git-rebase（git 变基）

Git Rebase with Intelligent Conflict Resolution

Quick Start

For most rebases with multiple commits, use the squash-first strategy to resolve conflicts only once:

# Step 1: Backup current state
bash scripts/pre-rebase-backup.sh

# Step 2: Squash commits (interactive rebase on current branch)
git rebase -i $(git merge-base HEAD origin/main)

# Step 3: Rebase onto target
git rebase origin/main

# Step 4: If conflicts, resolve them once (see workflow below)
# Then continue: git rebase --continue

# Step 5: Force push safely
git push origin $(git rev-parse --abbrev-ref HEAD) --force-with-lease

This approach resolves conflicts once instead of per-commit, saving time and mental overhead.

Core Workflow: Conflict Analysis & Resolution

Copy this checklist and mark progress:

Rebase Workflow:
- [ ] Step 1: Create safety backup
- [ ] Step 2: Fetch latest from target branch
- [ ] Step 3: Analyze conflict scope
- [ ] Step 4: Choose resolution strategy
- [ ] Step 5: Apply conflict resolutions
- [ ] Step 6: Validate merged code
- [ ] Step 7: Run tests
- [ ] Step 8: Force push safely

Step 1: Create Safety Backup

ALWAYS do this first. If rebase goes wrong, you can recover:

# Use the bundled script
bash scripts/pre-rebase-backup.sh

# Or manually create timestamped backup branch
TIMESTAMP=$(date +%Y%m%d_%H%M%S)
git branch backup-rebase-$TIMESTAMP

# Alternative: Create temporary ref to your current commit
git reflog  # Note your current HEAD SHA for manual recovery

This costs nothing and saves hours of work if something goes wrong.

Step 2: Fetch Latest Changes

Ensure you have the most recent remote state:

# Fetch without modifying local branches
git fetch origin

# View the divergence
git log --oneline origin/main..HEAD  # Your commits
git log --oneline HEAD..origin/main  # New commits on main

Understand how many commits you're rebasing and how much main has changed.

Step 3: Analyze Conflict Scope

Before starting the rebase, predict conflicts:

# See which files you changed
git diff --name-only origin/main...HEAD

# See which files main changed
git diff --name-only origin/main HEAD

# Likely conflict areas: files changed in both

Key insight: If you changed auth.ts and so did main, you WILL get conflicts in auth.ts.

Anticipating conflicts helps you understand how to resolve them.

Step 4: Choose Resolution Strategy

For detailed strategy comparison and decision matrix, see references/strategies.md.

Strategy A: Squash First (Recommended for 3+ commits)

When to use: Multiple feature commits with many conflicts expected

Why: Reduces conflicts to one resolution phase instead of per-commit

# Interactive rebase on current branch first
git rebase -i $(git merge-base HEAD origin/main)

# In editor, change all "pick" to "squash" (or 's') except first commit
# Save and exit - commits are squashed into one
# Edit commit message to describe the entire feature

# Now rebase the squashed commit
git rebase origin/main

# Resolve conflicts once, then git rebase --continue

Tradeoffs: Lose individual commit history, but simpler conflict resolution

Strategy B: Interactive Rebase with Conflict Awareness

When to use: 1-2 commits, clean history, or complex per-commit logic

git rebase -i origin/main

# In editor, you can:
# - Reorder commits to isolate conflict-prone ones
# - Drop commits that are already in main (git detects this)
# - Combine related commits before rebasing

# Save and exit - rebase proceeds, stopping at conflicts

Tradeoffs: More control, but more conflict-resolution iterations

Strategy C: Simple Linear Rebase (Fastest, Auto-Resolution)

When to use: Simple cases, no critical decisions, or in automated pipelines

# Rebase all commits at once
git rebase origin/main

# If no conflicts, done
# If conflicts, you resolve each one

Warning: Not recommended for complex scenarios. Use Strategies A or B instead.

Step 5: Apply Conflict Resolutions

When git rebase pauses with conflicts, use the analysis script:

# Analyze conflicts
bash scripts/analyze-conflicts.sh

# See which files conflict
git status

# For each conflicted file:
# - RECOMMENDED: Use merge tool for visual clarity
git mergetool --no-prompt

# - ALTERNATIVE: Manual edit in your editor
# Search for conflict markers: <<<<<<, ======, >>>>>>

Conflict Marker Anatomy

<<<<<<< HEAD
// Your current feature code
function authenticate(token) {
  validateToken(token);
  return true;
}
=======
// Main branch code (incoming)
function authenticate(token) {
  if (!token) throw new Error("No token");
  validateToken(token);
  setSession(token);
  return true;
}
>>>>>>> origin/main

Decision framework (before deleting markers):

1. Can you keep both? YES → Merge them intelligently function authenticate(token) {if (!token) throw new Error("No token"); // Keep main's validation validateToken(token); setSession(token); // Keep main's session setup return true; // Keep feature's return}
2. Conflicting logic? Understand WHY they differ, then decide

- Did main add critical security checks? → Keep main's version - Did your feature add essential functionality? → Keep feature's version - Are they trying to do different things? → Combine intentionally

1. Lost features? NEVER let a feature silently disappear

- If you added authentication logic, ensure it's in final version - If main improved database access, ensure that's preserved

For detailed resolution patterns, see references/resolution-patterns.md.

Key Resolution Principles

✅ DO:

• Keep both versions' important functionality when possible
• Use the merge tool for visual representation
• Add comments explaining merged conflicts: // Merged from both versions: main's validation + feature's session setup
• Test each file after resolution

❌ DON'T:

• Mindlessly pick one version without understanding both
• Delete conflict markers without understanding the conflict
• Keep duplicate code - merge intelligently
• Skip testing before continuing

Step 6: Validate Merged Code

After resolving conflicts, validate the merged code:

# Use the validation script
bash scripts/validate-merge.sh

# Manual checks:
# 1. Check syntax
npm run lint  # or eslint, pylint, etc.

# 2. Check types (if TypeScript)
npm run type-check  # or tsc --noEmit

# 3. Spot-check key files
git diff HEAD origin/main -- <conflicted-file>

# If validation fails:
# 1. Fix the issue in the file
# 2. git add <file>
# 3. git rebase --continue

Important: Validation catches mistakes BEFORE you commit them.

Step 7: Run Tests

This is your safety net:

# Run full test suite
npm test

# Or specific tests for changed areas
npm test -- --testPathPattern=auth  # If auth.ts was changed

# If tests fail:
# 1. Understand what broke
# 2. Fix in the files
# 3. git add <files>
# 4. git rebase --continue

Rule: Never force-push code that fails tests.

Step 8: Force Push Safely

Use --force-with-lease instead of --force. It protects against accidentally overwriting others' work:

# SAFE: Protects others' commits
git push origin $(git rev-parse --abbrev-ref HEAD) --force-with-lease

# UNSAFE: Can overwrite others' work
git push origin your-branch -f  # Don't do this

# If force-with-lease fails:
# Someone else pushed to your branch
# Coordinate with them before forcing

Common Scenarios & Strategies

Scenario 1: Many Small Conflicts Across 5+ Commits

Use: Squash-first strategy

git rebase -i $(git merge-base HEAD origin/main)
# Mark all but first commit as 's' (squash)
# Save - commits squash into one

git rebase origin/main
# Resolve conflicts once
git rebase --continue

# Only one conflict-resolution phase!

Why: Each commit might have conflicts. Squashing before rebasing means one pass.

Scenario 2: One Specific Commit Has Conflicts

Use: Target that commit with interactive rebase

git rebase -i origin/main

# In editor, move the problematic commit to the end
# Save - rebase proceeds, stopping at that commit

# When it stops, you know exactly which commit conflicts
git status  # See what changed in this commit

# Resolve, then continue
git rebase --continue

Why: Isolating the commit helps you understand what it's trying to do.

Scenario 3: Conflicts Keep Repeating (Same File, Different Commits)

Use: Git rerere (reuse recorded resolution)

# Enable rerere globally (one-time setup)
git config --global rerere.enabled true

# Now when you hit the same conflict in a second commit,
# Git automatically applies the first commit's resolution

git rebase origin/main

# Git remembers your first conflict resolution
# and replays it automatically for similar conflicts

Why: When rebasing and hitting the same file repeatedly, rerere saves manual work.

Scenario 4: Rebase Conflicts Are Too Complex

Emergency escape plan:

# Abort the rebase - return to original state
git rebase --abort

# Fall back to merge (safer for complex scenarios)
git merge origin/main

# Or try a different approach:
# - Squash your entire feature branch first
# - Cherry-pick main's critical changes selectively

Important: It's okay to abort and rethink. Better than a broken rebase.

Bundled Scripts

This skill includes helper scripts in scripts/:

• pre-rebase-backup.sh: Creates a safety backup before rebasing
• analyze-conflicts.sh: Analyzes current conflicts and provides detailed information
• validate-merge.sh: Validates that merge/rebase is clean and ready for testing

Run scripts with bash scripts/<script-name>.sh.

Reference Files

For detailed information on specific topics:

• Strategies: references/strategies.md - Complete strategy comparison and decision matrix
• Resolution Patterns: references/resolution-patterns.md - Common conflict resolution patterns and heuristics
• Troubleshooting: references/troubleshooting.md - Solutions to common rebase problems
• Automation: references/automation.md - Automated conflict resolution for CI/CD
• Scripts & Tools: references/scripts-tools.md - Additional git commands and tool usage

Checklist: Before You Commit to Rebasing

• Understand why you're rebasing (cleaner history, syncing with main, etc.)
• Backup your current branch: bash scripts/pre-rebase-backup.sh
• Run tests on current branch - they should pass
• Fetch latest: git fetch origin
• Understand what you'll be rebasing (5 commits? 50?)
• Understand main's recent changes (1 commit? Major refactor?)
• Choose your strategy (see references/strategies.md)
• Have your merge tool ready (if using GUI)
• Block 30 mins - don't rush conflict resolution
• Have tests ready to run after rebase

When NOT to Rebase

• Shared branches (use merge instead: git merge origin/main)
• Critical production code without comprehensive tests
• When multiple people are pushing to the same branch
• If you don't understand the conflicts you're seeing

Default to merge if uncertain. Merge is safer for collaborative work.

Use rebase when: Solo feature branch, clean history matters, no shared dependencies.

Summary: The Rebase Philosophy

Rebasing is a tool for creating clean, linear commit history. Used well, it makes debugging and code review easier. Used poorly, it loses work.

The key principle: Understand every conflict before resolving it. Don't automate away the thinking.

With this Skill, you can rebase with confidence, understanding each decision and protecting your code throughout the process.