Add cherry_pick.py script for ORT release

tools/python/cherry_pick.py

@@ -0,0 +1,318 @@
+# Copyright (c) Microsoft Corporation. All rights reserved.
+# Licensed under the MIT License.
+
+"""
+Cherry-Pick Helper Script
+-------------------------
+Description:
+    This script automates the process of cherry-picking commits for a release branch.
+    It fetches merged PRs with a specific label, sorts them by merge date, and generates:
+    1. A batch file (.cmd) with git cherry-pick commands.
+    2. A markdown file (.md) for the PR description.
+    It also checks for potential missing dependencies (conflicts) by verifying if files modified
+    by the cherry-picked commits have any other modifications in commits that are not in the
+    specified target branch and are not included in the cherry-pick list.
+
+Usage:
+    python cherry_pick.py --label "release:1.24.2" --output cherry_pick.cmd --branch "origin/rel-1.24.2"
+
+Requirements:
+    - Python 3.7+
+    - GitHub CLI (gh) logged in.
+    - Git available in PATH.
+"""
+
+import argparse
+import json
+import os
+import sys
+
+from cherry_pick_utils import (
+    check_preflight,
+    extract_pr_numbers,
+    get_pr_number_from_subject,
+    run_command,
+)
+
+
+def get_merged_prs(repo, label, limit=200):
+    """Fetch merged PRs with the specific label."""
+    print(f"Fetching merged PRs with label '{label}' from {repo}...")
+    cmd = [
+        "gh",
+        "pr",
+        "list",
+        "--repo",
+        repo,
+        "--label",
+        label,
+        "--state",
+        "merged",
+        "--json",
+        "number,title,mergeCommit,mergedAt",
+        "-L",
+        str(limit),
+    ]
+    output = run_command(cmd)
+    if not output:
+        return []
+
+    try:
+        return json.loads(output)
+    except json.JSONDecodeError as e:
+        print(f"Error parsing gh output: {e}", file=sys.stderr)
+        return []
+
+
+def get_changed_files(oid):
+    """Get list of files changed in a commit."""
+    output = run_command(["git", "diff-tree", "--no-commit-id", "--name-only", "-m", "-r", oid], silent=True)
+    if output:
+        return output.strip().splitlines()
+    return []
+
+
+def sanitize_title(title):
+    """Normalize PR titles for single-line text output."""
+    return title.replace("\n", " ").strip()
+
+
+def escape_markdown_table_cell(text):
+    """Escape markdown table delimiters in cell content."""
+    return sanitize_title(text).replace("|", "\\|")
+
+
+def get_existing_pr_numbers(branch, repo=None, log_limit=500):
+    """Get the set of PR numbers already present in the target branch."""
+    output = run_command(["git", "log", branch, "--oneline", "-n", str(log_limit)], silent=True)
+    if not output:
+        return set()
+    pr_numbers = set()
+
+    # Pre-fetch PR cache to avoid redundant gh calls
+    pr_cache = {}
+
+    # Process commit log
+    lines = output.strip().splitlines()
+    for line in lines:
+        parts = line.split(" ", 1)
+        if len(parts) < 2:
+            continue
+        subject = parts[1]
+
+        pr_num = get_pr_number_from_subject(subject)
+        if not pr_num:
+            continue
+
+        pr_num_int = int(pr_num)
+        pr_numbers.add(pr_num_int)
+
+        # Check if it's a cherry-pick / meta-PR
+        is_meta_pr = (
+            "cherry pick" in subject.lower() or "cherry-pick" in subject.lower() or "cherrypick" in subject.lower()
+        )
+
+        if is_meta_pr:
+            # Query gh to get more details (body/commits) to find squashed sub-PRs
+            if pr_num not in pr_cache:
+                gh_cmd = ["gh", "pr", "view", pr_num, "--json", "title,body,commits"]
+                if repo:
+                    gh_cmd.extend(["--repo", repo])
+                gh_out = run_command(gh_cmd, silent=True)
+                if gh_out:
+                    try:
+                        pr_cache[pr_num] = json.loads(gh_out)
+                    except json.JSONDecodeError:
+                        pr_cache[pr_num] = None
+                else:
+                    pr_cache[pr_num] = None
+
+            details = pr_cache.get(pr_num)
+            if details:
+                # Collect sub-PRs from title, body, and commits
+                extracted_nums = []
+                extracted_nums.extend(extract_pr_numbers(details.get("title", ""), strict=True))
+                extracted_nums.extend(extract_pr_numbers(details.get("body", ""), strict=True))
+
+                for commit in details.get("commits", []):
+                    extracted_nums.extend(extract_pr_numbers(commit.get("messageHeadline", ""), strict=True))
+
+                for num in set(extracted_nums):
+                    if num != pr_num_int:
+                        pr_numbers.add(num)
+
+    return pr_numbers
+
+
+def check_missing_dependencies(prs, branch):
+    """Check for potential missing dependencies (conflicts)."""
+    print("\nChecking for potential missing dependencies (conflicts)...")
+
+    # Collect OIDs being cherry-picked and all their ancestor commits
+    cherry_pick_oids = set()
+    for pr in prs:
+        if pr.get("mergeCommit"):
+            merge_oid = pr["mergeCommit"]["oid"]
+            cherry_pick_oids.add(merge_oid)
+            # Include ancestor commits of merge commits to avoid false-positive warnings
+            # for PRs that used a regular merge (not squash) strategy
+            ancestor_output = run_command(["git", "log", "--format=%H", merge_oid, "--not", branch], silent=True)
+            if ancestor_output:
+                for ancestor_oid in ancestor_output.strip().splitlines():
+                    cherry_pick_oids.add(ancestor_oid.strip())
+
+    conflicting_prs_count = 0
+    for pr in prs:
+        if not pr.get("mergeCommit"):
+            continue
+
+        oid = pr["mergeCommit"]["oid"]
+        number = pr["number"]
+
+        files = get_changed_files(oid)
+        if not files:
+            continue
+
+        # For each file, find commits that modified it between the target branch and the cherry-picked commit.
+        # Deduplicate warnings: group affected files by missing commit.
+        # missing_commits maps: missing_commit_oid -> {"title": ..., "files": [...]}
+        missing_commits = {}
+
+        for filepath in files:
+            # git log <cherry-pick-commit> --not <target-branch> -- <file>
+            output = run_command(["git", "log", oid, "--not", branch, "--format=%H %s", "--", filepath], silent=True)
+
+            if not output:
+                continue
+
+            for line in output.strip().splitlines():
+                parts = line.split(" ", 1)
+                c = parts[0]
+                title = parts[1] if len(parts) > 1 else ""
+
+                if c == oid:
+                    continue
+                if c not in cherry_pick_oids:
+                    entry = missing_commits.setdefault(c, {"title": title, "files": []})
+                    if not entry["title"]:
+                        entry["title"] = title
+                    entry["files"].append(filepath)
+
+        # Print deduplicated warnings
+        if missing_commits:
+            conflicting_prs_count += 1
+            for missing_oid, entry in missing_commits.items():
+                files_str = ", ".join(entry["files"])
+                print(
+                    f"WARNING: PR #{number} ({oid}) modifies files that were also changed by commit {missing_oid} ({entry['title']}), "
+                    f"which is not in the cherry-pick list. This may indicate missing related changes. Affected files: {files_str}"
+                )
+
+    if conflicting_prs_count == 0:
+        print("No potential missing dependencies found.")
+    else:
+        print(f"\nDone. Found potential missing dependencies for {conflicting_prs_count} PRs.")
+
+
+def main():
+    parser = argparse.ArgumentParser(description="Generate cherry-pick script from PRs with a specific label.")
+    parser.add_argument("--label", required=True, help="Label to filter PRs")
+    parser.add_argument(
+        "--output", required=True, help="Output script file path (.sh for bash, .cmd for Windows batch)"
+    )
+    parser.add_argument("--repo", default="microsoft/onnxruntime", help="Repository (default: microsoft/onnxruntime)")
+    parser.add_argument(
+        "--branch", default="HEAD", help="Target branch to compare against for dependency checks (default: HEAD)"
+    )
+    parser.add_argument("--limit", type=int, default=200, help="Maximum number of PRs to fetch (default: 200)")
+    parser.add_argument(
+        "--md-output",
+        help="Output markdown file path for the PR description (default: next to --output)",
+    )
+    args = parser.parse_args()
+
+    # Preflight Check
+    if not check_preflight():
+        return
+
+    # 1. Fetch Merged PRs
+    prs = get_merged_prs(args.repo, args.label, args.limit)
+
+    if not prs:
+        print(f"No PRs found with label '{args.label}'.")
+        return
+
+    # Sort by mergedAt (ISO 8601 strings sort correctly in chronological order)
+    prs.sort(key=lambda x: x["mergedAt"])
+
+    # 1.5. Check which PRs are already in the target branch
+    existing_prs = get_existing_pr_numbers(args.branch, repo=args.repo)
+    if existing_prs:
+        print(f"Found {len(existing_prs)} PRs already in branch '{args.branch}'.")
+
+    cherry_pick_prs = []
+    skipped_count = 0
+    for pr in prs:
+        number = pr["number"]
+        safe_title = sanitize_title(pr["title"])
+
+        if not pr.get("mergeCommit"):
+            print(f"Warning: PR #{number} has no merge commit OID. Skipping.", file=sys.stderr)
+            continue
+
+        if number in existing_prs:
+            print(f"Skipping PR #{number} (already in branch '{args.branch}'): {safe_title}")
+            skipped_count += 1
+            continue
+
+        cherry_pick_prs.append(pr)
+
+    # Determine output format based on file extension
+    is_shell = args.output.endswith(".sh")
+
+    # 2. Write Output Script
+    commit_count = len(cherry_pick_prs)
+    with open(args.output, "w", encoding="utf-8") as f:
+        if is_shell:
+            f.write("#!/bin/bash\n")
+            f.write(f"# Cherry-pick {args.label} commits\n")
+            f.write("# Sorted by merge time (oldest first)\n")
+            f.write("set -e\n\n")
+        else:
+            f.write("@echo off\n")
+            f.write(f"rem Cherry-pick {args.label} commits\n")
+            f.write("rem Sorted by merge time (oldest first)\n\n")
+
+        for pr in cherry_pick_prs:
+            number = pr["number"]
+            safe_title = sanitize_title(pr["title"])
+
+            oid = pr["mergeCommit"]["oid"]
+            comment = "#" if is_shell else "rem"
+            f.write(f"{comment} PR {number}: {safe_title}\n")
+            f.write(f"git cherry-pick {oid}\n\n")
+
+    print(f"Generated {args.output} with {commit_count} commits ({skipped_count} skipped, already in branch).")
+
+    # 3. Write PR Description Markdown (table format)
+    output_dir = os.path.dirname(args.output)
+    md_output = args.md_output or os.path.join(output_dir, "cherry_pick_pr_description.md")
+    with open(md_output, "w", encoding="utf-8") as f:
+        f.write("This cherry-picks the following commits for the release:\n\n")
+        f.write("| Commit ID | PR Number | Commit Title |\n")
+        f.write("|-----------|-----------|-------------|\n")
+        for pr in cherry_pick_prs:
+            number = pr["number"]
+            title = escape_markdown_table_cell(pr["title"])
+            oid = pr["mergeCommit"]["oid"]
+            short_oid = oid[:10]
+            f.write(f"| {short_oid} | #{number} | {title} |\n")
+
+    print(f"Generated {md_output} with {commit_count} commits.")
+
+    # 4. Dependency Check
+    check_missing_dependencies(cherry_pick_prs, args.branch)
+
+
+if __name__ == "__main__":
+    main()