CI fast path: nextest, bump-only skip, package releases without a host round, fast_gates.sh (#318) #344

Closed
buildagent wants to merge 0 commits from ci-fast into master
Member

This PR answers the operator's request: "we urgently need to reduce CI time and be more productive", and "every change need test, PR, merge, release. this for 4x2hours for every simple EBF." It also serves the goal of language packages: "a faster development process and better maintenance, since we can deploy independent".

Measured before (round on the 0.34.0 commit)

Job Wall Cause
ci-windows (one serial job) 85 min 61 min of cargo test running 428 binaries one at a time; ~15 min of rebuilds after the tests
ci.yml cargo test 51.5 min binaries run serially
package-ci ruby leg 23.7 min dev + release compile back to back after a 5.3 min shared clippy
release.yml ~37 min

Commits

  • E, bb7e59d2: a package release needs only its own CI legs.
    • The bug: package-release.yml required the host jobs to be green. A package-only commit has no host rows, so every package release would have waited 50 min and then needed a hand-dispatched ~85 min host round. No pkg/* tag had ever been pushed, so this never surfaced.
    • Now package_ci_gate.sh requires the package's own two legs to be green and the host jobs only to be "not red". This reverses a #309 rule; the operator approved it on 2026-10-08.
    • The ruby leg is split into package (ruby) (dev) and package corpus (ruby) (release), plus one shared lint job. No gate was removed.
    • The path is documented end to end in tests/packages/README.md.
  • A, 3da1ec61 and d1d86fe6: a version-bump-only commit costs no CI round.
    • It applies only when the diff against the parent changes nothing but the version string on the BUMP_SITES declared in version_site_registry.rs, read from the parent.
    • *.rs and CI files can never be skippable sites.
    • The parent must be green in its own rows of the same workflow, and a skip never chains off another skip.
    • require_ci_green decides on the tagged commit's own rows first.
    • d1d86fe6 closes a "ci-gate-bypass" finding from the automated security review. I mutated the .rs refusal myself, and version_only_means_the_version_string_and_nothing_else went RED.
  • B, 99c4edc4: the workspace suite runs under cargo nextest (profile ci).
    • Retries are 0, so a flaky test stays visible.
    • Doc-tests run separately via cargo test --doc; the -- --ignored suites stay on cargo test.
    • The workflow-text gates now grade the nextest invocation.
    • Measured locally: 47.6 min → 13.0 min (3.7×).
    • One binary failed only under parallel load: language_package_overhead, whose 30 s precompile budget is starved on a loaded box. It and host_extraction_epoch are serialised in a test group.
    • Windows installs pinned nextest 0.9.144 with a sha256 check.
  • D, 942a3861: fast_gates.sh runs all 61 tree-reading and registry gates in about 2 min. fast_gates_registry.rs keeps the list complete. Lanes run it before reporting; today a lane missed three such gates.
  • C, 0dd5cf89: the Windows job sets one RUSTFLAGS, so the second full dev build after the tests is gone.

This PR's own CI is the first real-runner test of

  • nextest on both runners;
  • the Windows RUSTFLAGS change;
  • the package-ci split timing.

Not in this PR (needs the operator)

  • The Windows cache never restores ("Cache not found" on every run), so every run builds cold. A persistent target/ on the runner would fix it, but it has the stale-build-hash hazard.
  • Release reuse of CI-built binaries is a next step.

🤖 Generated with Claude Code

https://claude.ai/code/session_0126PDDLB4wNHxKXvWM1VNmu

This PR answers the operator's request: "we urgently need to reduce CI time and be more productive", and "every change need test, PR, merge, release. this for 4x2hours for every simple EBF." It also serves the goal of language packages: "a faster development process and better maintenance, since we can deploy independent". ## Measured before (round on the 0.34.0 commit) | Job | Wall | Cause | |---|---|---| | ci-windows (one serial job) | 85 min | 61 min of `cargo test` running 428 binaries one at a time; ~15 min of rebuilds after the tests | | ci.yml `cargo test` | 51.5 min | binaries run serially | | package-ci ruby leg | 23.7 min | dev + release compile back to back after a 5.3 min shared clippy | | release.yml | ~37 min | | ## Commits - **E, bb7e59d2: a package release needs only its own CI legs.** - The bug: `package-release.yml` required the host jobs to be green. A package-only commit has no host rows, so every package release would have waited 50 min and then needed a hand-dispatched ~85 min host round. No `pkg/*` tag had ever been pushed, so this never surfaced. - Now `package_ci_gate.sh` requires the package's own two legs to be green and the host jobs only to be "not red". **This reverses a #309 rule; the operator approved it on 2026-10-08.** - The ruby leg is split into `package (ruby)` (dev) and `package corpus (ruby)` (release), plus one shared `lint` job. No gate was removed. - The path is documented end to end in `tests/packages/README.md`. - **A, 3da1ec61 and d1d86fe6: a version-bump-only commit costs no CI round.** - It applies only when the diff against the parent changes nothing but the version string on the `BUMP_SITES` declared in `version_site_registry.rs`, read from the parent. - `*.rs` and CI files can never be skippable sites. - The parent must be green in its own rows of the same workflow, and a skip never chains off another skip. - `require_ci_green` decides on the tagged commit's own rows first. - d1d86fe6 closes a "ci-gate-bypass" finding from the automated security review. I mutated the `.rs` refusal myself, and `version_only_means_the_version_string_and_nothing_else` went RED. - **B, 99c4edc4: the workspace suite runs under `cargo nextest` (profile `ci`).** - Retries are 0, so a flaky test stays visible. - Doc-tests run separately via `cargo test --doc`; the `-- --ignored` suites stay on `cargo test`. - The workflow-text gates now grade the nextest invocation. - **Measured locally: 47.6 min → 13.0 min (3.7×).** - One binary failed only under parallel load: `language_package_overhead`, whose 30 s precompile budget is starved on a loaded box. It and `host_extraction_epoch` are serialised in a test group. - Windows installs pinned nextest 0.9.144 with a sha256 check. - **D, 942a3861: `fast_gates.sh` runs all 61 tree-reading and registry gates in about 2 min.** `fast_gates_registry.rs` keeps the list complete. Lanes run it before reporting; today a lane missed three such gates. - **C, 0dd5cf89: the Windows job sets one `RUSTFLAGS`, so the second full dev build after the tests is gone.** ## This PR's own CI is the first real-runner test of - nextest on both runners; - the Windows `RUSTFLAGS` change; - the package-ci split timing. ## Not in this PR (needs the operator) - **The Windows cache never restores** ("Cache not found" on every run), so every run builds cold. A persistent `target/` on the runner would fix it, but it has the stale-build-hash hazard. - **Release reuse of CI-built binaries** is a next step. 🤖 Generated with [Claude Code](https://claude.com/claude-code) https://claude.ai/code/session_0126PDDLB4wNHxKXvWM1VNmu
`version_bump.sh --version-only <commit>` answers whether a commit
differs from its first parent ONLY by the workspace version string:
the version moved, every changed file is a site of the new version by
the SAME `sites` scan the bump uses (no second list), every changed line
is its parent line with old -> new (never a `requires_host` line), and
the byte counts agree. Line-paired, so a third-party lockfile entry that
sits at the old version stays an unchanged line. Run against real
history: both bumps made under the current rules (0.34.0, 0.34.1) are
version-only; the four older ones (pre guest-exclusion lockfiles, a
release commit with source edits) are refused, each for its reason.

ci_changes.sh decision 0 (push and pull_request): a version-only commit
whose PARENT is green on GREEN_JOBS runs nothing (`mode=version-bump`).

require_ci_green.sh: a version-only commit's jobs that have no row of
their own (absent, or skipped only) are decided by the parent's rows; a
row the bump has of its own still decides. Needed because a skipped job
has NO row in /actions/tasks (runs 1025/1026), so the release gate would
otherwise wait out its deadline on `absent`. release.yml's windows-gate
checkout fetches history so the parent exists.

Tests (ci_path_gating, mutations run): the predicate's five refusals,
decision 0 on both events, its fall-throughs, and the gate's inheritance.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_0126PDDLB4wNHxKXvWM1VNmu
THE BLOCKER. package-release.yml's ci-gate required the HOST jobs green
for the tagged commit and no package leg at all. A package-only commit
has no host rows (ci_changes.sh path-filters every host job off, and a
skipped job has no row in /actions/tasks), so every package release
waited out the 50-minute deadline and then needed a hand-dispatched host
round (~85 min on the single Windows runner) that grades nothing a
package release ships. No pkg/* tag has ever been pushed.

package_ci_gate.sh now requires `package (<n>)` and `package corpus
(<n>)` green and the host jobs only not RED (require_ci_green.sh's new
NOT_RED_JOBS arm). A green leg row can be a leg that exited at its scope
step, so on a commit that does not change tests/packages/<n>/ only
nightly / dispatched rows count (new REQUIRED_EVENTS filter).
release_gate.rs now holds HOST_JOBS equal to release.yml's
REQUIRED_JOBS, which package-release.yml's comment claimed and nothing
checked.

THE RUBY LEG (23.7 min, run 1024). Not a special gate: ruby is the only
package with corpus suites, and the leg built two profiles back to back
(~10 min dev, ~9 min release compiling) after a 5.3 min clippy of a
crate every leg shares. The leg is now `package (<n>)` (dev, steps 2-5)
beside `package corpus (<n>)` (release, steps 6-7), and the clippy is
one `lint` job per run. Same commands, env and order within each half;
no gate removed. Estimated wall ~11 min, not measured (needs dispatch).

tests/packages/README.md documents the path end to end with its costs.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_0126PDDLB4wNHxKXvWM1VNmu
The review flagged `version_bump.sh --version-only` as a CI-gate bypass.
The most likely referent: a "site" was any file that STATED the new
version (the `sites` scan), and `.forgejo/` was walked. A commit could
make a script or a workflow state the version in a green parent and then
flip it with a pure old -> new edit — `[ "$V" = "<old>" ] && exit 0`
becomes a different program — and skip CI. The same line-paired rule
also let a third-party dependency pinned at the version move in
Cargo.toml / Cargo.lock.

Closed, every rule a refusal:
1. Content: Cargo.toml may change only `[workspace.package] version`;
   Cargo.lock only `version` lines of packages with no `source` (the
   workspace's own). Other declared sites: old -> new per line, bytes
   accounted.
2. The parent's green is real and of THIS workflow: decision 0 needs
   WORKFLOW_ID (ci.yml / ci-windows.yml) and require_ci_green.sh filters
   rows by `workflow_id`; skipped / cancelled / absent are not green; the
   parent check runs with inheritance off (a skip never chains off a
   skip — also structurally impossible, which the test records).
3. Skippable files are BUMP_SITES from version_site_registry.rs, read
   out of the PARENT commit; `*.rs` and `.forgejo/.github/.gitlab` are
   refused even if listed.
4. The release gate still needs a real green row for the tag, or for
   the parent of a version-only tag; its own rows always decide first.

Mutations run: S1-S6, S8, S9 RED; S7 (drop the redundant
INHERIT_VERSION_ONLY=false) survives, for the structural reason above.
Both real bump commits (0.34.0, 0.34.1) are still version-only.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_0126PDDLB4wNHxKXvWM1VNmu
ci.yml and ci-windows.yml run `cargo nextest run --workspace --exclude
code-index-first-party-tests --no-fail-fast --profile ci` and, beside it,
`cargo test --doc` over the same set (nextest runs no doc-tests). The
`-- --ignored` suites stay on `cargo test`: they are timing and corpus
measurements that run one at a time on purpose.

MEASURED locally, same args, same warm build, 12 cores (shared box):
  cargo test     2857 s (47.6 min); load1 at start 4.18, median 4.2
  cargo nextest   782 s (13.0 min); load1 at start 26.1 (15-min 8.3),
                  median 15.9 (nextest's own parallelism)
  3.7x. One test failed under nextest only:
  language_package_overhead::php_psr4_layout_adds_no_reachability_the_
  builtin_lacks — its package refused with PRECOMPILE_WALL_CLOCK_BUDGET_MS
  observed 30000: a debug cranelift precompile starved of CPU by the
  other binaries (the whole binary took 33 s under cargo test). Those
  precompile-heavy binaries are a nextest test-group (max 2, 4 slots
  each). The group config was run on its two binaries (13 passed, 176 s);
  a full re-run under it was NOT made (two full runs was the budget).

.config/nextest.toml `ci` profile: retries = 0 (a flake stays red),
failure output immediate+final, slow tests reported every 120 s and
never killed.

Install: the Linux image ci-rust:1.94.0-v7 ships cargo-nextest 0.9.144
(probed); the step prints its version so an image without it fails by
name. Windows: a pinned 0.9.144 prebuilt zip, sha256 measured here
(upstream publishes none), skipped when that version is present.

Gates still grade the real invocation: ci_path_gating's package-home rule
reads `cargo nextest run` as a test invocation (and now expects four
host workspace runs), ci_cadence counts a nextest step as a grading step
and requires --profile ci and retries = 0, ci_disk_preflight treats
`cargo nextest` as a compiler invocation. Mutations run: retries = 2,
--profile ci dropped, --exclude dropped, the doc-test step's
`success() || failure()` dropped — each RED.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_0126PDDLB4wNHxKXvWM1VNmu
.forgejo/scripts/fast_gates.sh runs 61 targets — every
`crates/*/tests/*_gate.rs` and `*_registry.rs` but precision_gate (it
indexes fixtures through the daemon; CI runs it in release), plus the
suites that read the tree or the workflows (ci_path_gating, ci_cadence,
ignored_test_reachability, frozen_pins, reason_producers,
release_notes_quoting, corpus_require_floor, ...). No corpus. nextest
when present, cargo test otherwise.

MEASURED, warm dev build, 12 cores, load 5.9 at start: 483 tests, 105 s,
green. disclosure_derivation_registry is the long pole (82 s).

fast_gates_registry.rs keeps the list from rotting: every suffixed target
is named exactly once (GATES or NOT_HERE with a >= 30-char reason) and
every named path exists. Mutations run: a new zz_probe_gate.rs, a
dropped doc_citation_gate, a dead path, a 4-char reason — each RED.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_0126PDDLB4wNHxKXvWM1VNmu
ci-windows: one RUSTFLAGS for the whole job, one dev build (CI-FAST C)
Some checks failed
Package CI / package suites (clippy) (pull_request) Successful in 18m59s
CI / cargo clippy (pull_request) Successful in 22m59s
CI / cargo check (MSRV 1.98) (pull_request) Successful in 21m27s
Package CI / package (svelte) (pull_request) Successful in 33m7s
CI / CI lane wall-clock headroom (pull_request) Successful in 1m2s
Package CI / package (php) (pull_request) Failing after 35m5s
Package CI / package (xaml) (pull_request) Successful in 35m29s
CI / cargo test (pull_request) Has been cancelled
CI / cargo test (daemon transport) (pull_request) Has been cancelled
CI / Plugin path cost + pool throughput (nightly) (pull_request) Has been cancelled
CI / guest crates (fmt, clippy, doc) (pull_request) Has been cancelled
CI / cargo check (windows-gnu) (pull_request) Has been cancelled
Package CI / package (csharp) (pull_request) Has been cancelled
Package CI / package corpus (javascript) (pull_request) Has been cancelled
Package CI / package corpus (php) (pull_request) Has been cancelled
Package CI / package corpus (rust) (pull_request) Has been cancelled
Package CI / package corpus (ruby) (pull_request) Has been cancelled
Package CI / package (timeline) (pull_request) Has been cancelled
Package CI / package (javascript) (pull_request) Has been cancelled
CI / cargo test (abi, 32-bit + wasm32) (pull_request) Has been cancelled
CI / cargo doc (intra-doc links) (pull_request) Has been cancelled
Package CI / package (ruby) (pull_request) Has been cancelled
Package CI / package corpus (typescript) (pull_request) Has been cancelled
CI / OSS corpus (tier 1) (pull_request) Has been cancelled
Package CI / package (python) (pull_request) Has been cancelled
Package CI / package corpus (csharp) (pull_request) Has been cancelled
Package CI / package corpus (python) (pull_request) Has been cancelled
CI / cargo deny (pull_request) Has been cancelled
Package CI / package (rust) (pull_request) Has been cancelled
Package CI / package (typescript) (pull_request) Has been cancelled
0dd5cf89cb
WHY THE JOB REBUILT (run 1022, measured from its log). After 61 min of
tests the job compiled the workspace four more times:
  11:40  46 s   test profile  `cargo test -p code-index-cli --test
                installer_ps1_e2e`: `-p` alone changes feature
                unification against the `--workspace` build (NOT changed
                here; ~46 s)
  11:43  3m08s  dev           `cargo build --workspace --bins` under
  11:45  1m47s  dev           RUSTFLAGS=+crt-static (+ release_smoke):
                RUSTFLAGS is in every crate's fingerprint, so a full
                second dev graph
  11:51  5m11s  release       the release worker + smoke harness (#232),
  11:55  3m51s  release       under the same flags
The job-level RUSTFLAGS removes the second dev graph (~5 min): clippy,
build and test now build what the smoke steps build, and the tests run
the CRT linkage the archive ships. The release-profile smoke is KEPT: it
is the per-push evidence about the profile the archive ships, and moving
it to release.yml would trade ~9 min here for finding #232's class of
defect at tag time.

Also found: the job's actions/cache never restores ("Cache not found" for
every key) — its post-step tars a 42 GB target/ — so every run builds
cold. Not changed here: a persistent CARGO_TARGET_DIR on the runner is
the obvious fix and is exactly the stale-build-hash hazard recorded for
shared target dirs; it needs the operator.

UNVERIFIED ON WINDOWS: not dispatched (per instructions). release.yml's
build-windows already builds the full workspace, proc-macros included,
under this exact flag, which is the evidence it compiles; whether every
test passes under the static CRT is what the first dispatch will show.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_0126PDDLB4wNHxKXvWM1VNmu
ci-windows: install cargo-nextest into a job-local directory (CI-FAST B)
Some checks failed
CI / cargo clippy (pull_request) Successful in 23m21s
CI / cargo check (MSRV 1.98) (pull_request) Successful in 22m54s
CI (Windows) / fmt + clippy + build + test (windows) (pull_request) Failing after 30m32s
Package CI / package (python) (pull_request) Failing after 35m48s
Package CI / package (xaml) (pull_request) Successful in 36m44s
CI / CI lane wall-clock headroom (pull_request) Successful in 1m1s
Package CI / package (svelte) (pull_request) Successful in 37m4s
Package CI / package (timeline) (pull_request) Successful in 37m41s
CI / guest crates (fmt, clippy, doc) (pull_request) Successful in 1m39s
Package CI / package (php) (pull_request) Successful in 39m51s
Package CI / package (rust) (pull_request) Successful in 39m51s
Package CI / package (csharp) (pull_request) Successful in 40m41s
Package CI / package (typescript) (pull_request) Successful in 41m25s
CI / cargo check (windows-gnu) (pull_request) Successful in 23m7s
Package CI / package (javascript) (pull_request) Successful in 41m42s
Package CI / package (ruby) (pull_request) Successful in 42m17s
CI / cargo test (abi, 32-bit + wasm32) (pull_request) Successful in 16m38s
CI / cargo deny (pull_request) Successful in 15m55s
CI / cargo doc (intra-doc links) (pull_request) Successful in 8m36s
Package CI / package corpus (csharp) (pull_request) Successful in 52m35s
Package CI / package corpus (php) (pull_request) Successful in 52m38s
Package CI / package corpus (python) (pull_request) Failing after 52m43s
Package CI / package corpus (javascript) (pull_request) Successful in 52m48s
Package CI / package corpus (rust) (pull_request) Successful in 52m49s
Package CI / package corpus (typescript) (pull_request) Successful in 52m33s
Package CI / package corpus (ruby) (pull_request) Successful in 58m32s
CI / cargo test (pull_request) Successful in 29m35s
CI / OSS corpus (tier 1) (pull_request) Successful in 41m48s
CI / cargo test (daemon transport) (pull_request) Successful in 12m39s
CI / Plugin path cost + pool throughput (nightly) (pull_request) Has been skipped
a8d7d312cb
Run 1036 failed in 7 minutes: Expand-Archive was refused write access
to the runner's shared C:\ci\cargo\bin. The pinned, sha256-verified zip
now unpacks under RUNNER_TEMP, and that directory joins PATH for the
step and, through GITHUB_PATH, for the rest of the job.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_0126PDDLB4wNHxKXvWM1VNmu
The first Windows nextest run (1039) failed 8 tests, one class: under
cargo-nextest every test is its own process and binaries run side by
side, so an in-test build could rewrite an executable another test was
running.

1. `cargo build -p X --bin X` inside tests. Each `-p` selection is its own
   feature unification and so its own binary — MEASURED on Linux after
   `cargo build --workspace`: `cargo build -p code-index-daemon --bin
   code-index-daemon` uplifts a 150 MB daemon over the 216 MB workspace
   one (a different inode), the test build's dev-dependency features make
   a third (216.5 MB), and an identical repeat is a no-op (inode kept). So
   the relink happens even when nothing changed, and Windows refuses it:
   six query_cli tests, `failed to remove file …code-index-daemon.exe`.
   Seven build sites had no Windows present-exe skip (query_cli x2,
   blast_radius_hook, plugin_add_mcp_e2e, python_module_symbols_e2e,
   conformance_isolation, package_fixture); each now takes the
   harness-built binary on Windows in the one spelling
   inline_cargo_build_registry already accepts. query_cli's second,
   daemon-only build is gone: the test's "another build" is a rewritten
   registration, not another binary.
   blast_radius_hook returned whenever the sibling merely EXISTED on every
   platform, spelled `is_file()`, which the existing gate did not see; it
   now builds, and the gate reads `.is_file()` too.

2. `rustc -o <fixed path>`: plugin_host_handshake's stand-in worker was
   compiled by two test processes into one target/debug path at once
   (LNK1104, two failures). Per-process directory now; fork_window's
   stand-in likewise (unix, same hazard as ETXTBSY).

Gate: inline_cargo_build_registry gains
`no_in_test_build_replaces_an_executable_another_test_runs` — every
in-test build site has a `#[cfg(windows)]` present-exe skip in its own fn
or in a same-file helper it delegates to, and every `rustc -o` names the
process id — plus synthetic controls. Mutations RUN, all RED: drop
conformance_isolation's skip; drop daemon/common's delegated `build_once`
skip; drop the pid from foreign_worker; restore blast_radius_hook's
unguarded `is_file()` return (old gate).

Verified: clippy -D warnings for x86_64-pc-windows-gnu over the four
touched test crates (compiles the cfg(windows) arms); the touched suites
under nextest on Linux, 38/38. NOT verified on native Windows.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_0126PDDLB4wNHxKXvWM1VNmu
Every test that runs a real package precompiles its extractor through the
DEBUG `code-index-plugin-host --precompile`, and unoptimised cranelift
was the cost: MEASURED 8.0-10.2 s of CPU per first-party extractor,
single-threaded (load 8.4, 12 cores). The product kills a precompile at
PRECOMPILE_WALL_CLOCK_BUDGET_MS = 30 s, so a busy CI box ran debug tests
into it and the package was refused as `extractor.malformed`:
- package-ci run 1041, `package (python)`: package_release_scripts'
  `plugin check` verdict "staging REFUSED extractor.malformed";
- the local nextest run of language_package_overhead (observed: 30000).
Not caused by the CI-FAST E logic, but made likely by it: the dev/corpus
split doubled the concurrent jobs on the package box (21 in run 1041).

opt-level = 3 for the cranelift crates, regalloc2 and
wasmtime-internal-cranelift only; every other crate's debug build is
unchanged and the product bound is unchanged. MEASURED after:
  precompile, six extractors      0.7-0.9 s each (was 8.0-15.4 s)
  host_extraction_epoch + language_package_overhead under nextest
                                  54.8 s, load 5.1 (was 176 s, load 6.3)
  package_release_scripts         92.8 s, load 2.6 (cargo test: 337 s,
                                  load ~4; nextest full run: 437 s, contended)
  one-off cost: 53 s incremental rebuild of the plugin host.

ci_cadence::the_wasm_code_generator_is_optimised_in_debug_builds reads
the crate set from Cargo.lock (a wasmtime upgrade that adds one fails
until it is optimised); it found cranelift-native, which the first draft
missed. Mutation RUN: drop the cranelift-codegen override — RED naming it.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_0126PDDLB4wNHxKXvWM1VNmu
package-ci: one job per package again, one cargo selection per profile (CI-FAST E)
Some checks failed
CI (Windows) / changes (what this run must test) (pull_request) Successful in 28s
Package CI / changes (which packages this run must test) (pull_request) Successful in 23s
CI / changes (what this run must test) (pull_request) Successful in 42s
CI / cargo fmt (pull_request) Successful in 1m5s
CI / OSS corpus tier-3 scale (nightly) (pull_request) Has been skipped
CI / Grammar rebuild from source (nightly) (pull_request) Has been skipped
CI / CI lane wall-clock headroom (pull_request) Successful in 1m6s
CI / guest crates (fmt, clippy, doc) (pull_request) Successful in 1m58s
Package CI / package suites (clippy) (pull_request) Successful in 16m25s
CI / cargo deny (pull_request) Successful in 16m56s
CI / cargo test (abi, 32-bit + wasm32) (pull_request) Successful in 18m17s
CI / cargo check (MSRV 1.98) (pull_request) Successful in 20m51s
CI / cargo clippy (pull_request) Successful in 21m16s
CI / cargo check (windows-gnu) (pull_request) Successful in 22m44s
CI / cargo doc (intra-doc links) (pull_request) Successful in 12m29s
Package CI / package (xaml) (pull_request) Successful in 36m5s
Package CI / package (svelte) (pull_request) Successful in 36m19s
Package CI / package (timeline) (pull_request) Successful in 36m24s
CI (Windows) / fmt + clippy + build + test (windows) (pull_request) Successful in 40m8s
Package CI / package (rust) (pull_request) Successful in 1h7m18s
Package CI / package (javascript) (pull_request) Successful in 1h7m56s
Package CI / package (csharp) (pull_request) Successful in 1h8m1s
Package CI / package (php) (pull_request) Successful in 1h8m4s
Package CI / package (typescript) (pull_request) Successful in 1h8m27s
Package CI / package (python) (pull_request) Failing after 1h9m48s
Package CI / package (ruby) (pull_request) Successful in 1h14m53s
CI / OSS corpus (tier 1) (pull_request) Successful in 1h10m57s
CI / cargo test (pull_request) Failing after 1h4m44s
CI / cargo test (daemon transport) (pull_request) Has been skipped
CI / Plugin path cost + pool throughput (nightly) (pull_request) Has been skipped
4de58322f5
DECISION (b), from run 1041 against nightly 1018 — both "every package
in scope":
  1018, one job per package:   legs 43.8-50.8 min, wall ~51 min
  1041, dev + corpus split:     dev 35.8-42.3, corpus 52.5-58.5, lint
                                18.5; wall 58.5 min, 21 heavy jobs at
                                once, python's precompile ran into the
                                product's 30 s bound
Every change to package-ci.yml, crates/first-party-tests or a shared
guest crate puts every package in scope, as does the nightly, so on the
one package box the split was a net LOSS of ~8 min and a source of
contention failures. The leg is one job again; the once-per-run `lint`
job stays (it lints the shared crate, not a package).

What the split was meant to buy comes from the SELECTION. Each cargo
package selection is its own feature unification, so its own build of
the dependency graph; the leg used four dev selections and two release
ones (1041 python: 17m37 + 7m03 + 7m14 dev, 27m04 + 22m49 release of
compiling). Every dev `cargo test` now names the same four packages, and
every release `cargo test`/`cargo build` the same two. MEASURED locally
from an empty target dir, the dev steps only (load 7.6-11):
  old selections   81 + 65 + 39 + 28 s = 213 s, 406 crates compiled
  one selection    67 +  0 + 11 +  1 s =  79 s, 267 crates
(The in-test `cargo build -p code-index-plugin-host` of
package_release_scripts still builds its own variant: 15-43 s.)

package_ci_gate.sh requires the one leg `package (<name>)`.
ci_path_gating::a_leg_builds_each_profile_once replaces the two-halves
test. Mutations RUN, each RED: drop `-p code-index-plugin-host` from the
phase 3/4 step (two dev selections); restore the plugin-host-only
release build (two release selections); require the old corpus leg name
in the gate.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_0126PDDLB4wNHxKXvWM1VNmu
ci: wall-clock-bounded tests run alone, after the suite (PR #344 round 3)
Some checks failed
Package CI / changes (which packages this run must test) (pull_request) Successful in 20s
CI / changes (what this run must test) (pull_request) Successful in 26s
CI (Windows) / changes (what this run must test) (pull_request) Successful in 35s
CI / cargo fmt (pull_request) Successful in 1m1s
CI / OSS corpus tier-3 scale (nightly) (pull_request) Has been skipped
CI / Grammar rebuild from source (nightly) (pull_request) Has been skipped
CI / CI lane wall-clock headroom (pull_request) Successful in 1m54s
CI / guest crates (fmt, clippy, doc) (pull_request) Successful in 2m12s
CI / cargo doc (intra-doc links) (pull_request) Successful in 14m42s
Package CI / package suites (clippy) (pull_request) Successful in 17m8s
CI / cargo deny (pull_request) Successful in 16m57s
CI / cargo test (abi, 32-bit + wasm32) (pull_request) Successful in 17m45s
CI / cargo check (MSRV 1.98) (pull_request) Successful in 20m4s
CI / cargo clippy (pull_request) Successful in 20m24s
CI / cargo check (windows-gnu) (pull_request) Successful in 21m36s
Package CI / package (svelte) (pull_request) Successful in 33m7s
Package CI / package (xaml) (pull_request) Successful in 33m19s
Package CI / package (timeline) (pull_request) Successful in 33m36s
CI (Windows) / fmt + clippy + build + test (windows) (pull_request) Successful in 44m19s
Package CI / package (csharp) (pull_request) Successful in 1h1m3s
Package CI / package (rust) (pull_request) Successful in 1h1m41s
Package CI / package (php) (pull_request) Successful in 1h2m2s
Package CI / package (javascript) (pull_request) Successful in 1h2m7s
Package CI / package (typescript) (pull_request) Successful in 1h2m6s
Package CI / package (python) (pull_request) Failing after 1h3m33s
Package CI / package (ruby) (pull_request) Successful in 1h7m17s
CI / OSS corpus (tier 1) (pull_request) Successful in 1h13m1s
CI / cargo test (pull_request) Successful in 56m38s
CI / cargo test (daemon transport) (pull_request) Successful in 12m24s
CI / Plugin path cost + pool throughput (nightly) (pull_request) Has been skipped
0d35dfd265
Run 1045: m2_acceptance::m2_thousand_file_repo_finishes_quickly asserted
a 1000-file cold index under 6 s (its own comment: ~120 ms in a debug
build) and took 21.45 s beside the nextest suite on a box running ten
package jobs. The bound was right; the measurement was contended.

THE POPULATION, by rule rather than by name:
crates/test-support/tests/wall_clock_registry.rs lexes every #[test] /
#[tokio::test] in crates/*/{src,tests,benches,examples} and takes the
ones that assert an UPPER bound on a value derived from `.elapsed()`
(directly or through a `let` chain: `took`, `ratio`, ...), including
`if took > BUDGET { panic!() }`. Lower bounds ("it really waited") and
deadline loops (`assert!(Instant::now() < deadline)`) are left out.
31 non-ignored tests in 6 packages, m2 among them; #[ignore]d ones run
under `cargo test -- --ignored --test-threads=1` already. It cannot see a
bound asserted inside a helper the test calls.

THE MECHANISM: `[profile.ci]` gets `default-filter = 'not (…)'` over
those 31, and `[profile.ci-timing]` runs exactly them with
`test-threads = 1`, as a second nextest step in ci.yml and
ci-windows.yml (`success() || failure()`). No bound was raised.
NOT an exclusive override inside one run, and that was MEASURED: nine
binaries / 284 tests under 12 busy loops of external load,
  threads-required = "num-test-threads" override   676 s
  no override                                       244 s
  ci (excluding them) + ci-timing                   230 + 84 = 314 s
because nextest drains every running test before each exclusive one.
m2 under that load: 3.6-4.0 s beside the subset, 2.0 s alone. External
load (other jobs on the CI box) is not removed by any nextest setting.

Gates and mutations RUN, each RED:
  wall_clock_registry: drop m2 from both profiles (named missing);
  test-threads = 4; a stale term in both; m2 dropped from ci's exclusion
  only (the profiles disagree); synthetic take/leave controls.
  ci_cadence: delete ci-windows.yml's ci-timing step (the excluded tests
  would run nowhere).
fast_gates.sh lists the new registry (fast_gates_registry requires it):
62 targets, 488 tests, 79 s.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_0126PDDLB4wNHxKXvWM1VNmu
python_imports: a layout that restates the native identities proves nothing (#347)
All checks were successful
Package CI / changes (which packages this run must test) (push) Successful in 30s
CI / changes (what this run must test) (push) Successful in 44s
CI (Windows) / changes (what this run must test) (push) Successful in 53s
CI (Windows) / fmt + clippy + build + test (windows) (push) Has been skipped
CI / cargo fmt (push) Successful in 55s
CI / OSS corpus (tier 1) (push) Has been skipped
CI / OSS corpus tier-3 scale (nightly) (push) Has been skipped
CI / Grammar rebuild from source (nightly) (push) Has been skipped
CI / cargo clippy (push) Has been skipped
CI / cargo check (MSRV 1.98) (push) Has been skipped
CI / cargo check (windows-gnu) (push) Has been skipped
CI / cargo test (abi, 32-bit + wasm32) (push) Has been skipped
CI / cargo deny (push) Has been skipped
CI / CI lane wall-clock headroom (push) Has been skipped
CI / guest crates (fmt, clippy, doc) (push) Has been skipped
CI / cargo doc (intra-doc links) (push) Has been skipped
CI / cargo test (push) Has been skipped
CI / cargo test (daemon transport) (push) Has been skipped
CI / Plugin path cost + pool throughput (nightly) (push) Has been skipped
Package CI / package suites (clippy) (push) Successful in 10m50s
Package CI / package (xaml) (push) Successful in 23m56s
Package CI / package (svelte) (push) Successful in 24m18s
Package CI / package (timeline) (push) Successful in 24m46s
Package CI / package (csharp) (push) Successful in 44m12s
Package CI / package (rust) (push) Successful in 45m3s
Package CI / package (php) (push) Successful in 45m24s
Package CI / package (javascript) (push) Successful in 45m27s
Package CI / package (typescript) (push) Successful in 45m32s
Package CI / package (python) (push) Successful in 46m16s
Package CI / package (ruby) (push) Successful in 51m35s
05dcbee20c
The first-party de.h-dv.python package left 91 builtin binds on
python-flask unresolved (package-ci `package (python)`, parity axis
"references and targets", 15365 vs 15365 rows; #315's table). Every one
was the denied-refs arm: the package declares `[module_map]`, its hook
answered READY for all 83 files, and identities_for_pass turned each
answer into Proven -- so every file became layout-OWNED and every import
whose origin is not one complete object (`from pallets_sphinx_themes
import get_version`, `from flask import request` re-exported from a
globals.py that declares no `request`, `from flaskr import create_app`)
became negative evidence denying the native tiers' cross-file bind.

But flask's pyproject.toml is metadata-only, and python_layout's
non-explicit arm "preserves the native default": all 83 module names
were byte-equal to the host's own path derivation. The hook said nothing
the builtin did not already assume, yet it switched the resolver into a
different regime. module_map::namespace_candidates already states the
rule for the other layout consumer: a source-consistent layout "supplies
no additional proof", so a first-party package resolves bind for bind as
the builtin does.

One clause, over the WHOLE answer: when every dotted-module file's layout
identity equals native_module(path) -- or is PATH_UNMAPPED where the
native derivation names nothing either -- the pass resolves as with no
layout. Any file that differs or is withheld keeps the layout's module
set, and every file stays owned (never per file: the negative evidence
is about the set). The builtin leg has no layouts, so it cannot move.

Tests (mutations RUN, each RED):
  python_imports::module_metadata_tests::
    a_layout_restating_the_native_identities_resolves_as_no_layout
    - restates_native forced false                  -> restating arm red
    - PATH_UNMAPPED arm forced false (the .pyw)      -> restating arm red
    - decided per file instead of over the answer    -> differing arm red
  python_language_overhead (the package leg as package-ci spells it):
    red before (the 91 sites), green after.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_0126PDDLB4wNHxKXvWM1VNmu
Author
Member

Fast-forwarded into master at 05dcbee2, approved by the operator. Round 5 (runs 1050–1052) was green on every job. This is a CI-only change, so there is no release.

Measured on the real runners:

  • Windows: 85 min → 41 min.
  • Linux cargo test: 29.6 min on a quiet box, 56 min with the box overloaded.

Also in this PR:

  • #347 item 2: the python package parity fix, 910353e9.
  • Wall-clock-bounded tests: they now run alone, after the main suite.
  • No exe rebuild races on Windows: a test no longer rebuilds an exe that another test is running.
  • Package CI: one job per package again.

#347 item 1, the python cold-cost budget, stays open. The nightly run asserts it.

Forgejo does not mark a fast-forward as merged, so I'm closing this by hand.

Fast-forwarded into master at 05dcbee2, approved by the operator. Round 5 (runs 1050–1052) was green on every job. This is a CI-only change, so there is no release. Measured on the real runners: - **Windows:** 85 min → 41 min. - **Linux `cargo test`:** 29.6 min on a quiet box, 56 min with the box overloaded. Also in this PR: - **#347 item 2:** the python package parity fix, `910353e9`. - **Wall-clock-bounded tests:** they now run alone, after the main suite. - **No exe rebuild races on Windows:** a test no longer rebuilds an exe that another test is running. - **Package CI:** one job per package again. #347 item 1, the python cold-cost budget, stays open. The nightly run asserts it. Forgejo does not mark a fast-forward as merged, so I'm closing this by hand.
buildagent closed this pull request 2026-10-09 13:28:19 +02:00
All checks were successful
CI / changes (what this run must test) (pull_request) Successful in 21s
CI (Windows) / changes (what this run must test) (pull_request) Successful in 30s
Package CI / changes (which packages this run must test) (pull_request) Successful in 47s
CI / cargo fmt (pull_request) Successful in 1m10s
CI / OSS corpus tier-3 scale (nightly) (pull_request) Has been skipped
CI / Grammar rebuild from source (nightly) (pull_request) Has been skipped
CI / CI lane wall-clock headroom (pull_request) Successful in 2m18s
CI / guest crates (fmt, clippy, doc) (pull_request) Successful in 2m19s
CI / cargo doc (intra-doc links) (pull_request) Successful in 14m13s
CI / cargo deny (pull_request) Successful in 16m27s
Package CI / package suites (clippy) (pull_request) Successful in 16m44s
CI / cargo test (abi, 32-bit + wasm32) (pull_request) Successful in 16m42s
CI / cargo check (MSRV 1.98) (pull_request) Successful in 19m43s
CI / cargo clippy (pull_request) Successful in 20m5s
CI / cargo check (windows-gnu) (pull_request) Successful in 21m41s
Package CI / package (xaml) (pull_request) Successful in 33m16s
Package CI / package (svelte) (pull_request) Successful in 33m39s
Package CI / package (timeline) (pull_request) Successful in 33m43s
CI (Windows) / fmt + clippy + build + test (windows) (pull_request) Successful in 41m14s
Package CI / package (javascript) (pull_request) Successful in 59m37s
Package CI / package (rust) (pull_request) Successful in 1h0m47s
Package CI / package (php) (pull_request) Successful in 1h1m14s
Package CI / package (csharp) (pull_request) Successful in 1h1m11s
Package CI / package (typescript) (pull_request) Successful in 1h1m43s
Package CI / package (python) (pull_request) Successful in 1h2m51s
Package CI / package (ruby) (pull_request) Successful in 1h6m44s
CI / OSS corpus (tier 1) (pull_request) Successful in 1h12m48s
CI / cargo test (pull_request) Successful in 56m36s
CI / cargo test (daemon transport) (pull_request) Successful in 12m11s
CI / Plugin path cost + pool throughput (nightly) (pull_request) Has been skipped
Package CI / changes (which packages this run must test) (push) Successful in 30s
CI / changes (what this run must test) (push) Successful in 44s
CI (Windows) / changes (what this run must test) (push) Successful in 53s
CI (Windows) / fmt + clippy + build + test (windows) (push) Has been skipped
CI / cargo fmt (push) Successful in 55s
CI / OSS corpus (tier 1) (push) Has been skipped
CI / OSS corpus tier-3 scale (nightly) (push) Has been skipped
CI / Grammar rebuild from source (nightly) (push) Has been skipped
CI / cargo clippy (push) Has been skipped
CI / cargo check (MSRV 1.98) (push) Has been skipped
CI / cargo check (windows-gnu) (push) Has been skipped
CI / cargo test (abi, 32-bit + wasm32) (push) Has been skipped
CI / cargo deny (push) Has been skipped
CI / CI lane wall-clock headroom (push) Has been skipped
CI / guest crates (fmt, clippy, doc) (push) Has been skipped
CI / cargo doc (intra-doc links) (push) Has been skipped
CI / cargo test (push) Has been skipped
CI / cargo test (daemon transport) (push) Has been skipped
CI / Plugin path cost + pool throughput (nightly) (push) Has been skipped
Package CI / package suites (clippy) (push) Successful in 10m50s
Package CI / package (xaml) (push) Successful in 23m56s
Package CI / package (svelte) (push) Successful in 24m18s
Package CI / package (timeline) (push) Successful in 24m46s
Package CI / package (csharp) (push) Successful in 44m12s
Package CI / package (rust) (push) Successful in 45m3s
Package CI / package (php) (push) Successful in 45m24s
Package CI / package (javascript) (push) Successful in 45m27s
Package CI / package (typescript) (push) Successful in 45m32s
Package CI / package (python) (push) Successful in 46m16s
Package CI / package (ruby) (push) Successful in 51m35s

Pull request closed

Sign in to join this conversation.
No reviewers
No milestone
No project
No assignees
1 participant
Notifications
Due date
The due date is invalid or out of range. Please use the format "yyyy-mm-dd".

No due date set.

Dependencies

No dependencies set

Reference
h-dv/code-index!344
No description provided.