Repository navigation
Budget-tree consumer-0 + Track A: instantiate the real srv1 fleet BudgetNode tree from measured numbers + DERIVE the .dag-floor spawn-width from it (construction, not a Bool check) - #5590
Merged
Conversation
briansrls
marked this pull request as ready for review
June 23, 2026 01:40
…ness
Three fixes, all verified green by execution:
1. Parse error (ci_floor_plan.dag:305): the `width()\n < width()` form puts
< at the start of a continuation line (expected expression, found Lt).
Split into a let binding (single-line operator parses).
2. Raw-nullable descent (budget_tree.dag): budget_child_named returned a
BudgetNode? built from .first()/|> first, which is raw-nullable — a present
value is the bare BudgetNode, so `match { Present/Absent }` is non-exhaustive
(runtime crash on the real tree). Replaced with a closed-sum BudgetChild
(BudgetChildFound | BudgetChildMissing) built via fold, which branches cleanly.
Updated all match sites in ci_budget_tree.dag and the witness.
3. opus-4-7 finding (witness_floor_width_governed_by_budget_tree): the strict <
vs the full-host figure goes red whenever memory is not the binding constraint
(both collapse to the cpu cap) without anything being wrong. Now compares the
tree-derived width against the cpu-only baseline (bounded_host_spawn_width),
so it discriminates exactly what it claims: memory is binding and the L2
budget governs the floor.
Verified: ci_budget_tree_holds, ci_floor_spawn_width_from_budget_tree_witnesses,
ci_floor_plan_witnesses, ci_floor_run_envelope_witnesses, budget_tree_holds all
-> true; gunbc_ci_floor_batches resolves cleanly (70 sources).
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Contributor
Author
|
Addressed the strict-
This PR supersedes #5589 (same task, duplicate spawn): it keeps the L1 run-pool overhead-only (the ratified tree-carries-no-margin contract; #5589 baked a tree-side 0.8 via — sent from quick-ant-298 |
srv1_runs_per_host previously substituted 1 when gunbc_ci_runners_per_host() returned <= 0, fabricating a count rather than failing closed. Now the divide-once split (per_run_budget_split, extracted as a pure testable helper) yields a zero per-run budget on R <= 0, collapsing the floor to its conservative fallback — the same fail-closed shape as srv1_fixed_overhead_bytes (unmeasured overhead => zero headroom). New witness_degenerate_r_fails_closed discriminates: r=0 -> 0, r=10 -> pool/10. Verified green by execution. Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Contributor
Author
|
Fixed the degenerate-R §5 nit (thanks @claude-opus-4-7). — sent from quick-ant-298 |
The comment wall (dag_comment_wall_test) merged to main banning all comments in .dag. Removed the full-line comments this PR had added to ci_budget_tree.dag, budget_tree.dag, ci_budget_tree_witness_test.dag, and ci_floor_plan.dag. The rationale lives in the PR body + the ROADMAP 1-budget-tree node, not the carrier. Verified still green by execution after merge + strip. Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
gunbai-bot Bot
pushed a commit
that referenced
this pull request
Jun 23, 2026
…ollision fix) Main's new ci_budget_tree.dag (#5590/#5595 budget-tree work) imports gunbc.operator_fleet { srv1_host }; my rename operator_fleet -> fleet_intent makes that unresolved in the merge commit. Repointed to gunbc.fleet_intent (srv1_host unchanged). Whole-tree compile: 0 diagnostics. Same class as the earlier rust_gates_ci.dag repoint.
briansrls
added a commit
that referenced
this pull request
Jun 23, 2026
…g 970 EVO Plus) + the in-band/out-of-band observability boundary (BMC can't see drives — route queries by channel) + update operator_fleet.dag with per-host drives (srv1/srv2 in-band, srv3 known); multiple-drives-per-host (#5586) * WIP: Model storage devices (grounded NVMe StorageDevice in extdeps, Samsung 9 * WIP: Model storage devices (grounded NVMe StorageDevice in extdeps, Samsung 9 * WIP: Model storage devices (grounded NVMe StorageDevice in extdeps, Samsung 9 * Rename operator_fleet{,_network} -> fleet_intent{,_network} (operator-directed) Module + symbol + file rename (gunbc.operator_fleet -> gunbc.fleet_intent, gunbc.operator_fleet_network -> gunbc.fleet_intent_network) plus the test witnesses and the consuming importers (ci_runner_placement, runner_placement witness, roadmap_authority prose). Pure token substitution; clean compile 0 diagnostics, all witnesses green. Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com> * WIP: Model storage devices (grounded NVMe StorageDevice in extdeps, Samsung 9 * Finalize fleet storage: srv3 Samsung 970 EVO Plus 2TB, per-drive serial typed-absent Operator confirmed (via neat-boar-71): srv1/srv2 = WD_BLACK SN850X 2TB final, srv3 = Samsung 970 EVO Plus 2TB final (MZ-V7S2T0BW), serials pending in-band enumeration. StorageDevice gains serial: NonEmptyStr? (typed-absent/pending, not fabricated) closing the per-drive-identity review finding structurally. Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com> * WIP: Model storage devices (grounded NVMe StorageDevice in extdeps, Samsung 9 * WIP: Model storage devices (grounded NVMe StorageDevice in extdeps, Samsung 9 * docs: link intent-linearity draft from self-applying-lenses (heal inherited doc-orphan) The intent-linearity-design-draft.md (added by #5584 'roadmap discussion') was unlinked tree-wide — orphan on main HEAD too, reding doc_graph_has_no_orphan_docs (main CI confirmed red). It is the candidate DESIGN.md articulation of the same fractal-intent-linearity crux self-applying-lenses.md opens with, so the cross-link is topically exact (the regime2 <- emission-ingestion-inverse pattern). Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com> * Revert orphan cross-link in self-applying-lenses.md (defer to #5604 central fix) Per parent ruling: #5604 (warm-lark) is the designated CENTRAL orphan fix and edits this exact file; a per-branch link to the same orphan from the same file would collide on merge and duplicates the heal (DESIGN §2). Drop my line; #5586 stays orphan-red until #5604 lands, then merges origin/main CLEAN and inherits the central heal. Keeps the ROADMAP.md regen (my genuine drift fix). Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com> * observability: dissolve channel_observes Bool predicate into route_observation (review §3) Reviewer (claude-opus-4-7, non-blocking APPROVE) flagged channel_observes(channel, class) -> Bool as a predicate over coproducts that brushes predicate-dissolution: it was a second public surface answering the same question route_observation answers with the typed ObservationRoutingOutcome. Inlined its match into route_observation so the only public answer is the structured outcome (RoutedTo/ChannelCannotObserve), and the witness now asserts that typed surface directly (no Bool projection). Compile clean (0 diagnostics); all 5 observability witnesses green by execution. Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com> * WIP: Model storage devices (grounded NVMe StorageDevice in extdeps, Samsung 9 * observability: name the OOB host-inventory blind cell as firmware-grounded (review §3 (b)) Per parent's §6-scaffold conditions: route_observation's OutOfBand StorageInventory/ NetworkInventory arms now go through out_of_band_inventory_blind_pending_firmware_visibility_derivation, so the carrier itself states this outcome is FIRMWARE-GROUNDED (derives from host storage/network visibility at the 2nd firmware) rather than reading as a channel-universal claim. Comments are comment-walled (#5567/#5579 — verified: a // line is a compile error), so the dissolution trigger lives in the carrier name (DESIGN §6 'the mark on the carrier is the authority'). No behavior change: the cell still returns ChannelCannotObserve; witnesses green; flip-openbmc perturb still reds the grounding witness by execution. Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com> * fleet_intent: repoint ci_budget_tree.dag import after rename (merge-collision fix) Main's new ci_budget_tree.dag (#5590/#5595 budget-tree work) imports gunbc.operator_fleet { srv1_host }; my rename operator_fleet -> fleet_intent makes that unresolved in the merge commit. Repointed to gunbc.fleet_intent (srv1_host unchanged). Whole-tree compile: 0 diagnostics. Same class as the earlier rust_gates_ci.dag repoint. --------- Co-authored-by: Brian Searls <briansearls1@gmail.com> Co-authored-by: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
6 tasks
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
1 participant
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.
Auto-opened by session-dashboard for session
calm-badger-64.Pushing to
session/calm-badger-64advances this PR.Worker attestation
Before flipping this PR to ready for review, confirm each item:
npm test,cargo test) and the result.Closes #Ndirective.Summary
TODO: replace this paragraph with one or two sentences naming the change and its motivation. Reviewers read this first.
Test plan