreServerWsRpcGroupsplit keeps upstream's issue RPCs, handlers and split prefixes. Both sides' settings additions are kept.readThreadPagetakes upstream's newrunCount. FleetService passes its full run count.t3_thread_launchkeepsviaFleetand the Home logic, plus upstream's retry-project check, itsexisting_worktreereplay lookup and its worktree default.ProviderTurnStartServicekeeps the fork'sinjectHistoryguard on uncertain delivery and adds upstream'snativeSessionLacksKeptTurns.RunExecutionServicekeeps the fork's fall-through, which already holds any non-continuing terminal while the thread probe reports work; it is a superset of upstream's failed-only hold.agent-context-increase).wakingtext. Plain reconnecting text moves to upstream's header. The draft's project stamp feeds the fork's environment pick, so a stamp on another chat's box is still steered off. The outbox keepsclaimFirstTurnCloudBoxandholdQueuedBoxDemandin upstream's fiber rewrite. AconnectionStatusLabelthat auto-merged twice intoThreadRouteContentwas renamed.refusedRouteIds.ClaudeBackgroundWorkAfterLimit.test.tsnow expects upstream'srunIdon background tasks, and it sends the task notification and recorded wake turn a limited CLI really sends. Upstream (fix(server): background tasks survive a failed Claude turn on a live process pingdotgg/t3code#31840) keeps a finished task listed until its result wakes the agent; adapter unchanged.Blast Radius
New upstream migrations: 097 creates the empty
issue_tracker_bindingstable, which is instant. 098 builds two partial expression indexes, over active rows oforchestration_v2_projection_threadsand overorchestration_v2_projection_provider_threads. The PC has 35,801 active threads and 19,381 provider threads (read-only count). Inferred cost is seconds to well under a minute, since it does not touch turn items. Still, expect a slower first start than a no-migration deploy.Verification
vp run typecheck: contracts, server, client-runtime, mobile, web.vp fmt --checkandvp linton all 768 files that differ from upstream.vp run knip:check.vp test run: client-runtime, mobile, contracts. Server: full suite run; the 14 files failing locally fail only on this Mac (tmpdir symlinks, APFS case, Chrome) and pass on CI's 8 server shards. Web: all 34 fork-touched test files (24 unit, 10 browser) plus SidebarDraftBlock.Not verified locally: the full web and desktop suites (left to CI), and a real-client pass of the mobile connection status and new-task flow.
Follow-ups: FleetService
threads.readstill loads every run; it could use upstream's limited loading from pingdotgg#37915.Claude Opus 5.5 did the sync in Claude Code.
🤖 Generated with Claude Code