PR 0: future-size measurement harness + build-std/nm CI gates (#125) by JustinKovacich · Pull Request #127 · luminartech/simple_someip

JustinKovacich · 2026-06-10T21:11:57Z

First PR of the #125 / phase-22 close-out stack
(docs/simple_someip/plans/2026-06-09-phase22-125-memory-reduction-design.md).

Stacked on #124 (feat/no-alloc-bare-metal) — when that merges, this
retargets to feature/phase21_api_symmetry.

Host-proxy future-size witness tests (budgets = baseline x1.25) for client
run/socket-loop + server run futures, tokio AND static-channel configs.
tools/capture_type_sizes.sh + size_probe client-future instantiation:
authoritative thumbv7em -Zprint-type-sizes table. Baseline committed under
docs/simple_someip/plans/baselines/.
CI: -Zbuild-std=core gate (halo no-alloc-sysroot certification; the
prebuilt-sysroot job ships alloc and cannot catch this class) + nm
alloc audit extended to server,bare_metal (alloc-free since feat(server): make server alloc-free on bare_metal, and surface inbound non-SD unicast #124).

No production-code behavior changes; transport::probe adds public
zero-behavior stubs (documented probe-only). First commit is rustfmt/clippy
cleanup over the #124 base; review commit-by-commit if that's noisy.

🤖 Generated with Claude Code

Pure formatting/lint fixes on Feliciano's #124 commits: import ordering, line reflow, allow(unused_imports) on the no-tracing log aliases (every call site compiles out under --no-default-features), and Box::pin on the embassy example's ~16 KiB setup future (clippy::large_futures). No behavior changes. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>

Moves the zero-behavior NullSocket/NullFactory/NullTimer/ NullE2ERegistry/NullInterface impls out of #[cfg(test)] into transport::probe (cfg'd test-or-bare_metal) and adds NullSpawner, so tools/size_probe can instantiate the client futures on thumbv7em-none-eabihf for -Zprint-type-sizes layout capture. Documented probe-only: not for production use (#125 PR 0). Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>

Host-arch proxy budgets (baseline x1.25, rounded up to 64) for Inner::run_future and the spawned socket loop, measured via size_of_val before tokio::spawn moves them. Trips on layout regressions; authoritative thumb numbers come from tools/capture_type_sizes.sh (#125 PR 0). Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>

Same proxy-budget rules as the client witness. Also rustfmt reflow of the #124-era test bodies in this file. (#125 PR 0) Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>

…nnels Bare-metal-channel configuration witnesses for the client run/socket- loop futures and the server run future — the configuration TC4 actually ships, host-arch proxy numbers (#125 PR 0). Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>

…seline - tools/size_probe grows a client feature: instantiates Inner::run_future / socket_loop_future no_std over transport::probe Null* deps and a probe-local heapless ProbePayload (mirrors TestPayload — cross-ref note added there), so -Zprint-type-sizes reports real on-target layouts. - tools/capture_type_sizes.sh: one-command host + thumb capture into target/type-sizes/summary.md. - docs/.../baselines/pr0-size-baseline.md: committed pre-optimization numbers the #125 PRs diff against (thumb table is client-only in PR 0; server thumb probe lands with PR 3's static plumbing). - docs/.../2026-06-09-phase22-125-memory-reduction-design.md: the spec this stack implements, referenced by CI comments and the baseline. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>

…are_metal - New build_std_core job: builds client/server/combined bare_metal with -Zbuild-std=core on thumbv7em — halo's TC4 proxy has NO alloc in its sysroot, and the prebuilt-sysroot job ships alloc so it cannot catch an extern-crate-alloc regression. - no_std job: server,bare_metal gets the same clean-build + nm alloc-symbol audit as the client (alloc-free since #124); stale Arc<EventPublisher> comments fixed. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>

Also updates the 0.8.0 'known limitation: server pulls alloc' note, which PR #124 made false (recorded deviation — #124 didn't touch the changelog, and this PR's CI now asserts the opposite). Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>

Copilot

Pull request overview

Adds a measurement/regression harness for async future layout growth (issue #125) and strengthens CI gates to ensure thumbv7em bare-metal builds remain alloc-free, including under -Zbuild-std=core. This fits the codebase’s embedded/no-alloc goals by making memory regressions detectable early and by providing authoritative on-target type-size capture tooling.

Changes:

Introduces thumbv7em -Zprint-type-sizes capture tooling (tools/capture_type_sizes.sh) and a tools/size_probe no_std probe that instantiates client run/socket-loop futures for authoritative layout measurement.
Adds host “future size witness” tests (tokio + bare-metal static-channel configs) with +25% proxy budgets to trip on regressions.
Extends CI with a nightly -Zbuild-std=core thumb gate and expands alloc-symbol (nm) audits to cover server,bare_metal.

Reviewed changes

Copilot reviewed 22 out of 23 changed files in this pull request and generated 2 comments.

Show a summary per file

File	Description
tools/size_probe/src/lib.rs	Adds a no_std client-future instantiation entrypoint and probe-local payload/channel types for type-size capture.
tools/size_probe/Cargo.toml	Updates probe crate purpose/docs and adds deps needed for the probe-local payload implementation.
tools/size_probe/Cargo.lock	Locks new transitive deps introduced by enabling `client` + adding probe-local deps.
tools/capture_type_sizes.sh	New script to generate host + thumbv7em `-Zprint-type-sizes` summaries.
tests/no_alloc_server_witness.rs	Formatting-only update to static subscription storage initialization.
tests/bare_metal_server.rs	Minor import/formatting adjustments around `NonSdRequestCallback` and a test assertion.
tests/bare_metal_e2e.rs	Adds documentation about shared static pools + adds bare-metal future-size witness test with budgets.
src/transport.rs	Promotes “null” dependency stubs into `transport::probe` for layout probing / trait conformance.
src/server/subscription_manager.rs	Formatting-only change to the boxed-future associated type.
src/server/runtime.rs	Minor formatting and log message formatting improvements.
src/server/mod.rs	Formatting-only adjustments and import consolidation.
src/protocol/sd/test_support.rs	Notes coupling between test payload and size-probe payload layout mirroring.
src/log.rs	Suppresses `unused_imports` when `tracing` is off and callsites are cfg’d out.
src/lib.rs	Minor module ordering change (`log` moved).
src/client/socket_manager.rs	Import ordering change for `crate::log` aliases.
src/client/mod.rs	Adds tokio client future-size witness test + budgets; minor import ordering change.
src/client/inner.rs	Import ordering change for `crate::log` aliases.
simple-someip-embassy-net/tests/loopback.rs	Formatting-only change to boxed future associated types.
examples/embassy_net_client/src/main.rs	Avoids large async stack frame via `Box::pin`; tweaks buffer allocation style.
docs/simple_someip/plans/baselines/pr0-size-baseline.md	Adds PR0 baseline capture data and regeneration instructions.
docs/simple_someip/plans/2026-06-09-phase22-125-memory-reduction-design.md	Adds design/plan doc for the #125 memory reduction PR stack.
CHANGELOG.md	Documents new infra (witness tests, probes, capture script, CI gates) and updated alloc-free server status.
.github/workflows/ci.yml	Adds `-Zbuild-std=core` thumb gate and extends alloc-symbol audit coverage.

💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.

JustinKovacich · 2026-06-10T21:36:37Z

+    let run_size = core::mem::size_of_val(&run_fut);
+    let _run_handle = tokio::spawn(run_fut);
+    client.bind_discovery().await.expect("bind_discovery");
+    let loop_size = max_spawned.load(Ordering::SeqCst);
+
+    println!("FUTURE_SIZE bm_client_run_future {run_size}");
+    println!("FUTURE_SIZE bm_client_socket_loop {loop_size}");
+    println!("FUTURE_SIZE bm_server_run_future {server_run_size}");
+
+    assert!(loop_size > 0, "spawner never received the socket loop");
+    assert!(
+        run_size <= BM_CLIENT_RUN_FUTURE_BUDGET,
+        "client run future grew: {run_size} B > budget {BM_CLIENT_RUN_FUTURE_BUDGET} B"
+    );
+    assert!(
+        loop_size <= BM_CLIENT_SOCKET_LOOP_BUDGET,
+        "socket loop future grew: {loop_size} B > budget {BM_CLIENT_SOCKET_LOOP_BUDGET} B"
+    );
+    assert!(
+        server_run_size <= BM_SERVER_RUN_FUTURE_BUDGET,
+        "server run future grew: {server_run_size} B > budget {BM_SERVER_RUN_FUTURE_BUDGET} B"
+    );
+    client.shut_down();
+}


Declining this one. The pool-budget header at the top of the file (lines ~46-53) already accounts for it: each Client::new_with_deps claims one ControlMessage + one ClientUpdate slot, both pools hold 4, and this witness is counted in the documented "3 client-constructing tests worst-case" — with the growth rule spelled out for whoever adds a 4th. Also, #[tokio::test] drops its runtime when the test fn returns, which drops the spawned run future at test exit — a claim can't outlive its test, so there's no cross-test leak beyond the budgeted concurrent overlap. Awaiting teardown with a timeout would add machinery the budget already makes unnecessary (and the sibling e2e test in this file uses the same spawn-then-abort shape).

JustinKovacich · 2026-06-10T21:36:39Z

+# `client,bare_metal` — the audited alloc-free combo (no `server`,
+# which pulls `extern crate alloc` per the lib.rs feature table).
+# `bare_metal` covers the codec-only FFI surface (matches halo PR


Fixed in ef988f7 — and this comment turned out to be one of four sites still claiming server pulls extern crate alloc: the lib.rs feature-table row, the extern crate alloc rationale block in lib.rs, and Cargo.toml's _alloc marker comment (which contradicted the server feature comment 20 lines below it). All four now state the post-#124 reality: _alloc is implied by std / embassy_channels only and gates the Arc-backed conveniences.

PR #124 made the server engine alloc-free but left four doc/comment sites claiming otherwise: the lib.rs feature table row, the extern-crate-alloc rationale block, Cargo.toml's _alloc marker comment, and tools/size_probe's dependency comment (the flagged one). All now state the post-#124 reality: _alloc comes from std/embassy_channels only, gating the Arc-backed conveniences. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>

JustinKovacich · 2026-06-29T20:39:35Z

Folded into main via the 0.8.0 consolidation (#138). Content verified present in main; closing and deleting the branch as part of post-0.8.0 cleanup.

JustinKovacich and others added 8 commits June 10, 2026 17:06

test(server): future-size witness for the tokio run future

1a6d84c

Same proxy-budget rules as the client witness. Also rustfmt reflow of the #124-era test bodies in this file. (#125 PR 0) Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>

JustinKovacich requested a review from Copilot June 10, 2026 21:14

Copilot started reviewing on behalf of JustinKovacich June 10, 2026 21:14 View session

Copilot AI reviewed Jun 10, 2026

View reviewed changes

JustinKovacich mentioned this pull request Jun 11, 2026

PR 1: #124 follow-ups — parsed union NonSdRequestCallback + doc trues-up (#125 stack) #129

Closed

JustinKovacich deleted the branch feat/no-alloc-bare-metal June 17, 2026 01:45

JustinKovacich closed this Jun 17, 2026

JustinKovacich reopened this Jun 17, 2026

JustinKovacich closed this Jun 29, 2026

JustinKovacich deleted the feature/pr0_measurement_harness branch June 29, 2026 20:40

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Uh oh!

PR 0: future-size measurement harness + build-std/nm CI gates (#125)#127

PR 0: future-size measurement harness + build-std/nm CI gates (#125)#127
JustinKovacich wants to merge 9 commits into
feat/no-alloc-bare-metalfrom
feature/pr0_measurement_harness

JustinKovacich commented Jun 10, 2026

Uh oh!

Copilot AI left a comment

Uh oh!

JustinKovacich Jun 10, 2026

Uh oh!

JustinKovacich Jun 10, 2026

Uh oh!

JustinKovacich commented Jun 29, 2026

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

2 participants

Uh oh!

Conversation

JustinKovacich commented Jun 10, 2026

Uh oh!

Copilot AI left a comment

Choose a reason for hiding this comment

Pull request overview

Reviewed changes

Uh oh!

JustinKovacich Jun 10, 2026

Choose a reason for hiding this comment

Uh oh!

JustinKovacich Jun 10, 2026

Choose a reason for hiding this comment

Uh oh!

JustinKovacich commented Jun 29, 2026

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

2 participants