Histogram Cleanup

Code Health: Histogram Cleanup

Identify and safely remove "dead code" associated with expired histograms. This includes removing the recording calls in C++/Java, cleaning up the metadata in histograms.xml, and addressing any dependent tests.

Overview

Expired histograms that are not intentionally kept for diagnostics represent technical debt. Act as an expert Chromium contributor specializing in Metrics to clean up these resources while ensuring no test regressions occur.

Goal: Remove expired Chromium histograms (dead metrics) from XML files and all code recording sites.

Relevant Resources & Style Guides

• Chromium Metrics - Histograms Guide
• Implementation Patterns: patterns.md (C++, Java, XML examples)
• Analysis Guidelines: analysis_guidelines.md (Safety checks and effort estimation)
• Bug Discovery: bug_discovery.md (Finding existing trackers)
• Discovery Script: find_candidates.py (Identify expired histograms)
• Automated Review Protocol: automated_review.md

Scope & Proactivity

• Primary Scope: Focus exclusively on the identified histogram by default. Do NOT suggest removing random additional histograms across the file or component just because they are expired.
• Exception for Coupled/Shared Recording Code: If other histograms share the exact same recording logic, helper methods, or calculated variables (e.g., co-located in the same helper function or branching from the same condition) as the target histogram, their removal MAY be bundled into the same cleanup plan IF AND ONLY IF:
  1. Expiry Verification: Confirm in histograms.xml that the co-located histograms are also expired and lack the <expired_intentionally> tag.
  2. Safety Verification: Perform the same safety checks (no test dependencies, no external repo references) for each bundled histogram.
  3. Atomic Benefit: Removing them together cleanly eliminates shared boilerplate (e.g., timer calculations, string building, parameter passing) that would otherwise be left as dead or awkward code.
• Related Dead Code: If dead code is found (e.g., constants, enums, or helper methods, or calculated variables like timers) that is directly related to the histogram(s) being removed, include its removal in the cleanup plan. Present these as part of the primary task, not as separate "proactive suggestions."

Workflow

[!IMPORTANT] Execution Protocol: Execute all steps sequentially one by one. Do not skip any step. Do not use edit-code or grep. Use rg (ripgrep) for searches.

Step 1: Workspace Preparation

Follow the workspace preparation steps in workspace_preparation.md to ensure a clean and updated environment.

Step 2: Discovery & Batch Selection

Follow the Discovery & Batch Selection workflow.

Step 3: Deep Dive & Safety Analysis

1. Comprehensive Analysis: Delegate the entire deep dive to the generalist sub-agent with this exact prompt:

   "Read references/analysis_guidelines.md and follow the 'Generalist Deep Dive Prompt' instructions for the histogram <HistogramName>. ALL read-only discovery (including rg and cs) is pre-authorized; DO NOT ask for permission. Assume rg is available in the environment."

2. Present Findings & Evaluate Confidence: Evaluate the Confidence Score returned by the generalist.

   • If Confidence >= 9: Inform the user: "Confidence is high ([X]/10). Proceeding with cleanup based on this plan: [Removal Plan Summary]." Output this message, then proceed directly to the Refactoring & Implementation phase. Do NOT ask for permission.
   • If Confidence is between 1 and 8: Present the findings and prompt the user using ask_user (type='choice'):
     • header: "Confidence Check"
     • question: "This histogram is safe to remove from [Files] and [Tests]. My confidence for this cleanup is [X]/10 because [Justification]. Shall I proceed with the cleanup diff?"
     • options:
       • label: "Proceed with Diff", description: "Generate and apply the cleanup changes"
       • label: "Discard", description: "Discard this candidate and stop."
     • Action based on selection:
       • If "Proceed with Diff": Proceed to the Refactoring & Implementation phase.
       • If "Discard": Stop the workflow.
   • If Confidence is 0: Inform the user: "Confidence is zero (0/10) because [Justification]. An alternate expired histogram will be found." Immediately restart the workflow from the Discovery & Batch Selection phase to identify a different candidate. Ensure the same histogram is not selected again in this session.