Skip to content

Capture child-only JIT listings in frozen native replay - #694

Merged
thomhurst merged 1 commit into
mainfrom
issue-692-native-jit
Nov 13, 2026
Merged

thomhurst merged 1 commit into
mainfrom
issue-692-native-jit

Conversation

@thomhurst

@thomhurst thomhurst commented Nov 13, 2026 •

Copy link
Copy Markdown
Owner

The frozen native-parent launcher can replay original scoped dispatch images but cannot capture their compiled code without instrumenting the parent too. Add opt-in --jit-method / workflow jit_method using .NET's native JIT disassembly settings, with distinct child-only output per launch. Capture failures retain every outcome and fail visibly. Diagnostic measurements live under diagnosticResults and are explicitly excluded from timing/allocation evidence; default uninstrumented output is unchanged.

Reuses native DOTNET_JitDisasm, DOTNET_JitDisasmDiffable, DOTNET_JitStdOutFile, BenchmarkDotNet's existing native parent, the existing shell-quoted child environment wrapper, frozen inventories, selected-runner receipt validation and result parser. No native capability is duplicated. Missing/inlined/precompiled/interleaved code and unobserved tiers remain limitations: this does not reconstruct historical execution or prove equivalence.

Validation: all 25 test_frozen_bdn_control.py tests pass, including parent/child isolation, real shell quoting, mixed-mode rejection and preservation of all four launches after failed capture/processes. A small local .NET 10.0.14 ARM64 smoke fixture emitted a valid native JIT listing and printed 42. This validates runtime diagnostic support only; the original Linux x64 capture has not run. No benchmark or production tests required.

Validation sources, binaries and logs retained at /Users/tom/git/Relayer-evidence/pr-694/6179915c70bd98996612e37d0c4a01c55c5d562f/ while PR #694 remains open; all 32 files independently read-back hash/length verified against validation-sha256.json (SHA-256 3f09113c1ac012b91f92286f44a08fbd1e0e3d7828162753949263a395064ae6).

Per-message cost: unchanged, 0 B; internal investigation tooling only. No docs impact.

Closes #692

Next diagnostic for #613; does not close its historical throughput investigation. After this reaches trusted main, capture one native-parent scoped-dispatch ABBA with original run 43409168876, runtime 10.0.15 and method filter *DirectScopedDispatch*; preserve all raw listings/receipts/outcomes and compare observed code/tiers before choosing further work. Exclude instrumented timings and allocations.

@coderabbitai

coderabbitai Bot commented Nov 13, 2026 •

Copy link
Copy Markdown

Important

  • Trigger review

This repository does not receive automatic reviews because it has fewer than 10 stars.

Run configuration

  • Source mode: Auto
  • Repository: Unavailable
  • Branch: Unavailable
  • File: No configuration file from this repository applied
  • Commit: Unavailable
  • Layers: Organization UI
  • Review profile: CHILL
  • Plan: Advanced
  • Run ID: 4cebc6fa-e504-44ca-bf74-05b3db0f277b
  • Autofix · Keep fixing CodeRabbit findings and required CI, and resolving merge conflicts

Thanks for using CodeRabbit! It's free for OSS, and your support helps us grow. If you like it, consider giving us a shout-out.

Share

@thomhurst

Copy link
Copy Markdown
Owner Author

@codex review

@github-actions

Copy link
Copy Markdown

Review of #694: no blocking issues found.

  • Reuse: the change uses the runtime's own DOTNET_JitDisasm, DOTNET_JitDisasmDiffable and DOTNET_JitStdOutFile settings. It reuses the existing shell-quoted child environment wrapper, so the managed parent stays uninstrumented. It does not add a parallel capture mechanism.
  • Correctness:
    • --jit-method is rejected without the native parent, with a trace tool or allocation profiler, or with an empty or NUL-containing filter.
    • Each launch gets its own N-side-jit.asm file.
    • A missing or invalid listing header sets jitError and fails the run, and every launch outcome is still kept.
    • Timings go under diagnosticResults and are marked excluded, so they cannot be mistaken for evidence.
    • protocol.instrumented is true when --jit-method is set, and default output is unchanged.
  • Workflow: the input arrives through an env variable, so there is no shell injection. It is appended to control_args as a quoted array element.
  • Tests: the new tests cover the none, missing, empty, invalid and launch-failure cases. They also check parent/child environment isolation, real shell quoting of a hostile filter, rejection of mixed instrumentation, and that all four outcomes are kept.
  • Performance and docs: this is CI tooling only, so there is no per-message cost and no docs impact.

Minor, non-blocking: the listing check only looks for the ; Assembly listing for method header. That is enough for "something was captured", and the PR already states the completeness limits.

@chatgpt-codex-connector

chatgpt-codex-connector Bot commented Nov 13, 2026 •

Copy link
Copy Markdown
Codex Review Summary: ✅ Completed

This comment shows the latest Codex review activity on this pull request.

Review Status Commit Review trigger
📝 Code Review ✅ Completed 2026-11-13T14:05:37.232173Z 6179915 Draft marked ready
ℹ️ About Codex in GitHub

Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you

  • Open a pull request for review
  • Mark a draft as ready
  • Comment "@codex review" or "@codex security review".

Codex reacts with 👀 while any review is running, comments if it has suggestions, and reacts with 👍 once all reviews finish with no findings.

@chatgpt-codex-connector

Copy link
Copy Markdown

Codex Review: Didn't find any major issues. Another round soon, please!

Reviewed commit: 6179915c70

ℹ️ About Codex in GitHub

Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you

  • Open a pull request for review
  • Mark a draft as ready
  • Comment "@codex review".

If Codex has suggestions, it will comment; otherwise it will react with 👍.

Codex can also answer questions or update the PR. Try commenting "@codex address that feedback".

@thomhurst

Copy link
Copy Markdown
Owner Author

Current-head review disposition: Claude CLEAR and Codex completed without findings both cover 6179915. Claude’s minor listing-header observation is non-blocking: the capture intentionally establishes that a listing exists; the PR documents completeness limits. No code change is needed.

Reverified all 32 archived validation files against their lengths and SHA-256 inventory, and verified the manifest hash recorded in the PR body. The durable location remains /Users/tom/git/Relayer-evidence/pr-694/6179915c70bd98996612e37d0c4a01c55c5d562f/ for the lifetime of this open PR. Original disposable worktree is already removed; this verification used the previously independently verified archive and its recorded manifest. No tests rerun because the reviewed code is unchanged.

@thomhurst
thomhurst marked this pull request as ready for review November 13, 2026 14:04
@thomhurst
thomhurst enabled auto-merge (squash) November 13, 2026 14:04
@thomhurst
thomhurst merged commit 8a4ccbb into main Nov 13, 2026
25 checks passed
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.

Capture child-only JIT disassembly in frozen native-parent replay

1 participant