Skip to content

pkg/aflow: expect reproducer instability / unrelated kernel crashes #7093

@a-nogikh

Description

@a-nogikh

Originally brought up in #7081 (comment)


As part of the patching and the new patch-iteration workflow, we re-reproduce the original bug before working on it. That is good per se, but may be problematic:

  • Some reproducers are somewhat flaky. What if it didn't crash by chance?
  • Some reproducers may trigger several different bugs, which may require different fixes.

It feels like crash.Reproduce should do not one, but N repro attempts. And then we should ask some sub-agent to determine which of the results we should pass on?

When we know the original bug title / crash report, it would be nice to use it as well.

Metadata

Metadata

Assignees

No one assigned

    Labels

    AI patchingFeature requests and bugs related to AI-based kernel bug fix generation.enhancement

    Type

    No type

    Projects

    No projects

    Milestone

    No milestone

    Relationships

    None yet

    Development

    No branches or pull requests

    Issue actions