julia-debugging

name: julia-debugging description: Diagnose, reproduce, and fix Julia errors, stack traces, package-loading failures, logging issues, type-instability problems, test failures, Pkg/environment issues, notebook failures, model failures, and performance regressions.

Julia Debugging

Use this skill to debug Julia code systematically. Prefer official Julia and Pkg documentation for language behavior, package loading, stack traces, logging, profiling, performance, tests, and environments.

Read ../../references/julia-official-workflow.md and ../../references/analysis-workflow-handoff.md for substantial work.

Debug Loop

1. Capture the exact failing command and smallest useful output.
2. Reproduce with the narrowest command possible.
3. Confirm Julia version, active project, Project.toml, Manifest.toml, imports, and package versions.
4. Classify the failure: exception, wrong result, package/environment, precompilation, type stability, allocation/performance, notebook, model/data, or external dependency.
5. Start with the exception type and first useful project-owned stack frame.
6. Form one hypothesis at a time.
7. Apply the smallest fix.
8. Rerun the narrow failing command.
9. Rerun broader verification after the narrow check passes.

Standard Surfaces

• Stack traces and minimal examples for correctness bugs.
• Pkg.status(), Pkg.instantiate(), and active-project checks for environment failures.
• Test, @test_throws, and @test_logs for behavior and log assertions.
• @debug, @info, JULIA_DEBUG, and Logging for runtime state.
• @code_warntype, allocation checks, Profile, JET, and BenchmarkTools for type/performance questions.

Common Failure Modes

• UndefVarError: missing import, scope issue, renamed API, or Unicode lookalike.
• MethodError: wrong argument types, missing module qualification, or API mismatch.
• Package load/precompile failure: wrong active environment, stale manifest, missing dependency, or compatibility bounds.
• Slow code: global state, type instability, abstract containers, allocation-heavy hot paths, or top-level loops.

Required Handoff

If a failure blocks validation, hand off once to debug-test-failures from analysis-engineering-workflow. If the root cause changes architecture, security, data, or AI/ML risk, hand off to framework-compliance-review before finalizing.