Repository navigation
Conversation
There was a problem hiding this comment.
Copilot was unable to review this pull request because the user who requested the review has reached their quota limit.
An organization owner can choose the repository organization as the billing source for reviews requested by members with a paid Copilot license. Learn about Copilot code review billing.
6813f21 to
767e402
Compare
767e402 to
3b01497
Compare
|
merger: held out of batch 44. At head 3b01497, 7 tests fail that pass on main 26c083b: test_commander_problem_orders::test_check_repo_resolves_once_for_send_and_correction[.|absolute|agent-comms] (GitCommandFailed origin/HEAD), test_order_check_replaces_a_wrong_check_so_done_runs_the_corrected_one, test_order_run_carry_gives_a_pre_p511_order_its_declared_run, test_connector.py::test_connector_lists_every_tool (RecursionError at src/agent_comms/declared_family.py:164 -> /mcp 500), test_standing_rules.py:202 wheel_spin ('; distance 2, projected done ' missing). Run: timeout 300 pytest on these node ids. |
3b01497 to
dd59630
Compare
…rs at the boundary
455846f to
0b3d9cb
Compare
|
Sent back by desktop merger mB: 4 conflict hunks in commander/order.py and todos.py against main. Main's overdue alarm returns one joined OVERDUE line using t.deadline.clock(); this branch returns one line per order using the deleted hhmm(). Main deleted hhmm, so do not bring it back. Merge current main, keep t.deadline.clock(), and pick ONE alarm shape (joined or per-order). Delete the other, along with its tests. Due 00:40. |
|
Not merged: conflicts with main in commander/order.py and todos.py. Rebase. |
|
Replaced by merged #2153 with P271, since both change the Todo/PR progress owner. Actual CLI preserves earlier orders and closes a replacement order using its existing real merged PR. Initial batch 200 passed/21 failed, no repeat; final source/tools ratchets 0/0. |
P283: retain problem progress across orders and reused checkouts
Ready head:
3b014971e9e27ea9a7df9a1ae37789aa7c5ac1be.Rebased onto requested main
d302838978e45dffc307bd1c6c26e7583d26f6fe; no conflicts. All source is pushed and the checkout is clean.Owners and replaced decisions
CheckedProblemOrder.precedingretains the full preceding order. Assignment, reissue and link transitions preserve the chain.Todo.held_order_idsderives identities from that chain: no stored ID list or side table.PullRequest.belongs_toconsumes the Todo's problem declaration or held order identities.PullRequests.listedno longer filters by only the current order. Body references are decoded once; titles and guessed branch names are not used. Core PR creation emits its explicit problem declaration.StorageRule.checkout_pathsowns candidate discovery and is shared with storage accounting.WorkBranches.worktreessupplies actual branch/checkouts from Git metadata.PresentCheckoutuses existing Git refs, excluding unborn branch names, and observes owned local branches plus retained origin refs. The named problem checkout is no longer the only progress source.--git-dir=<worktree .git file>override is deleted.P270's inherited merged-order lifecycle and current Commander-context methods remain intact. No deleted module-level commander functions restored.
Final verification on this head
Both ratchets ran as separate commands, using the shared QA Python and existing audit scripts:
Both reports pin base
d302838978e45dffc307bd1c6c26e7583d26f6feand head3b014971e9e27ea9a7df9a1ae37789aa7c5ac1be. No checker or baseline changes.Background test run, outer
timeout 300,-X faulthandler,PYTHONPATH=src, existing shared QA interpreter, owned scratch basetemp:The same suites on main
d302838978e45dffc307bd1c6c26e7583d26f6fereturned 66 passed, 14 failed. Every candidate failure occurs on main: new failures []. The new family test passes through a reused worktree, a real replacement order and persisted history, old-order/problem-body PR linkage, a real push, checkout removal and origin-ref progress. The warning is the shared pytest environment's unknowntimeoutconfiguration; the outer timeout was active.Actual configured live-bus observation
Read the real configured bus through this head's
OrderProgress.read, the owner consumed by Core's progress readers. No problem mutation, provider input or manual closure was requested:That is not
no commit since the order. PR #1965's body declares P241, so the reissued order no longer hides its merge even though previously discarded order history cannot be reconstructed.Earlier actual overdue output also showed P215 PR #1951 and P275 PR #2051. That historical output remains retained, not presented as a new-head CLI pass. A later ordinary overdue attempt failed with TodoConflict in
Commander.first_calls; that path records first-call events and is not read-only. Earlier writes may have committed. No successful complete overdue/heartbeat pass or installation is claimed here.Source evidence and retained records
Before evidence was inherited from the explicit pi-cn-phase2 takeover; after evidence uses the existing audit Package/Repository parser across production, tools and tests. The retained census names declarations, inheritance, imports and candidate consumers, with parse omissions reported. Dynamic receiver bindings remain ambiguous, not proved by the syntax census.
Current evidence:
/home/ts/.cache/agent-scratch/pi-cn-phase3/p283/rebase-4837/(separate ratchets, candidate/baseline tests, actual configured P241 read, AST output). The interrupted final inspection returned no normal tool result; retained complete reports and the AST footer were inspected afterwards, not replayed. The earlier cwd-recovery write was UNKNOWN and was not retried.Assumed: only explicit body problem declarations/headings or retained order references link a PR; pushed work without such a PR remains truthfully pushed with no linked PR. Origin refs are observed publication evidence. Discarded historical orders are not invented. Merger owns rehearsal, installation and installed service acceptance; the persistent goal remains blocked.
Agent-Comms-Problem: P283
Agent-Comms-Order: 1a45039659e5