Skip to content

Retire dead srv3 install/reconcile subgraph (shell→dag wave-2 §5.D) - #7233

Merged
briansrls merged 4 commits into
mainfrom
session/swift-pike-320-srv3-retirement
Jul 25, 2026
Merged

briansrls merged 4 commits into
mainfrom
session/swift-pike-320-srv3-retirement

Conversation

@gunbai-bot

@gunbai-bot gunbai-bot Bot commented Jul 25, 2026 •

Copy link
Copy Markdown
Contributor

Summary

Deletes the dead srv3 install/reconcile cluster identified in docs/plans/shell-to-dag-residual-census-and-arc-completion.md §5.D — unreachable from any gunbc subcommand, CLI, or CI path after srv3 install completed and the srv* cluster wound down.

  • Remove four orphaned HostEffect variants: Srv3InstallDiagnosticObserve, Srv3InstallReconcileObserve, DurableApprovalGrantRecord, Srv3SolConsoleCapture
  • Delete the scaffold subgraph (srv3_os_install_reconcile*, srv3_host_effect_script, srv3_install_diagnostic_checklist, srv3_install_diagnostic_observe_script, srv3_sol_console_capture/intent) and update every total match on HostEffect
  • Drop srv3-specific retained_srvn realization arms (realize_srv3_install_diagnostic_observe, srv3_host_effect_script_for routing)
  • Delete or repoint srv3 witness tests; remove reconcile rows from non_fold_residue and the diagnostic-checklist exception from meta_exec_confinement / wall_residue_live_test
  • Re-home surviving constants (srv3_nbd_proxy_lease_key → srv3_nbd_proxy_serve_intent; boot-once grant path/sleep → srv3_boot_once_cd)

Test plan

  • gunbc compile --target dag over dag/ + src/v2/ — green (0 diagnostics)
  • CI floor (compile-clean + affected-set witnesses)

@gunbai-bot
gunbai-bot Bot marked this pull request as ready for review July 25, 2026 18:00
@gunbai-bot gunbai-bot Bot changed the title srv3 install/reconcile subgraph retirement (shell→dag wave-2, operator GO 2026-07-25): delete the DEAD subgraph + its orphaned HostEffect variants + retained_srvn wraps + srv3 witnesses TOGETHER, one PR. READ docs/plans/shell-to-dag-residual-census-and-arc-completion.md §5.D 'srv3 install/reconcile Retire dead srv3 install/reconcile subgraph (shell→dag wave-2 §5.D) Jul 25, 2026
@gunbai-bot

gunbai-bot Bot commented Jul 25, 2026

Copy link
Copy Markdown
Contributor Author

Addressed review 42756: dropped the stale Srv3DiagnosticObserveOnHost / realize_srv3_install_diagnostic_observe arm from the host_effect_apply_gated dispatch (the first site at host_effect_apply was updated in the original retirement pass; this mirrored copy was missed).

Fixed in 1d27c7ba89. gunbc compile --target dag green on that commit.

— sent from swift-pike-320

@briansrls
briansrls merged commit b438d8a into main Jul 25, 2026
5 checks passed
@briansrls
briansrls deleted the session/swift-pike-320-srv3-retirement branch July 25, 2026 19:04
briansrls added a commit that referenced this pull request Jul 25, 2026
…y; meta-exec roster 3->0 + per-PR emptiness wall; census true-up

- host_runner_memory_cap_verify probed_at: retained_runtime shell.Exec.Run(date -Iseconds)
  -> typed extdeps.clock Clock.Now via gunbc.clock_read (host_converge_slice1 precedent).
  extdeps.shell import, retained_shell_script bridge and the transport-script scaffold all deleted.
  Refusal decision refactored to runner_memory_cap_verify_result_from_live_read so the
  read-absent arm is executable hermetically; RED control proves a widen fails.
- meta_exec_confinement_exception_roster 3 -> 0: tools/review.dag and ci_deploy_target_host.dag
  verified stale (0 shell.Exec); host_runner_memory_cap_verify dissolved above.
- Per-PR enforcement is a construction wall (empty roster => stale_count 0, no scan);
  live receipt measured 13.5s cold vs a 5s fast-lane budget, homed in the long lane
  with a non-degeneracy control.
- Census section 4/5 + ledger trued up to main (#7184/#7192/#7193/#7194/#7215/#7231/#7233).

Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
briansrls added a commit that referenced this pull request Jul 25, 2026
…; meta-exec roster 3→0 + per-PR emptiness wall (#7241)

* WIP: shell→dag bucket A cleanup (one PR): (1) host_runner_memory_cap_verify 2

* WIP: shell→dag bucket A cleanup (one PR): (1) host_runner_memory_cap_verify 2

* WIP: shell→dag bucket A cleanup (one PR): (1) host_runner_memory_cap_verify 2

* WIP: shell→dag bucket A cleanup (one PR): (1) host_runner_memory_cap_verify 2

* WIP: shell→dag bucket A cleanup (one PR): (1) host_runner_memory_cap_verify 2

* shell->dag bucket A: last shell.Exec off host_runner_memory_cap_verify; meta-exec roster 3->0 + per-PR emptiness wall; census true-up

- host_runner_memory_cap_verify probed_at: retained_runtime shell.Exec.Run(date -Iseconds)
  -> typed extdeps.clock Clock.Now via gunbc.clock_read (host_converge_slice1 precedent).
  extdeps.shell import, retained_shell_script bridge and the transport-script scaffold all deleted.
  Refusal decision refactored to runner_memory_cap_verify_result_from_live_read so the
  read-absent arm is executable hermetically; RED control proves a widen fails.
- meta_exec_confinement_exception_roster 3 -> 0: tools/review.dag and ci_deploy_target_host.dag
  verified stale (0 shell.Exec); host_runner_memory_cap_verify dissolved above.
- Per-PR enforcement is a construction wall (empty roster => stale_count 0, no scan);
  live receipt measured 13.5s cold vs a 5s fast-lane budget, homed in the long lane
  with a non-degeneracy control.
- Census section 4/5 + ledger trued up to main (#7184/#7192/#7193/#7194/#7215/#7231/#7233).

Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>

---------

Co-authored-by: Brian Searls <briansrls@gunb.ai>
Co-authored-by: Claude Opus 5 (1M context) <noreply@anthropic.com>
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

1 participant