carlos / platform Public
Review brief: region HA plan 1a (foundations), round 2
Read CODEX-PLAN-REGION-HA-1A-BRIEF-1.md first: its "Fixed: do not re-argue" and "What to check" sections still apply, to the plan at the commit named in the prompt.
Round 2
This revision answers your round 1 (CODEX-PLAN-REGION-HA-1A-R1.md); the plan ends with a dispositions table. Since round 1 the plan's code was dry-run by its author: every new package and both patch tasks were written into the tree, built, cross-built for darwin and windows and tested, then removed. The code blocks for internal/boxid, internal/hostfiles, internal/termfence, internal/activator/installed* and internal/console/boxchoice* are those files verbatim. So a "would not compile" finding should be rare: if you raise one, quote the line and say exactly what is undefined.
First give a disposition table for your nineteen round 1 findings: resolved, partly resolved (say what is left), or not resolved. Then report only new findings, including any the fixes introduce. In particular look at:
readStable,syncRoot,Restore,removeInside,writeInsideand their use ofos.Root(Go 1.26): any remaining way out of the root, any way to block, any path where a tombstone or a write is skipped that should not be, or the reverse.- The run loop's degraded, repair and dead-watcher handling, and whether each test can pass with its mechanism disabled.
edgesFileWrite'srefreshand the seeded gate.foldHomes's tie reporting anddialBeatsagainst every existing home-map rule.- The SQS queue policy's Deny: does it block the EventBridge rule's own delivery, or anything else that must work?
- The plan's claim that an alert on two unstamped home entries would be wrong today. Check it against
edgesFilePassandeverywhereClass.
Two things the author could not run and says so: the six real-inotify tests in watch_linux_test.go (read watch_linux.go with extra care) and the IMDS signature check.
Do not re-raise anything you mark resolved.
Output
As round 1. End with a verdict: ready to build, or not ready.