Skip to content

Four-compile redundancy collapse (#4091 §1.2): collapse ci_v4 four 332-source closure compiles to one path (~14m savings) - #4175

Closed
briansrls wants to merge 2 commits into
mainfrom
session/cool-fox-718
Closed

briansrls wants to merge 2 commits into
mainfrom
session/cool-fox-718

Conversation

@briansrls

@briansrls briansrls commented Jun 1, 2026 •

Copy link
Copy Markdown
Contributor

Four-compile redundancy collapse (#4091 §1.2) — model-only slice

Sequenced behind #4171 (primary carrier). Intentionally DRAFT per parent's dedup ruling — do not flip ready / merge before #4171 lands. This PR is the complementary modeling-hygiene slice (collapse of the modeled ci_pipeline shadow jobs in src/v4/workflow/ci.dag); #4171 carries the real required-path multi-target-emit win.

What this changes

Three modeled jobs in ci_pipeline each independently re-ran a full src/v4 332-source closure compile while only declaring a bare needs: [v2_compile_src_v4] ordering dep. This turns those into real artifact-consumption edges by adding ci_upsert_upstream_job_input(job: v2_compile_src_v4) to each job's CiUpsertStep inputs — mirroring the established v4_t15 / testclaim_corpus_eval pattern (ci.dag:807, "M1 rust emit is upstream compile authority"):

  • m1_rust_emit_probe_execution
  • lens_ci_registry_execution
  • phase1_nat_semiring_rung_gate_execution

Each consumes v2_compile_src_v4's target-independent resolved closure (front-end shared once) and re-emits its own target from it.

Correctness (manager-requested confirmations)

  1. Front-end-only collapse — emit + strictness unchanged. Only the redundant front-end recompile (parse/lower/infer over the 332-source closure) is deduplicated. Each job still re-emits its own target (M1 still --target rust at full V4_M1_RUST_EMIT_PROBE_STRICT=1); no emit target is collapsed into another. Coverage is identical, just deduplicated.
  2. Affected-set invalidation unchanged per job. Each job keeps its existing src/v4/** file-set input (and m1 its scripts/v4-m1* / Cargo.{toml,lock}); the upstream edge is added, not substituted. What triggers a rerun is neither widened nor narrowed.

Discipline

  • No new ci.dag carrier (existing ci_upsert_upstream_job_input only) — no DFS-arbiter sign-off needed.
  • No substrate / generated-CLI / emit-pipeline touch (that is shape B = Four-compile redundancy collapse (#4091 §1.2): single 332-source closure path (~14m) #4171's lane).
  • bootstrap-dag split out: v2_bootstrap_smoke_execution's collapse is not a clean modeled edge — the workflow_local_bootstrap_dag_upstream carveout (ci.dag:336) has dissolution_target: ModelMissingSubstrate { what: v4_bootstrap_dag_emit_ci_job } (= substrate). Deliberately untouched; tracked as a follow-up.

Receipt

New integration test v4_workflow_ci_four_compile_collapse_jobs_consume_v2_compile_src_v4_closure block-scopes the assertion that each of the three jobs carries the upstream edge.

Post-#4171 rebase note

After #4171 lands, rebase onto main and drop any model/test edits #4171 already covers; keep only the residue (the shadow-pipeline upstream edges #4171 doesn't touch).

@briansrls

Copy link
Copy Markdown
Contributor Author

Closing per CI Manager + smart-newt-797 agreement: model-only slice sequenced behind #4171 (held on #4179/M1-main). Substantial rework expected at rebase — will open a fresh post-#4171 PR from node adhoc-11b7a2da-46a / cool-fox for ci_pipeline residue only.

Tracking node and plan preserved; not abandoned.

@briansrls briansrls closed this Jun 1, 2026
@briansrls
briansrls deleted the session/cool-fox-718 branch June 1, 2026 18:42
briansrls added a commit that referenced this pull request Jun 1, 2026
…e v2_compile_src_v4 closure (#4091 §1.2) (#4214)

* WIP: Next-wave: respawn cool-fox — #4175 model-only ci.dag hygiene (post-#417

* chore: apply cargo fmt on four-compile collapse test

Co-authored-by: Cursor <cursoragent@cursor.com>

* test(ci): add P5(b) same-path receipt for four-compile collapse pin

Codex RC: new hand-Rust in an EXPECTED_HAND_AUTHORED_TEST path needs an
inline Mechanism (b) receipt (SG-0 delta 0; ROADMAP T-PB-B deferral).

Co-authored-by: Cursor <cursoragent@cursor.com>

---------

Co-authored-by: Cursor <cursoragent@cursor.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