debugging
Use when encountering any bug, test failure, solver issue, or unexpected behavior, before proposing fixes
- License 许可
- Apache-2.0
- 最近更新
- 2026-11-12
- Stars
- ★ 7
- Forks
- 2
01 前言 Front Matter
- name
- debugging
- agents
- []
- surface
- public
- description
- Use when encountering any bug, test failure, solver issue, or unexpected behavior, before proposing fixes
02 资源文件 2
- 参考文档defense-in-depth.md3.6 KB
- 参考文档root-cause-tracing.md5.2 KB
03 SKILL.md 正文
Systematic Debugging
Overview
Random fixes waste time and create new bugs. Quick patches mask underlying issues.
Core principle: ALWAYS find root cause before attempting fixes. Symptom fixes are failure.
Violating the letter of this process is violating the spirit of debugging.
The Iron Law
NO FIXES WITHOUT ROOT CAUSE INVESTIGATION FIRST
If you haven't completed Phase 1, you cannot propose fixes.
When to Use
Use for ANY technical issue:
- Test failures
- Bugs in production
- Unexpected behavior
- Performance problems
- Build failures
- Integration issues
Solver and optimizer issues:
- Stagnation: cost not decreasing over iterations, optimizer stuck at local minimum
- Constraint violation: dynamics constraints not satisfied, trajectory drifting
- Integrator divergence: NaN/Inf appearing in trajectories, timestep too large
- Infeasibility: optimizer reports problem infeasible, conflicting constraints
Use this ESPECIALLY when:
- Under time pressure (emergencies make guessing tempting)
- "Just one quick fix" seems obvious
- You've already tried multiple fixes
- Previous fix didn't work
- You don't fully understand the issue
Don't skip when:
- Issue seems simple (simple bugs have root causes too)
- You're in a hurry (rushing guarantees rework)
- Manager wants it fixed NOW (systematic is faster than thrashing)
Quantum-Specific Sanity Checks
Before diving into debugging quantum optimal control code, verify these fundamentals:
- Hamiltonian Hermiticity - Is $H = H^\dagger$? Non-Hermitian generators produce non-unitary propagators.
- Control bounds - Are controls within physically realizable bounds? Check
a_boundsin the trajectory. - Time grid resolution - Is $\Delta t$ small enough for the dynamics? If $|H| \cdot \Delta t \gg 1$, the integrator may diverge.
- Operator dimensions - Are all operators, states, and trajectories dimensionally consistent? Mismatched Hilbert space dimensions cause silent errors.
- Basis ordering - Is the computational basis consistent between the Hamiltonian, initial state, and target gate?
- Embedded operator subspace - For
EmbeddedOperator, is the computational subspace correctly specified?
Solver Diagnosis Patterns
Stagnation (cost plateau)
- Symptoms: Cost decreases initially then flatlines; optimizer runs to max iterations without convergence.
- Common causes: Poor initial guess, insufficient control authority, local minimum, objective landscape too flat.
- Investigation: Plot cost vs iteration. Check gradient norms (near-zero = stuck). Try different initial trajectories. Verify objective is correctly formulated.
Constraint Violation (dynamics drift)
- Symptoms:
dynamicsconstraint violation remains large; trajectory does not satisfy equations of motion. - Common causes: Integrator mismatch (BilinearIntegrator vs SplineIntegrator), timestep too large, incorrect Hamiltonian coupling structure.
- Investigation: Check constraint violation magnitude at each timestep. Verify the integrator matches the control parameterization. Use
duration(pulse)notsum(get_timesteps)for duration checks.
Integrator Divergence (NaN/Inf)
- Symptoms: NaN or Inf values appear in state or control trajectories mid-solve.
- Common causes: Timestep too large for operator norms, stiff dynamics, numerical overflow in matrix exponential.
- Investigation: Reduce $\Delta t$. Check $|H| \cdot \Delta t$. Examine which timestep first produces NaN. Verify no division by zero in custom objectives.
The Vacuous Solve (silent control-drop) — 2026-10, measured
- Symptoms: The solver consumes thousands of iterations, the variables MOVE (extracted knots differ from the seed), but the internal objective/fidelity is bit-identical to the seed value across every option, setting, and formulation change (max_iter, mu_strategy, member count). Post-solve quality is entirely explained by regularizer drift.
- Common causes: The problem's dynamics integrator never received the controls — template
integrator auto-resolution silently resolving to a control-free path (Piccolo 2.2.0
measured:
SplineIntegratorauto-wiring left the objective constant); or reading a STALEfidelity(qtraj)(the construction-time ODE solution) that masks movement in the other direction. - Investigation: Evaluate the objective at the SEED and assert the post-solve value has MOVED (a descent method cannot end where it started, bit-identical). Pass integrators explicitly. Verify extracted solutions through independent scorers, never the problem's internal readouts. Provenance: harmoniqs/diraq-esr-realism 2026-10-23 overnight §4.
The Silent Edit (no-op transformations) — six measured instances
- Symptoms: An edit reports success but behavior never changes; archaeology shows the pattern never matched (trailing comma, already-replaced form, reverted block). Worst case: stacked no-ops make the file's edit history untrustworthy and force a clean rewrite.
- Common causes:
str.replace/sedsucceed on no-match by design; kwargs that miss their consumer get swallowed; later edits revert earlier ones through shared patterns; LLM memory of what landed is not evidence. - Investigation/Prevention: Every programmatic edit ASSERTS its pattern matched
(
assert old in s) and the critical line is grep-verified after the write; multi-edit confusion resolves by rewriting clean with git as arbiter; numbers are never hand-copied between files when a formula can be evaluated (machine-pinned literals, rtol ≤ 1e-9). Provenance: six instances in harmoniqs/diraq-esr-realism loops 4–10.
The Silent OOM (death right after initialization)
- Symptoms: A process dies immediately after "optimizer initialization complete" with no error in any log — the OOM killer SIGKILLs without a traceback.
- Common causes: Dense interior-point Hessians over unitary trajectories scale as (2·d²·N)² — a 36-dim member block at N=21 needs ~25 GB. Member counts are memory decisions.
- Investigation: Compute the Hessian memory before launching; prefer small-d blocks.
Check
dmesg/journalctl for oom-kill records where readable.
Background process death at harness boundaries
- Symptoms:
nohup … &jobs are gone by the next interaction; logs end mid-init. - Cause: Sandboxed tool environments reap the process group when the call returns.
- Fix:
setsid nohup … &detaches the session (measured both ways, 2026-10-23).
Infeasibility (conflicting constraints)
- Symptoms: Solver reports "infeasible" or "locally infeasible"; no solution found.
- Common causes: Control bounds too tight for the target gate, conflicting time and fidelity constraints, target gate unreachable in given time.
- Investigation: Relax bounds and re-solve to identify which constraint is binding. Check if the target is achievable with unconstrained controls. Verify gate time is sufficient.
The Four Phases
You MUST complete each phase before proceeding to the next.
Phase 1: Root Cause Investigation
BEFORE attempting ANY fix:
-
Read Error Messages Carefully
- Don't skip past errors or warnings
- They often contain the exact solution
- Read stack traces completely
- Note line numbers, file paths, error codes
-
Reproduce Consistently
- Can you trigger it reliably?
- What are the exact steps?
- Does it happen every time?
- If not reproducible, gather more data, don't guess
-
Check Recent Changes
- What changed that could cause this?
- Git diff, recent commits
- New dependencies, config changes
- Environmental differences
-
Gather Evidence in Multi-Component Systems
WHEN system has multiple components (optimizer, integrator, objective, constraints):
BEFORE proposing fixes, add diagnostic instrumentation:
For EACH component boundary: - Log what data enters component - Log what data exits component - Verify environment/config propagation - Check state at each layer Run once to gather evidence showing WHERE it breaks THEN analyze evidence to identify failing component THEN investigate that specific componentExample (quantum control pipeline):
# Layer 1: Problem setup println("=== Hamiltonian dimensions: $(size(H_drift)) ===") println("=== Number of controls: $(length(a_labels)) ===") # Layer 2: Trajectory println("=== Trajectory timesteps: $(length(times)) ===") println("=== Control bounds: $(a_bounds) ===") # Layer 3: Integrator println("=== Integrator type: $(typeof(integrator)) ===") println("=== Max ||H||*dt: $(maximum(norms .* dts)) ===") # Layer 4: Optimizer output println("=== Final cost: $(result.objective) ===") println("=== Constraint violation: $(result.constraint_violation) ===")This reveals: Which layer fails (setup OK, integrator OK, optimizer diverges)
-
Trace Data Flow
WHEN error is deep in call stack:
See
root-cause-tracing.mdin this directory for the complete backward tracing technique.Quick version:
- Where does bad value originate?
- What called this with bad value?
- Keep tracing up until you find the source
- Fix at source, not at symptom
Phase 2: Pattern Analysis
Find the pattern before fixing:
-
Find Working Examples
- Locate similar working code in same codebase
- What works that's similar to what's broken?
-
Compare Against References
- If implementing pattern, read reference implementation COMPLETELY
- Don't skim - read every line
- Understand the pattern fully before applying
-
Identify Differences
- What's different between working and broken?
- List every difference, however small
- Don't assume "that can't matter"
-
Understand Dependencies
- What other components does this need?
- What settings, config, environment?
- What assumptions does it make?
Phase 3: Hypothesis and Testing
Scientific method:
-
Form Single Hypothesis
- State clearly: "I think X is the root cause because Y"
- Write it down
- Be specific, not vague
-
Test Minimally
- Make the SMALLEST possible change to test hypothesis
- One variable at a time
- Don't fix multiple things at once
-
Verify Before Continuing
- Did it work? Yes, proceed to Phase 4
- Didn't work? Form NEW hypothesis
- DON'T add more fixes on top
-
When You Don't Know
- Say "I don't understand X"
- Don't pretend to know
- Ask for help
- Research more
Phase 4: Implementation
Fix the root cause, not the symptom:
-
Create Failing Test Case
- Simplest possible reproduction
- Automated test if possible
- One-off test script if no framework
- MUST have before fixing
- Use the
tddskill for writing proper failing tests
-
Implement Single Fix
- Address the root cause identified
- ONE change at a time
- No "while I'm here" improvements
- No bundled refactoring
-
Verify Fix
- Test passes now?
- No other tests broken?
- Issue actually resolved?
-
If Fix Doesn't Work
- STOP
- Count: How many fixes have you tried?
- If < 3: Return to Phase 1, re-analyze with new information
- If >= 3: STOP and question the architecture (step 5 below)
- DON'T attempt Fix #4 without architectural discussion
-
If 3+ Fixes Failed: Question Architecture
Pattern indicating architectural problem:
- Each fix reveals new shared state/coupling/problem in different place
- Fixes require "massive refactoring" to implement
- Each fix creates new symptoms elsewhere
STOP and question fundamentals:
- Is this pattern fundamentally sound?
- Are we "sticking with it through sheer inertia"?
- Should we refactor architecture vs. continue fixing symptoms?
Discuss with your human partner before attempting more fixes
This is NOT a failed hypothesis - this is a wrong architecture.
Red Flags - STOP and Follow Process
If you catch yourself thinking:
- "Quick fix for now, investigate later"
- "Just try changing X and see if it works"
- "Add multiple changes, run tests"
- "Skip the test, I'll manually verify"
- "It's probably X, let me fix that"
- "I don't fully understand but this might work"
- "Pattern says X but I'll adapt it differently"
- "Here are the main problems: [lists fixes without investigation]"
- Proposing solutions before tracing data flow
- "One more fix attempt" (when already tried 2+)
- Each fix reveals new problem in different place
ALL of these mean: STOP. Return to Phase 1.
If 3+ fixes failed: Question the architecture (see Phase 4, step 5)
Your Human Partner's Signals You're Doing It Wrong
Watch for these redirections:
- "Is that not happening?" - You assumed without verifying
- "Will it show us...?" - You should have added evidence gathering
- "Stop guessing" - You're proposing fixes without understanding
- "Ultrathink this" - Question fundamentals, not just symptoms
- "We're stuck?" (frustrated) - Your approach isn't working
When you see these: STOP. Return to Phase 1.
Common Rationalizations
| Excuse | Reality |
|---|---|
| "Issue is simple, don't need process" | Simple issues have root causes too. Process is fast for simple bugs. |
| "Emergency, no time for process" | Systematic debugging is FASTER than guess-and-check thrashing. |
| "Just try this first, then investigate" | First fix sets the pattern. Do it right from the start. |
| "I'll write test after confirming fix works" | Untested fixes don't stick. Test first proves it. |
| "Multiple fixes at once saves time" | Can't isolate what worked. Causes new bugs. |
| "Reference too long, I'll adapt the pattern" | Partial understanding guarantees bugs. Read it completely. |
| "I see the problem, let me fix it" | Seeing symptoms != understanding root cause. |
| "One more fix attempt" (after 2+ failures) | 3+ failures = architectural problem. Question pattern, don't fix again. |
Quick Reference
| Phase | Key Activities | Success Criteria |
|---|---|---|
| 1. Root Cause | Read errors, reproduce, check changes, gather evidence | Understand WHAT and WHY |
| 2. Pattern | Find working examples, compare | Identify differences |
| 3. Hypothesis | Form theory, test minimally | Confirmed or new hypothesis |
| 4. Implementation | Create test, fix, verify | Bug resolved, tests pass |
When Process Reveals "No Root Cause"
If systematic investigation reveals issue is truly environmental, timing-dependent, or external:
- You've completed the process
- Document what you investigated
- Implement appropriate handling (retry, timeout, error message)
- Add monitoring/logging for future investigation
But: 95% of "no root cause" cases are incomplete investigation.
Supporting Techniques
These techniques are part of systematic debugging and available in this directory:
root-cause-tracing.md- Trace bugs backward through call stack to find original triggerdefense-in-depth.md- Add validation at multiple layers after finding root cause
Related skills:
- tdd - For creating failing test case (Phase 4, Step 1)
- test - For running Julia package tests to verify fixes
Real-World Impact
From debugging sessions:
- Systematic approach: 15-30 minutes to fix
- Random fixes approach: 2-3 hours of thrashing
- First-time fix rate: 95% vs 40%
- New bugs introduced: Near zero vs common
04 同仓库更多 Skills
malformed
Frontmatter that parses as YAML but is missing the required name field.
duplicated
Second skill carrying the duplicated frontmatter name.
duplicated
First skill carrying the duplicated frontmatter name.
drifted
A skill whose API claims have drifted from the fixture packages.
clean
A clean skill whose every extracted claim resolves against the fixture packages.
broken-paths
A skill with a broken relative link and a broken cited package path.