Repository navigation
Offer a review's fix only where it can run, and log a refusal once - #197
Merged
madskronborg merged 14 commits intoOct 26, 2026
Merged
madskronborg merged 14 commits into
madskronborg merged 14 commits into
Conversation
added 7 commits
October 23, 2026 06:28
A review offered Fix findings for a target with no branch to fix, refused every answer the same way, and replays wrote the refusal again each time. The review now asks the host before its first menu whether the reviewed branch can be established, by the checkout's own resolution and cutting nothing, and withholds what cannot run, naming why. A fix refused when it starts stands for the rest of the Run, and each refusal is recorded inside the Activity that met it. Closes #185.
The host writes a check's running-checks record only after spawning it, but the tests ended or restarted the host as soon as the child noted its pid. On a loaded machine the record was not there yet, so the next host found the check by its markers alone and the record count was zero. Both tests failed on master the same way.
…r-vm-mk-run-e4868a8c-now-claude-is-3a3fzk
The gate's check failed at 064b000 only because this worktree had no desktop/node_modules, which master's typecheck now reads. Installing them changes nothing git tracks, and Collie does not rerun a Run's check on a tree it already failed on.
added 5 commits
October 23, 2026 15:01
…r-vm-mk-run-e4868a8c-now-claude-is-3a3fzk # Conflicts: # src/processes.ts
A reader that found the file between its creation and its write read an empty reason, which failed the idle-budget test in CI and would show the board a Run parked for nothing.
Git 2.47+ (CI has 2.55) detaches the maintenance a commit starts, so it can still be writing into a test's repository after a fast test removed it, and the suite reports the directory as left behind.
…r-vm-mk-run-e4868a8c-now-claude-is-3a3fzk
#194 made the host's placement need ForgeHttp, which reviewedBranch now gets from the host's services as checkoutReviewed does, and gave worldWith an adapters object.
This was referenced Oct 23, 2026
added 2 commits
October 26, 2026 07:49
…r-vm-mk-run-e4868a8c-now-claude-is-3a3fzk # Conflicts: # test/host.test.ts
…r-vm-mk-run-e4868a8c-now-claude-is-3a3fzk
This was referenced Oct 26, 2026
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.
A
reviewRun used to offer Fix findings, and Fix findings in a full implement run, even when the reviewed branch could not be established, for example from a checkout with no remote. It then refused every answer and asked the same question again, and each replay wrote the refusal into the log again. Now, before its first menu, the review asks the host whether the reviewed branch can be established. The host answers with the same resolution it uses for the checkout, and cuts nothing. Where the answer is no, the review withholds both fix options and the question says why. The one exception: Fix findings stays on offer when a Run in this one's lineage may have a live implementer to hand the findings to. A fix refused when it starts stays refused for the rest of the Run, and each refusal is logged once. The amendment to ADR-0054 records the rules, and the workflows and SDK docs describe them.This is a recovery of run-e4868a8c: its four commits are carried over unchanged. On top of them,
test/host.test.tsnow waits for the host to record a held check before ending or restarting the host. That race failed two host tests on master too. Squash-merge, because three of the commits arefixup!s.Closes #185.
Verified on the branch as it stands: only
typecheckpassed on this tree (by agent).host-check-restart,lint,formatandreview-scenariospassed by agent, but on an earlier tree.checkfailed (by collie); see below.Verifications this run is held to
/home/mk/.local/share/mise/shims/bun run check(in/home/mk/.herdr/worktrees/.collie/madskronborg/recover-vm-mk-run-e4868a8c-now-claude-is-3a3fzk)The two gate fixes after the last review were checked by their own tests and dispositions, not by a reviewer. Gate fix 1 changed no code: it installed
desktop/node_modules, which master's typecheck now reads, and pushed Collie's merge of master (064b000). Gate fix 2 is the empty commit fb2cc25, made so the gate would run again with those dependencies installed.Not settled by the run