research: fix repeat sample manifest projection

[ErenAri]

Summary

Fix the first manual research-repeat-v1 execution failure without changing the
frozen pilot v1 evidence or the repeat sample.

Failed run

• workflow run: 35443831085
• source commit: 804c021ddc47a1d6b479a70690af4032c8dd79db
• artifact: 10585281049
• artifact SHA-256:
  b32ca592c79a01f4c70cf0d85607dfe07a0efe23deddee4dd817187483ac02fa

The workflow successfully materialized the frozen inputs and the three command
mode tuples completed stably, but every libbpf-backed tuple failed before VM
execution.

The repeat runner creates a one-profile matrix for each sampled tuple. Frozen
v1 libbpf manifests intentionally list all ten study profiles under
required_profiles. BPFCompat correctly rejects a matrix that omits any
manifest-required profile, so those sampled runs exited before producing a
report:

bpfcompat test failed: manifest required profile "..." is not present in matrix

The failed staging run therefore observed only 9/21 planned repeats. Its 9
completed command-mode observations are diagnostic only; the run is not promoted
as repeat-study evidence.

Fix

For libbpf-backed repeat tuples only:

• verify the exact frozen source manifest Git blob recorded in
  research/corpus/v1/study-plan.json;
• create a repeat-only manifest projection whose sole semantic scheduling change
  is narrowing required_profiles to the sampled profile;
• retain the projected manifest plus metadata containing:
  • source Git blob;
  • source SHA-256;
  • original required-profile list;
  • selected profile;
  • projected SHA-256;
  • projection rule;
• preserve all program/attach/validation fields unchanged;
• record the projection helper SHA-256 in repeat provenance.

This does not rewrite the frozen corpus manifest. It adapts only its matrix
membership assertion for a purposefully sampled single-profile rerun.

Regression coverage

The repeat preflight now:

• compiles and self-tests the projection helper;
• verifies every sampled libbpf case has a frozen manifest identity;
• recomputes the source manifest Git blob and compares it with the frozen study
  plan;
• verifies the sampled profile is present in the frozen required-profile list.

After merge

Dispatch the repeat workflow again from main:

gh workflow run research-repeat-v1.yml --repo Kernel-Guard/bpfcompat --ref main

Only a complete post-fix run should be considered for the repeat-stability
dataset.

[Copilot]

Copilot was unable to review this pull request because the user who requested the review has reached their quota limit.

[coderabbitai]

Warning

Review limit reached

Next included review available in 12 minutes.

Check out review usage here.

View limit details

Limit details: You’ve used the included review currently available.

You've used all free OSS reviews for now. Wait for the free limit to reset to keep reviewing this public repository.

Learn how review limits work.

Review configuration:

⚙️ Run configuration

Configuration used: Organization UI

Review profile: CHILL

Plan: Advanced

Run ID: 461baa12-5324-42ba-8f10-436ff11dcf44

📥 Commits

Reviewing files that changed from the base of the PR and between 804c021 and 8c49776.

📒 Files selected for processing (4)

• .github/workflows/research-repeat-v1.yml
• research/repeat/v1/README.md
• scripts/research/project-repeat-manifest-v1.py
• scripts/research/run-repeat-v1.sh

---

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

• X
• Mastodon
• Reddit
• LinkedIn

_{Comment @coderabbitai help to get the list of available commands.}