Refactor trace builder

[MauroToscano]

No description provided.

[github-actions]

Codex Code Review

No issues found in the PR diff.

I reviewed only the changes between b1f365b2...912830e4 and did not find security vulnerabilities, logic regressions, or significant performance problems introduced by this refactor. The phase extraction into collect_all_ops and build_traces appears behavior-preserving, including legacy from_logs behavior (elf=None keeps PAGE-related work disabled).

Residual risk: I did not run tests in this review, so runtime equivalence still depends on existing test coverage for both from_elf_and_logs and from_logs paths.

[claude]

Review: Refactor trace builder

Clean deduplication — the logic is semantically equivalent in both call paths and the elf=None guard correctly preserves the legacy behaviour (no PAGE bitwise lookups, empty page tables).

One pre-existing code smell made visible by the unification: num_padding_rows is computed twice inside build_traces with identical expressions (once for collect_byte_check_ops_for_padding, once for decode_lookups). Inline suggestion posted.

No security issues, logic bugs, or regressions found.