Fix backtrace deduplication for Clang debug info

[Hzfengsy]

Deduplicate stack frames by PC address to handle Clang's DWARF generation issues where multiple virtual frames are reported for the same program counter with incorrect source locations. Keep only the last frame for each unique PC as it typically represents the actual call site.

Also add "object.h" to the filename filter list to exclude Python object header frames from backtraces.

[gemini-code-assist]

Summary of Changes

Hello @Hzfengsy, I'm Gemini Code Assist¹! I'm currently reviewing this pull request and will post my feedback shortly. In the meantime, here's a summary to help you and other reviewers quickly get up to speed!

This pull request enhances the backtrace generation logic by introducing a frame deduplication strategy to mitigate issues with Clang's debug information, specifically when dealing with inlined functions. Additionally, it refines the backtrace filtering to exclude irrelevant Python internal frames, leading to more accurate and readable stack traces.

Highlights

• Backtrace Deduplication: Implemented a new mechanism to deduplicate stack frames based on their Program Counter (PC) address. This addresses an issue with Clang's DWARF generation where multiple 'virtual' frames might be reported for the same PC. The logic retains only the last frame for each unique PC, which is typically the most accurate call site.
• Filename Filter Enhancement: Added 'object.h' to the list of filenames to be filtered out from backtraces. This change aims to exclude Python object header frames, improving the clarity of generated stack traces.

🧠 New Feature in Public Preview: You can now enable Memory to help Gemini Code Assist learn from your team's feedback. This makes future code reviews more consistent and personalized to your project's style. Click here to enable Memory in your admin console.

Using Gemini Code Assist

The full guide for Gemini Code Assist can be found on our documentation page, here are some quick tips.

Invoking Gemini

You can request assistance from Gemini at any point by creating a comment using either /gemini <command> or @gemini-code-assist <command>. Below is a summary of the supported commands on the current page.

Feature	Command	Description
Code Review	/gemini review	Performs a code review for the current pull request in its current state.
Pull Request Summary	/gemini summary	Provides a summary of the current pull request in its current state.
Comment	@gemini-code-assist	Responds in comments when explicitly tagged, both in pull request comments and review comments.
Help	/gemini help	Displays a list of available commands.

Customization

To customize Gemini Code Assist for GitHub experience, repository maintainers can create a configuration file and/or provide a custom code review style guide (such as PEP-8 for Python) by creating and adding files to a .gemini/ folder in the base of the repository. Detailed instructions can be found here.

Limitations & Feedback

Gemini Code Assist may make mistakes. Please leave feedback on any instances where its feedback is incorrect or counter productive. You can react with 👍 and 👎 on @gemini-code-assist comments. If you're interested in giving your feedback about your experience with Gemini Code Assist for Github and other Google products, sign up here.

You can also get AI-powered code generation, chat, as well as code reviews directly in the IDE at no cost with the Gemini Code Assist IDE Extension.

Footnotes

1. Review the Privacy Notices, Generative AI Prohibited Use Policy, Terms of Service, and learn how to configure Gemini Code Assist in GitHub here. Gemini can make mistakes, so double check it and use code with caution. ↩

[gemini-code-assist]

Code Review

This pull request introduces a fix for backtrace deduplication to handle issues with Clang's debug information, and adds object.h to the filename filter list. The changes look good, but I have one suggestion to improve the robustness of the deduplication logic. The current implementation assumes that frames with the same program counter are always consecutive, which could be brittle. My suggestion provides a more robust solution that handles non-consecutive duplicate frames as well.

[gemini-code-assist]

The current deduplication logic assumes that stack frames with the same program counter (PC) are always consecutive in the reversed_frames vector. While this might be true for the specific Clang DWARF issue this change is addressing, this assumption makes the implementation fragile. If frames with the same PC were to appear non-consecutively for any reason, this logic would fail to deduplicate them correctly.

A more robust approach would be to iterate through the frames from most-recent to least-recent and use a set to track seen PCs. This ensures that only the most recent frame for each unique PC is kept, regardless of its position in the stack trace. This would correctly handle both consecutive and non-consecutive duplicate frames.

Note: This change would require adding #include <unordered_set> at the top of the file.

  std::vector<StackFrame> deduplicated_frames;
  std::unordered_set<uintptr_t> seen_pcs;
  // Iterate from most-recent to least-recent to keep the most recent frame for each PC.
  for (auto it = reversed_frames.rbegin(); it != reversed_frames.rend(); ++it) {
    const auto& frame = *it;
    if (frame.pc == 0 || seen_pcs.find(frame.pc) == seen_pcs.end()) {
      deduplicated_frames.push_back(frame);
      if (frame.pc != 0) {
        seen_pcs.insert(frame.pc);
      }
    }
  }
  // Reverse to get the correct order for display (least-recent to most-recent).
  std::reverse(deduplicated_frames.begin(), deduplicated_frames.end());

[Copilot]

Pull request overview

This PR addresses Clang's DWARF debug information generation issues by implementing PC-based deduplication of stack frames in backtraces, and adds Python's object.h to the filename filter list.

• Implements deduplication logic to handle multiple virtual frames reported for the same program counter
• Adds "object.h" to the backtrace filename filter to exclude Python C API header frames

---

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

[Copilot]

The deduplication implementation only removes consecutive frames with the same PC address. While this may be sufficient for Clang's inline frame generation (which likely produces consecutive duplicates), the PR description states "Deduplicate stack frames by PC address" and "keep only the last frame for each unique PC", which suggests complete deduplication is intended. If non-consecutive frames can have the same PC, consider using an unordered_map to track the last frame for each unique PC value. Otherwise, clarify in the comments that only consecutive duplicates are handled.

[Copilot]

The new PC-based deduplication logic lacks test coverage. Consider adding a test case that verifies frames with duplicate PC addresses are correctly deduplicated, especially to ensure the behavior works as intended for Clang's DWARF generation issues. This could be done by creating a mock scenario with duplicate PC values or by testing with actual Clang-compiled code that exhibits the issue.

[Copilot]

The comment "Python object.h" could be more specific to clarify this filters out Python C API header frames, not a local header file. Consider changing to "Python C API object.h" or "CPython object.h" for clarity.

Suggested change

	"object.h", // Python object.h
	"object.h", // Python C API object.h (CPython header)