From 4639c597d9b3aa06644feed014632248941ad5db Mon Sep 17 00:00:00 2001
From: Tianlei WU <tlwu@microsoft.com>
Date: Thu, 12 Feb 2026 15:14:02 -0800
Subject: [PATCH 01/11] Add cherry-pick script

---
 tools/python/cherry_pick.py | 162 ++++++++++++++++++++++++++++++++++++
 1 file changed, 162 insertions(+)
 create mode 100644 tools/python/cherry_pick.py

diff --git a/tools/python/cherry_pick.py b/tools/python/cherry_pick.py
new file mode 100644
index 0000000000000..fc2fa8cb55b9b
--- /dev/null
+++ b/tools/python/cherry_pick.py
@@ -0,0 +1,162 @@
+# Copyright (c) Microsoft Corporation. All rights reserved.
+# Licensed under the MIT License.
+#
+# Usage:
+# python create_cherry_pick.py --label "release:1.24.2" --output cherry_pick.cmd --branch "origin/rel-1.24.2"
+#
+# Arguments:
+#   --label: Label to filter PRs (required)
+#   --output: Output cmd file path (required)
+#   --repo: Repository (default: microsoft/onnxruntime)
+#   --branch: Target branch to compare against for dependency checks (default: HEAD)
+#
+# This script fetches merged PRs with the specified label from the onnxruntime repository,
+# sorts them by merge date, and generates:
+# 1. A batch file (specified by --output) containing git cherry-pick commands.
+# 2. A markdown file (cherry_pick_pr_description.md) summarizing the cherry-picked PRs for pull request description.
+#
+# It also checks for potential missing dependencies (conflicts) by verifying if files modified
+# by the cherry-picked commits have any other modifications in the target branch history
+# that are not included in the cherry-pick list.
+import argparse
+import subprocess
+import json
+import sys
+from collections import defaultdict
+
+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 cmd file path")
+    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)")
+    args = parser.parse_args()
+
+    # Fetch merged PRs with the specified label using gh CLI
+    print(f"Fetching merged PRs with label '{args.label}' from {args.repo}...")
+    cmd = [
+        "gh", "pr", "list",
+        "--repo", args.repo,
+        "--label", args.label,
+        "--state", "merged",
+        "--json", "number,title,mergeCommit,mergedAt",
+        "-L", "200"
+    ]
+
+    try:
+        result = subprocess.run(cmd, capture_output=True, text=True, check=True)
+        prs = json.loads(result.stdout)
+    except subprocess.CalledProcessError as e:
+        print(f"Error running gh command: {e}", file=sys.stderr)
+        print(e.stderr, file=sys.stderr)
+        sys.exit(1)
+    except json.JSONDecodeError as e:
+        print(f"Error parsing gh output: {e}", file=sys.stderr)
+        print(f"Output was: {result.stdout}", file=sys.stderr)
+        sys.exit(1)
+
+    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'])
+
+    # Write to output cmd file
+    commit_count = 0
+    with open(args.output, "w", encoding="utf-8") as f:
+        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 prs:
+            number = pr['number']
+            title = pr['title']
+            safe_title = title.replace('\n', ' ')
+
+            if not pr.get('mergeCommit'):
+                print(f"Warning: PR #{number} has no merge commit OID. Skipping.", file=sys.stderr)
+                continue
+
+            oid = pr['mergeCommit']['oid']
+            f.write(f"rem PR {number}: {safe_title}\n")
+            f.write(f"git cherry-pick {oid}\n\n")
+            commit_count += 1
+
+    print(f"Generated {args.output} with {commit_count} commits.")
+
+    # Write to markdown file. You can use it as the pull request description.
+    md_output = "cherry_pick_pr_description.md"
+    with open(md_output, "w", encoding="utf-8") as f:
+        f.write(f"This cherry-picks the following commits for the release:\n")
+        for pr in prs:
+            if not pr.get('mergeCommit'):
+                continue
+            number = pr['number']
+            f.write(f"- #{number}\n")
+
+    print(f"Generated {md_output} with {commit_count} commits.")
+
+    # Check for potential missing dependencies
+    print("\nChecking for potential missing dependencies (conflicts)...")
+
+    # Collect OIDs being cherry-picked
+    cherry_pick_oids = set()
+    for pr in prs:
+        if pr.get('mergeCommit'):
+            cherry_pick_oids.add(pr['mergeCommit']['oid'])
+
+    for pr in prs:
+        if not pr.get('mergeCommit'):
+            continue
+
+        oid = pr['mergeCommit']['oid']
+        number = pr['number']
+
+        # Get files changed by this commit
+        try:
+            res = subprocess.run(
+                ["git", "diff-tree", "--no-commit-id", "--name-only", "-r", oid],
+                capture_output=True, text=True, check=True
+            )
+            files = res.stdout.strip().splitlines()
+        except subprocess.CalledProcessError as e:
+            print(f"Error getting changed files for {oid}: {e}", file=sys.stderr)
+            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, [list of affected files])
+        missing_commits = defaultdict(lambda: ("", []))
+        for filepath in files:
+            try:
+                res = subprocess.run(
+                    ["git", "log", oid, "--not", args.branch, "--format=%H %s", "--", filepath],
+                    capture_output=True, text=True, check=True
+                )
+                for line in res.stdout.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:
+                        existing_title, existing_files = missing_commits[c]
+                        if not existing_title:
+                            existing_title = title
+                        existing_files.append(filepath)
+                        missing_commits[c] = (existing_title, existing_files)
+
+            except subprocess.CalledProcessError as e:
+                print(f"Error checking history for {filepath}: {e}", file=sys.stderr)
+                continue
+
+        # Print deduplicated warnings
+        for missing_oid, (title, affected_files) in missing_commits.items():
+            files_str = ", ".join(affected_files)
+            print(f"WARNING: PR #{number} ({oid}) depends on commit {missing_oid} ({title}) "
+                  f"which is not in the cherry-pick list. Affected files: {files_str}")
+
+if __name__ == "__main__":
+    main()

From dd0d487acaa8233efda159da4a4da9f8eacf09d8 Mon Sep 17 00:00:00 2001
From: Tianlei WU <tlwu@microsoft.com>
Date: Thu, 12 Feb 2026 15:21:06 -0800
Subject: [PATCH 02/11] lintrunner

---
 tools/python/cherry_pick.py | 71 +++++++++++++++++++++++--------------
 1 file changed, 44 insertions(+), 27 deletions(-)

diff --git a/tools/python/cherry_pick.py b/tools/python/cherry_pick.py
index fc2fa8cb55b9b..035d923725e4e 100644
--- a/tools/python/cherry_pick.py
+++ b/tools/python/cherry_pick.py
@@ -19,28 +19,38 @@
 # by the cherry-picked commits have any other modifications in the target branch history
 # that are not included in the cherry-pick list.
 import argparse
-import subprocess
 import json
+import subprocess
 import sys
 from collections import defaultdict
 
+
 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 cmd file path")
     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(
+        "--branch", default="HEAD", help="Target branch to compare against for dependency checks (default: HEAD)"
+    )
     args = parser.parse_args()
 
     # Fetch merged PRs with the specified label using gh CLI
     print(f"Fetching merged PRs with label '{args.label}' from {args.repo}...")
     cmd = [
-        "gh", "pr", "list",
-        "--repo", args.repo,
-        "--label", args.label,
-        "--state", "merged",
-        "--json", "number,title,mergeCommit,mergedAt",
-        "-L", "200"
+        "gh",
+        "pr",
+        "list",
+        "--repo",
+        args.repo,
+        "--label",
+        args.label,
+        "--state",
+        "merged",
+        "--json",
+        "number,title,mergeCommit,mergedAt",
+        "-L",
+        "200",
     ]
 
     try:
@@ -60,7 +70,7 @@ def main():
         return
 
     # Sort by mergedAt (ISO 8601 strings sort correctly in chronological order)
-    prs.sort(key=lambda x: x['mergedAt'])
+    prs.sort(key=lambda x: x["mergedAt"])
 
     # Write to output cmd file
     commit_count = 0
@@ -70,15 +80,15 @@ def main():
         f.write("rem Sorted by merge time (oldest first)\n\n")
 
         for pr in prs:
-            number = pr['number']
-            title = pr['title']
-            safe_title = title.replace('\n', ' ')
+            number = pr["number"]
+            title = pr["title"]
+            safe_title = title.replace("\n", " ")
 
-            if not pr.get('mergeCommit'):
+            if not pr.get("mergeCommit"):
                 print(f"Warning: PR #{number} has no merge commit OID. Skipping.", file=sys.stderr)
                 continue
 
-            oid = pr['mergeCommit']['oid']
+            oid = pr["mergeCommit"]["oid"]
             f.write(f"rem PR {number}: {safe_title}\n")
             f.write(f"git cherry-pick {oid}\n\n")
             commit_count += 1
@@ -88,11 +98,11 @@ def main():
     # Write to markdown file. You can use it as the pull request description.
     md_output = "cherry_pick_pr_description.md"
     with open(md_output, "w", encoding="utf-8") as f:
-        f.write(f"This cherry-picks the following commits for the release:\n")
+        f.write("This cherry-picks the following commits for the release:\n")
         for pr in prs:
-            if not pr.get('mergeCommit'):
+            if not pr.get("mergeCommit"):
                 continue
-            number = pr['number']
+            number = pr["number"]
             f.write(f"- #{number}\n")
 
     print(f"Generated {md_output} with {commit_count} commits.")
@@ -103,21 +113,23 @@ def main():
     # Collect OIDs being cherry-picked
     cherry_pick_oids = set()
     for pr in prs:
-        if pr.get('mergeCommit'):
-            cherry_pick_oids.add(pr['mergeCommit']['oid'])
+        if pr.get("mergeCommit"):
+            cherry_pick_oids.add(pr["mergeCommit"]["oid"])
 
     for pr in prs:
-        if not pr.get('mergeCommit'):
+        if not pr.get("mergeCommit"):
             continue
 
-        oid = pr['mergeCommit']['oid']
-        number = pr['number']
+        oid = pr["mergeCommit"]["oid"]
+        number = pr["number"]
 
         # Get files changed by this commit
         try:
             res = subprocess.run(
                 ["git", "diff-tree", "--no-commit-id", "--name-only", "-r", oid],
-                capture_output=True, text=True, check=True
+                capture_output=True,
+                text=True,
+                check=True,
             )
             files = res.stdout.strip().splitlines()
         except subprocess.CalledProcessError as e:
@@ -132,10 +144,12 @@ def main():
             try:
                 res = subprocess.run(
                     ["git", "log", oid, "--not", args.branch, "--format=%H %s", "--", filepath],
-                    capture_output=True, text=True, check=True
+                    capture_output=True,
+                    text=True,
+                    check=True,
                 )
                 for line in res.stdout.strip().splitlines():
-                    parts = line.split(' ', 1)
+                    parts = line.split(" ", 1)
                     c = parts[0]
                     title = parts[1] if len(parts) > 1 else ""
 
@@ -155,8 +169,11 @@ def main():
         # Print deduplicated warnings
         for missing_oid, (title, affected_files) in missing_commits.items():
             files_str = ", ".join(affected_files)
-            print(f"WARNING: PR #{number} ({oid}) depends on commit {missing_oid} ({title}) "
-                  f"which is not in the cherry-pick list. Affected files: {files_str}")
+            print(
+                f"WARNING: PR #{number} ({oid}) depends on commit {missing_oid} ({title}) "
+                f"which is not in the cherry-pick list. Affected files: {files_str}"
+            )
+
 
 if __name__ == "__main__":
     main()

From 6a44998644beac778c5138bf756fadd3c9c5fa2e Mon Sep 17 00:00:00 2001
From: Tianlei WU <tlwu@microsoft.com>
Date: Thu, 12 Feb 2026 15:46:21 -0800
Subject: [PATCH 03/11] address copilot feedback

---
 tools/python/cherry_pick.py | 287 ++++++++++++++++++++++--------------
 1 file changed, 177 insertions(+), 110 deletions(-)

diff --git a/tools/python/cherry_pick.py b/tools/python/cherry_pick.py
index 035d923725e4e..2d44f02da665f 100644
--- a/tools/python/cherry_pick.py
+++ b/tools/python/cherry_pick.py
@@ -1,23 +1,27 @@
 # Copyright (c) Microsoft Corporation. All rights reserved.
 # Licensed under the MIT License.
-#
-# Usage:
-# python create_cherry_pick.py --label "release:1.24.2" --output cherry_pick.cmd --branch "origin/rel-1.24.2"
-#
-# Arguments:
-#   --label: Label to filter PRs (required)
-#   --output: Output cmd file path (required)
-#   --repo: Repository (default: microsoft/onnxruntime)
-#   --branch: Target branch to compare against for dependency checks (default: HEAD)
-#
-# This script fetches merged PRs with the specified label from the onnxruntime repository,
-# sorts them by merge date, and generates:
-# 1. A batch file (specified by --output) containing git cherry-pick commands.
-# 2. A markdown file (cherry_pick_pr_description.md) summarizing the cherry-picked PRs for pull request description.
-#
-# It also checks for potential missing dependencies (conflicts) by verifying if files modified
-# by the cherry-picked commits have any other modifications in the target branch history
-# that are not included in the cherry-pick list.
+
+"""
+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 subprocess
@@ -25,45 +29,170 @@
 from collections import defaultdict
 
 
-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 cmd file path")
-    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)"
-    )
-    args = parser.parse_args()
-
-    # Fetch merged PRs with the specified label using gh CLI
-    print(f"Fetching merged PRs with label '{args.label}' from {args.repo}...")
+def run_command(command_list, cwd=None, silent=False):
+    """Run a command using a list of arguments for security (no shell=True)."""
+    try:
+        result = subprocess.run(command_list, check=False, capture_output=True, text=True, cwd=cwd, encoding="utf-8")
+        if result.returncode != 0:
+            if not silent:
+                log_str = " ".join(command_list)
+                print(f"Error running command: {log_str}", file=sys.stderr)
+                if result.stderr:
+                    print(f"Stderr: {result.stderr.strip()}", file=sys.stderr)
+            return None
+        return result.stdout
+    except FileNotFoundError:
+        if not silent:
+            cmd = command_list[0]
+            print(f"Error: '{cmd}' command not found.", file=sys.stderr)
+            if cmd == "gh":
+                print(
+                    "Please install GitHub CLI (https://cli.github.com/) and ensure 'gh' is available on your PATH.",
+                    file=sys.stderr,
+                )
+        return None
+    except Exception as e:
+        if not silent:
+            print(f"Exception running command {' '.join(command_list)}: {e}", file=sys.stderr)
+        return None
+
+
+def check_preflight():
+    """Verify gh CLI and git repository early."""
+    # Check git
+    git_check = run_command(["git", "rev-parse", "--is-inside-work-tree"], silent=True)
+    if not git_check:
+        print("Error: This script must be run inside a git repository.", file=sys.stderr)
+        return False
+
+    # Check gh
+    gh_check = run_command(["gh", "--version"], silent=True)
+    if not gh_check:
+        print("Error: GitHub CLI (gh) not found or not in PATH.", file=sys.stderr)
+        print(
+            "Please install GitHub CLI (https://cli.github.com/) and ensure 'gh' is available on your PATH.",
+            file=sys.stderr,
+        )
+        return False
+
+    gh_auth = run_command(["gh", "auth", "status"], silent=True)
+    if not gh_auth:
+        print("Error: GitHub CLI not authenticated. Please run 'gh auth login'.", file=sys.stderr)
+        return False
+
+    return True
+
+
+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",
-        args.repo,
+        repo,
         "--label",
-        args.label,
+        label,
         "--state",
         "merged",
         "--json",
         "number,title,mergeCommit,mergedAt",
         "-L",
-        "200",
+        str(limit),
     ]
+    output = run_command(cmd)
+    if not output:
+        return []
 
     try:
-        result = subprocess.run(cmd, capture_output=True, text=True, check=True)
-        prs = json.loads(result.stdout)
-    except subprocess.CalledProcessError as e:
-        print(f"Error running gh command: {e}", file=sys.stderr)
-        print(e.stderr, file=sys.stderr)
-        sys.exit(1)
+        return json.loads(output)
     except json.JSONDecodeError as e:
         print(f"Error parsing gh output: {e}", file=sys.stderr)
-        print(f"Output was: {result.stdout}", file=sys.stderr)
-        sys.exit(1)
+        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", "-r", oid], silent=True)
+    if output:
+        return output.strip().splitlines()
+    return []
+
+
+def check_missing_dependencies(prs, branch):
+    """Check for potential missing dependencies (conflicts)."""
+    print("\nChecking for potential missing dependencies (conflicts)...")
+
+    # Collect OIDs being cherry-picked
+    cherry_pick_oids = set()
+    for pr in prs:
+        if pr.get("mergeCommit"):
+            cherry_pick_oids.add(pr["mergeCommit"]["oid"])
+
+    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, [list of affected files])
+        missing_commits = defaultdict(lambda: ("", []))
+
+        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:
+                    existing_title, existing_files = missing_commits[c]
+                    if not existing_title:
+                        existing_title = title
+                    existing_files.append(filepath)
+                    missing_commits[c] = (existing_title, existing_files)
+
+        # Print deduplicated warnings
+        for missing_oid, (title, affected_files) in missing_commits.items():
+            files_str = ", ".join(affected_files)
+            print(
+                f"WARNING: PR #{number} ({oid}) modifies files that were also changed by commit {missing_oid} ({title}), "
+                f"which is not in the cherry-pick list. This may indicate missing related changes. Affected files: {files_str}"
+            )
+
+
+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 cmd file path")
+    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="Wait limitation for PR fetching (default: 200)")
+    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}'.")
@@ -72,7 +201,7 @@ def main():
     # Sort by mergedAt (ISO 8601 strings sort correctly in chronological order)
     prs.sort(key=lambda x: x["mergedAt"])
 
-    # Write to output cmd file
+    # 2. Write Output Script
     commit_count = 0
     with open(args.output, "w", encoding="utf-8") as f:
         f.write("@echo off\n")
@@ -95,7 +224,7 @@ def main():
 
     print(f"Generated {args.output} with {commit_count} commits.")
 
-    # Write to markdown file. You can use it as the pull request description.
+    # 3. Write PR Description Markdown
     md_output = "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")
@@ -103,76 +232,14 @@ def main():
             if not pr.get("mergeCommit"):
                 continue
             number = pr["number"]
-            f.write(f"- #{number}\n")
+            title = pr["title"]
+            # Markdown link format: - #123 Title
+            f.write(f"- #{number} {title}\n")
 
     print(f"Generated {md_output} with {commit_count} commits.")
 
-    # Check for potential missing dependencies
-    print("\nChecking for potential missing dependencies (conflicts)...")
-
-    # Collect OIDs being cherry-picked
-    cherry_pick_oids = set()
-    for pr in prs:
-        if pr.get("mergeCommit"):
-            cherry_pick_oids.add(pr["mergeCommit"]["oid"])
-
-    for pr in prs:
-        if not pr.get("mergeCommit"):
-            continue
-
-        oid = pr["mergeCommit"]["oid"]
-        number = pr["number"]
-
-        # Get files changed by this commit
-        try:
-            res = subprocess.run(
-                ["git", "diff-tree", "--no-commit-id", "--name-only", "-r", oid],
-                capture_output=True,
-                text=True,
-                check=True,
-            )
-            files = res.stdout.strip().splitlines()
-        except subprocess.CalledProcessError as e:
-            print(f"Error getting changed files for {oid}: {e}", file=sys.stderr)
-            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, [list of affected files])
-        missing_commits = defaultdict(lambda: ("", []))
-        for filepath in files:
-            try:
-                res = subprocess.run(
-                    ["git", "log", oid, "--not", args.branch, "--format=%H %s", "--", filepath],
-                    capture_output=True,
-                    text=True,
-                    check=True,
-                )
-                for line in res.stdout.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:
-                        existing_title, existing_files = missing_commits[c]
-                        if not existing_title:
-                            existing_title = title
-                        existing_files.append(filepath)
-                        missing_commits[c] = (existing_title, existing_files)
-
-            except subprocess.CalledProcessError as e:
-                print(f"Error checking history for {filepath}: {e}", file=sys.stderr)
-                continue
-
-        # Print deduplicated warnings
-        for missing_oid, (title, affected_files) in missing_commits.items():
-            files_str = ", ".join(affected_files)
-            print(
-                f"WARNING: PR #{number} ({oid}) depends on commit {missing_oid} ({title}) "
-                f"which is not in the cherry-pick list. Affected files: {files_str}"
-            )
+    # 4. Dependency Check
+    check_missing_dependencies(prs, args.branch)
 
 
 if __name__ == "__main__":

From 1e6bbf915e5b27c09a259b463f01365f1a88d3b7 Mon Sep 17 00:00:00 2001
From: Tianlei Wu <tlwu@microsoft.com>
Date: Thu, 19 Feb 2026 09:10:29 +0000
Subject: [PATCH 04/11] update compile contributors patterns

---
 tools/python/compile_contributors.py | 3 ++-
 1 file changed, 2 insertions(+), 1 deletion(-)

diff --git a/tools/python/compile_contributors.py b/tools/python/compile_contributors.py
index 494b0f91c5381..79e5166cfb4b7 100644
--- a/tools/python/compile_contributors.py
+++ b/tools/python/compile_contributors.py
@@ -219,6 +219,7 @@ def extract_pr_numbers(text, strict=False):
         # And it avoids matching truncated headlines like (#25... as PR #25
         patterns = [
             r"\(#(\d+)\)",  # (#123)
+            r"(?:^|\s|-)#(\d+)(?:\s|$)",  # #123 at start, or preceded by space/dash, and followed by space or end
             r"microsoft/onnxruntime/pull/(\d+)",
         ]
         results = []
@@ -280,7 +281,7 @@ def get_prs_from_log(log_output, prs_base=None, log_file=None, scan_depth=100):
                 # Reuse commits already fetched in get_pr_details to avoid an extra gh CLI call
                 for commit in details.get("commits", []):
                     all_extracted_nums.extend(extract_pr_numbers(commit.get("messageHeadline", ""), strict=True))
-                    all_extracted_nums.extend(extract_pr_numbers(commit.get("messageBody", ""), strict=True))
+                    # DO NOT scan messageBody for expansion to avoid historical context PRs
 
                 # Filter and Normalize
                 current_pr_int = int(pr_num_str)

From 1c40b36abf04f1f87e62ce1e7b26b5499be25c0d Mon Sep 17 00:00:00 2001
From: Tianlei Wu <tlwu@microsoft.com>
Date: Fri, 27 Feb 2026 01:38:39 +0000
Subject: [PATCH 05/11] output sh and markdown table

---
 tools/python/cherry_pick.py          | 76 ++++++++++++++++++++++++----
 tools/python/compile_contributors.py |  7 ++-
 2 files changed, 70 insertions(+), 13 deletions(-)

diff --git a/tools/python/cherry_pick.py b/tools/python/cherry_pick.py
index 2d44f02da665f..12beb44fb05bd 100644
--- a/tools/python/cherry_pick.py
+++ b/tools/python/cherry_pick.py
@@ -24,6 +24,7 @@
 
 import argparse
 import json
+import re
 import subprocess
 import sys
 from collections import defaultdict
@@ -120,6 +121,31 @@ def get_changed_files(oid):
     return []
 
 
+def get_pr_number_from_subject(subject):
+    """Extract PR number from a commit subject like 'Some title (#12345)'."""
+    match = re.search(r"\(#(\d+)\)$", subject.strip())
+    if match:
+        return match.group(1)
+    return None
+
+
+def get_existing_pr_numbers(branch):
+    """Get the set of PR numbers already present in the target branch."""
+    output = run_command(["git", "log", branch, "--oneline", "-n", "500"], silent=True)
+    if not output:
+        return set()
+    pr_numbers = set()
+    for line in output.strip().splitlines():
+        parts = line.split(" ", 1)
+        if len(parts) < 2:
+            continue
+        subject = parts[1]
+        pr_num = get_pr_number_from_subject(subject)
+        if pr_num:
+            pr_numbers.add(int(pr_num))
+    return pr_numbers
+
+
 def check_missing_dependencies(prs, branch):
     """Check for potential missing dependencies (conflicts)."""
     print("\nChecking for potential missing dependencies (conflicts)...")
@@ -179,7 +205,9 @@ def check_missing_dependencies(prs, branch):
 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 cmd file path")
+    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)"
@@ -201,12 +229,27 @@ def main():
     # 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)
+    if existing_prs:
+        print(f"Found {len(existing_prs)} PRs already in branch '{args.branch}'.")
+
+    # Determine output format based on file extension
+    is_shell = args.output.endswith(".sh")
+
     # 2. Write Output Script
     commit_count = 0
+    skipped_count = 0
     with open(args.output, "w", encoding="utf-8") as f:
-        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")
+        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 prs:
             number = pr["number"]
@@ -217,24 +260,35 @@ def main():
                 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
+
             oid = pr["mergeCommit"]["oid"]
-            f.write(f"rem PR {number}: {safe_title}\n")
+            comment = "#" if is_shell else "rem"
+            f.write(f"{comment} PR {number}: {safe_title}\n")
             f.write(f"git cherry-pick {oid}\n\n")
             commit_count += 1
 
-    print(f"Generated {args.output} with {commit_count} commits.")
+    print(f"Generated {args.output} with {commit_count} commits ({skipped_count} skipped, already in branch).")
 
-    # 3. Write PR Description Markdown
+    # 3. Write PR Description Markdown (table format)
     md_output = "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")
+        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 prs:
             if not pr.get("mergeCommit"):
                 continue
             number = pr["number"]
-            title = pr["title"]
-            # Markdown link format: - #123 Title
-            f.write(f"- #{number} {title}\n")
+            if number in existing_prs:
+                continue
+            title = pr["title"].replace("\n", " ")
+            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.")
 
diff --git a/tools/python/compile_contributors.py b/tools/python/compile_contributors.py
index 79e5166cfb4b7..5324c90aad94c 100644
--- a/tools/python/compile_contributors.py
+++ b/tools/python/compile_contributors.py
@@ -214,14 +214,17 @@ def extract_pr_numbers(text, strict=False):
         return []
 
     if strict:
-        # Strict mode: Only look for (#123) with closing paren or full onnxruntime URLs
+        # Strict mode: Only look for (#123) with closing paren, full onnxruntime URLs,
+        # or PR numbers in markdown table cells (| #123 |), or standalone #123 with clear boundaries.
         # This avoids noise from version numbers or external repo PRs
         # And it avoids matching truncated headlines like (#25... as PR #25
         patterns = [
             r"\(#(\d+)\)",  # (#123)
-            r"(?:^|\s|-)#(\d+)(?:\s|$)",  # #123 at start, or preceded by space/dash, and followed by space or end
             r"microsoft/onnxruntime/pull/(\d+)",
+            r"(?:^|\s|-)#(\d+)(?:\s|$)",  # #123 at start, or preceded by space/dash, and followed by space or end
+            r"\|\s*#(\d+)\s*\|",  # | #123 | (markdown table cell)
         ]
+
         results = []
         for p in patterns:
             results.extend(re.findall(p, text))

From 90d5604e007b59c1ed93772134c676a73101ced0 Mon Sep 17 00:00:00 2001
From: Tianlei Wu <tlwu@microsoft.com>
Date: Sat, 28 Feb 2026 20:55:14 +0000
Subject: [PATCH 06/11] refactoring

---
 tools/python/cherry_pick.py          | 117 ++++++++----------
 tools/python/cherry_pick_utils.py    |  92 ++++++++++++++
 tools/python/compile_contributors.py | 171 ++++++++-------------------
 3 files changed, 194 insertions(+), 186 deletions(-)
 create mode 100644 tools/python/cherry_pick_utils.py

diff --git a/tools/python/cherry_pick.py b/tools/python/cherry_pick.py
index 12beb44fb05bd..85bc7cba03617 100644
--- a/tools/python/cherry_pick.py
+++ b/tools/python/cherry_pick.py
@@ -25,63 +25,14 @@
 import argparse
 import json
 import re
-import subprocess
 import sys
 from collections import defaultdict
-
-
-def run_command(command_list, cwd=None, silent=False):
-    """Run a command using a list of arguments for security (no shell=True)."""
-    try:
-        result = subprocess.run(command_list, check=False, capture_output=True, text=True, cwd=cwd, encoding="utf-8")
-        if result.returncode != 0:
-            if not silent:
-                log_str = " ".join(command_list)
-                print(f"Error running command: {log_str}", file=sys.stderr)
-                if result.stderr:
-                    print(f"Stderr: {result.stderr.strip()}", file=sys.stderr)
-            return None
-        return result.stdout
-    except FileNotFoundError:
-        if not silent:
-            cmd = command_list[0]
-            print(f"Error: '{cmd}' command not found.", file=sys.stderr)
-            if cmd == "gh":
-                print(
-                    "Please install GitHub CLI (https://cli.github.com/) and ensure 'gh' is available on your PATH.",
-                    file=sys.stderr,
-                )
-        return None
-    except Exception as e:
-        if not silent:
-            print(f"Exception running command {' '.join(command_list)}: {e}", file=sys.stderr)
-        return None
-
-
-def check_preflight():
-    """Verify gh CLI and git repository early."""
-    # Check git
-    git_check = run_command(["git", "rev-parse", "--is-inside-work-tree"], silent=True)
-    if not git_check:
-        print("Error: This script must be run inside a git repository.", file=sys.stderr)
-        return False
-
-    # Check gh
-    gh_check = run_command(["gh", "--version"], silent=True)
-    if not gh_check:
-        print("Error: GitHub CLI (gh) not found or not in PATH.", file=sys.stderr)
-        print(
-            "Please install GitHub CLI (https://cli.github.com/) and ensure 'gh' is available on your PATH.",
-            file=sys.stderr,
-        )
-        return False
-
-    gh_auth = run_command(["gh", "auth", "status"], silent=True)
-    if not gh_auth:
-        print("Error: GitHub CLI not authenticated. Please run 'gh auth login'.", file=sys.stderr)
-        return False
-
-    return True
+from cherry_pick_utils import (
+    run_command,
+    check_preflight,
+    get_pr_number_from_subject,
+    extract_pr_numbers,
+)
 
 
 def get_merged_prs(repo, label, limit=200):
@@ -121,28 +72,62 @@ def get_changed_files(oid):
     return []
 
 
-def get_pr_number_from_subject(subject):
-    """Extract PR number from a commit subject like 'Some title (#12345)'."""
-    match = re.search(r"\(#(\d+)\)$", subject.strip())
-    if match:
-        return match.group(1)
-    return None
-
-
 def get_existing_pr_numbers(branch):
     """Get the set of PR numbers already present in the target branch."""
     output = run_command(["git", "log", branch, "--oneline", "-n", "500"], silent=True)
     if not output:
         return set()
     pr_numbers = set()
-    for line in output.strip().splitlines():
+
+    # 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 pr_num:
-            pr_numbers.add(int(pr_num))
+        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_out = run_command(["gh", "pr", "view", pr_num, "--json", "title,body,commits"], 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", "")))
+                extracted_nums.extend(extract_pr_numbers(details.get("body", "")))
+
+                for commit in details.get("commits", []):
+                    extracted_nums.extend(extract_pr_numbers(commit.get("messageHeadline", "")))
+
+                for num in set(extracted_nums):
+                    if num != pr_num_int:
+                        pr_numbers.add(num)
+
     return pr_numbers
 
 
diff --git a/tools/python/cherry_pick_utils.py b/tools/python/cherry_pick_utils.py
new file mode 100644
index 0000000000000..34e3f8f0f87eb
--- /dev/null
+++ b/tools/python/cherry_pick_utils.py
@@ -0,0 +1,92 @@
+import subprocess
+import sys
+import re
+import json
+
+def run_command(command_list, cwd=None, silent=False):
+    """Run a command using a list of arguments for security (no shell=True)."""
+    try:
+        result = subprocess.run(command_list, check=False, capture_output=True, text=True, cwd=cwd, encoding="utf-8")
+        if result.returncode != 0:
+            if not silent:
+                log_str = " ".join(command_list)
+                print(f"Error running command: {log_str}", file=sys.stderr)
+                if result.stderr:
+                    print(f"Stderr: {result.stderr.strip()}", file=sys.stderr)
+            return None
+        return result.stdout
+    except FileNotFoundError:
+        if not silent:
+            cmd = command_list[0]
+            print(f"Error: '{cmd}' command not found.", file=sys.stderr)
+            if cmd == "gh":
+                print(
+                    "Please install GitHub CLI (https://cli.github.com/) and ensure 'gh' is available on your PATH.",
+                    file=sys.stderr,
+                )
+        return None
+    except Exception as e:
+        if not silent:
+            print(f"Exception running command {' '.join(command_list)}: {e}", file=sys.stderr)
+        return None
+
+
+def check_preflight():
+    """Verify gh CLI and git repository early."""
+    # Check git
+    git_check = run_command(["git", "rev-parse", "--is-inside-work-tree"], silent=True)
+    if not git_check:
+        print("Error: This script must be run inside a git repository.", file=sys.stderr)
+        return False
+
+    # Check gh
+    gh_check = run_command(["gh", "--version"], silent=True)
+    if not gh_check:
+        print("Error: GitHub CLI (gh) not found or not in PATH.", file=sys.stderr)
+        print(
+            "Please install GitHub CLI (https://cli.github.com/) and ensure 'gh' is available on your PATH.",
+            file=sys.stderr,
+        )
+        return False
+
+    gh_auth = run_command(["gh", "auth", "status"], silent=True)
+    if not gh_auth:
+        print("Error: GitHub CLI not authenticated. Please run 'gh auth login'.", file=sys.stderr)
+        return False
+
+    return True
+
+
+def get_pr_number_from_subject(subject):
+    """Extract PR number from a commit subject like 'Some title (#12345)'."""
+    match = re.search(r"\(#(\d+)\)$", subject.strip())
+    if match:
+        return match.group(1)
+    return None
+
+
+def extract_pr_numbers(text, strict=False):
+    if not text:
+        return []
+
+    if strict:
+        # Strict mode: Only look for (#123) with closing paren, full onnxruntime URLs,
+        # or PR numbers in markdown table cells (| #123 |), or standalone #123 with clear boundaries.
+        # This avoids noise from version numbers or external repo PRs
+        # And it avoids matching truncated headlines like (#25... as PR #25
+        patterns = [
+            r"\(#(\d+)\)",  # (#123)
+            r"microsoft/onnxruntime/pull/(\d+)",
+            r"(?:^|\s|-)#(\d+)(?:\s|$)",  # #123 at start, or preceded by space/dash, and followed by space or end
+            r"\|\s*#(\d+)\s*\|",  # | #123 | (markdown table cell)
+        ]
+
+        results = []
+        for p in patterns:
+            results.extend(re.findall(p, text))
+        return [int(x) for x in set(results)]
+
+    # Matches patterns like #123 or https://github.com/microsoft/onnxruntime/pull/123
+    # Also handles ( #123) or similar in titles
+    prs = re.findall(r"(?:#|/pull/)(\d+)", text)
+    return [int(x) for x in set(prs)]
diff --git a/tools/python/compile_contributors.py b/tools/python/compile_contributors.py
index 5324c90aad94c..892f0f1648502 100644
--- a/tools/python/compile_contributors.py
+++ b/tools/python/compile_contributors.py
@@ -30,8 +30,12 @@
 import json
 import os
 import re
-import subprocess
-
+from cherry_pick_utils import (
+    run_command,
+    check_preflight,
+    get_pr_number_from_subject as get_pr_number,
+    extract_pr_numbers,
+)
 
 def log_event(message, log_file=None):
     """Log a message to the console and an optional log file."""
@@ -40,100 +44,9 @@ def log_event(message, log_file=None):
     print(message)  # Clean print for console UI
     if log_file:
         log_file.write(full_message + "\n")
-
-
-def run_command(command_list, cwd=".", silent=False):
-    """Run a command using a list of arguments for security (no shell=True)."""
-    result = subprocess.run(command_list, check=False, capture_output=True, text=True, cwd=cwd, encoding="utf-8")
-    if result.returncode != 0:
-        if not silent:
-            log_str = " ".join(command_list)
-            print(f"Error running command: {log_str}")
-            if result.stderr:
-                print(f"Stderr: {result.stderr.strip()}")
-        return None
-    return result.stdout
-
-
-def check_preflight():
-    """Verify gh CLI and git repository early."""
-    # Check git
-    git_check = run_command(["git", "rev-parse", "--is-inside-work-tree"], silent=True)
-    if not git_check:
-        print("Error: This script must be run inside a git repository.")
-        return False
-
-    # Check gh
-    gh_check = run_command(["gh", "--version"], silent=True)
-    if not gh_check:
-        print("Error: GitHub CLI (gh) not found or not in PATH.")
-        return False
-
-    gh_auth = run_command(["gh", "auth", "status"], silent=True)
-    if not gh_auth:
-        print("Error: GitHub CLI not authenticated. Please run 'gh auth login'.")
-        return False
-
-    return True
-
-
-# Constants
-PR_CACHE = {}  # Cache for PR details to speed up multiple rounds referencing same PRs
 NAME_TO_LOGIN = {}  # Map full names to GitHub logins for consolidation
 VERIFIED_LOGINS = set()  # Track IDs known to be valid GitHub logins (vs free-form names)
-
-# Bots to exclude from contributor lists
-BOT_NAMES = {
-    "Copilot",
-    "dependabot[bot]",
-    "app/dependabot",
-    "github-actions[bot]",
-    "app/copilot-swe-agent",
-    "CI Bot",
-    "github-advanced-security[bot]",
-    "GitHub Actions",
-    "dependabot",
-    "github-actions",
-    "Gemini",
-    "CI",
-}
-
-
-def is_bot(name):
-    if not name:
-        return True
-    name_clean = name.strip().lstrip("@")
-    # Known bots and patterns
-    if name_clean in BOT_NAMES:
-        return True
-    if "[bot]" in name_clean.lower():
-        return True
-    if name_clean.lower().startswith("app/"):
-        return True
-    return False
-
-
-def is_invalid(name):
-    if not name:
-        return True
-    # If it's a bot, it's considered a valid identity for the CSV
-    if is_bot(name):
-        return False
-
-    name_clean = name.strip().lstrip("@")
-    # Paths, brackets, and code extensions
-    if "/" in name_clean or "\\" in name_clean or "[" in name_clean or "]" in name_clean:
-        return True
-    if any(name_clean.lower().endswith(ext) for ext in [".cmake", ".py", ".h", ".cc", ".cpp", ".txt", ".md"]):
-        return True
-    return False
-
-
-def get_pr_number(subject):
-    match = re.search(r"\(#(\d+)\)$", subject.strip())
-    if match:
-        return match.group(1)
-    return None
+PR_CACHE = {}  # Cache for PR details to speed up multiple rounds referencing same PRs
 
 
 def get_pr_details(pr_number):
@@ -150,7 +63,6 @@ def get_pr_details(pr_number):
     PR_CACHE[pr_number] = None
     return None
 
-
 def extract_authors_from_pr(details):
     authors = set()
     if not details:
@@ -208,32 +120,51 @@ def extract_authors_from_commit(commit_id):
 
     return authors
 
+# Bots to exclude from contributor lists
+BOT_NAMES = {
+    "Copilot",
+    "dependabot[bot]",
+    "app/dependabot",
+    "github-actions[bot]",
+    "app/copilot-swe-agent",
+    "CI Bot",
+    "github-advanced-security[bot]",
+    "GitHub Actions",
+    "dependabot",
+    "github-actions",
+    "Gemini",
+    "CI",
+}
 
-def extract_pr_numbers(text, strict=False):
-    if not text:
-        return []
-
-    if strict:
-        # Strict mode: Only look for (#123) with closing paren, full onnxruntime URLs,
-        # or PR numbers in markdown table cells (| #123 |), or standalone #123 with clear boundaries.
-        # This avoids noise from version numbers or external repo PRs
-        # And it avoids matching truncated headlines like (#25... as PR #25
-        patterns = [
-            r"\(#(\d+)\)",  # (#123)
-            r"microsoft/onnxruntime/pull/(\d+)",
-            r"(?:^|\s|-)#(\d+)(?:\s|$)",  # #123 at start, or preceded by space/dash, and followed by space or end
-            r"\|\s*#(\d+)\s*\|",  # | #123 | (markdown table cell)
-        ]
-
-        results = []
-        for p in patterns:
-            results.extend(re.findall(p, text))
-        return [int(x) for x in set(results)]
-
-    # Matches patterns like #123 or https://github.com/microsoft/onnxruntime/pull/123
-    # Also handles ( #123) or similar in titles
-    prs = re.findall(r"(?:#|/pull/)(\d+)", text)
-    return [int(x) for x in set(prs)]
+
+def is_bot(name):
+    if not name:
+        return True
+    name_clean = name.strip().lstrip("@")
+    # Known bots and patterns
+    if name_clean in BOT_NAMES:
+        return True
+    if "[bot]" in name_clean.lower():
+        return True
+    if name_clean.lower().startswith("app/"):
+        return True
+    return False
+
+
+def is_invalid(name):
+    if not name:
+        return True
+    # If it's a bot, it's considered a valid identity for the CSV
+    if is_bot(name):
+        return False
+
+    name_clean = name.strip().lstrip("@")
+    # Paths, brackets, and code extensions
+    if "/" in name_clean or "\\" in name_clean or "[" in name_clean or "]" in name_clean:
+        return True
+    if any(name_clean.lower().endswith(ext) for ext in [".cmake", ".py", ".h", ".cc", ".cpp", ".txt", ".md"]):
+        return True
+    return False
 
 
 def get_prs_from_log(log_output, prs_base=None, log_file=None, scan_depth=100):

From ff6f523b24355ec75c0b8763fbc3e77311c20fd9 Mon Sep 17 00:00:00 2001
From: Tianlei Wu <tlwu@microsoft.com>
Date: Sat, 28 Feb 2026 20:56:01 +0000
Subject: [PATCH 07/11] format

---
 tools/python/cherry_pick.py          |  6 +++---
 tools/python/cherry_pick_utils.py    |  7 +++++--
 tools/python/compile_contributors.py | 12 ++++++++++--
 3 files changed, 18 insertions(+), 7 deletions(-)

diff --git a/tools/python/cherry_pick.py b/tools/python/cherry_pick.py
index 85bc7cba03617..3557ea467448c 100644
--- a/tools/python/cherry_pick.py
+++ b/tools/python/cherry_pick.py
@@ -24,14 +24,14 @@
 
 import argparse
 import json
-import re
 import sys
 from collections import defaultdict
+
 from cherry_pick_utils import (
-    run_command,
     check_preflight,
-    get_pr_number_from_subject,
     extract_pr_numbers,
+    get_pr_number_from_subject,
+    run_command,
 )
 
 
diff --git a/tools/python/cherry_pick_utils.py b/tools/python/cherry_pick_utils.py
index 34e3f8f0f87eb..cd58f3b589c35 100644
--- a/tools/python/cherry_pick_utils.py
+++ b/tools/python/cherry_pick_utils.py
@@ -1,7 +1,10 @@
+# Copyright (c) Microsoft Corporation. All rights reserved.
+# Licensed under the MIT License.
+
+import re
 import subprocess
 import sys
-import re
-import json
+
 
 def run_command(command_list, cwd=None, silent=False):
     """Run a command using a list of arguments for security (no shell=True)."""
diff --git a/tools/python/compile_contributors.py b/tools/python/compile_contributors.py
index 892f0f1648502..bb02c2807d08c 100644
--- a/tools/python/compile_contributors.py
+++ b/tools/python/compile_contributors.py
@@ -30,13 +30,17 @@
 import json
 import os
 import re
+
 from cherry_pick_utils import (
-    run_command,
     check_preflight,
-    get_pr_number_from_subject as get_pr_number,
     extract_pr_numbers,
+    run_command,
+)
+from cherry_pick_utils import (
+    get_pr_number_from_subject as get_pr_number,
 )
 
+
 def log_event(message, log_file=None):
     """Log a message to the console and an optional log file."""
     timestamp = datetime.datetime.now().strftime("%Y-%m-%d %H:%M:%S")
@@ -44,6 +48,8 @@ def log_event(message, log_file=None):
     print(message)  # Clean print for console UI
     if log_file:
         log_file.write(full_message + "\n")
+
+
 NAME_TO_LOGIN = {}  # Map full names to GitHub logins for consolidation
 VERIFIED_LOGINS = set()  # Track IDs known to be valid GitHub logins (vs free-form names)
 PR_CACHE = {}  # Cache for PR details to speed up multiple rounds referencing same PRs
@@ -63,6 +69,7 @@ def get_pr_details(pr_number):
     PR_CACHE[pr_number] = None
     return None
 
+
 def extract_authors_from_pr(details):
     authors = set()
     if not details:
@@ -120,6 +127,7 @@ def extract_authors_from_commit(commit_id):
 
     return authors
 
+
 # Bots to exclude from contributor lists
 BOT_NAMES = {
     "Copilot",

From 33687c55f8b5bcc137dc0411f2a0b4b500a63cfa Mon Sep 17 00:00:00 2001
From: Tianlei Wu <tlwu@microsoft.com>
Date: Wed, 4 Mar 2026 16:03:47 -0800
Subject: [PATCH 08/11] status

---
 tools/python/cherry_pick.py | 20 ++++++++++++++------
 1 file changed, 14 insertions(+), 6 deletions(-)

diff --git a/tools/python/cherry_pick.py b/tools/python/cherry_pick.py
index 3557ea467448c..a26cfb6d92ddd 100644
--- a/tools/python/cherry_pick.py
+++ b/tools/python/cherry_pick.py
@@ -141,6 +141,7 @@ def check_missing_dependencies(prs, branch):
         if pr.get("mergeCommit"):
             cherry_pick_oids.add(pr["mergeCommit"]["oid"])
 
+    conflicting_prs_count = 0
     for pr in prs:
         if not pr.get("mergeCommit"):
             continue
@@ -179,12 +180,19 @@ def check_missing_dependencies(prs, branch):
                     missing_commits[c] = (existing_title, existing_files)
 
         # Print deduplicated warnings
-        for missing_oid, (title, affected_files) in missing_commits.items():
-            files_str = ", ".join(affected_files)
-            print(
-                f"WARNING: PR #{number} ({oid}) modifies files that were also changed by commit {missing_oid} ({title}), "
-                f"which is not in the cherry-pick list. This may indicate missing related changes. Affected files: {files_str}"
-            )
+        if missing_commits:
+            conflicting_prs_count += 1
+            for missing_oid, (title, affected_files) in missing_commits.items():
+                files_str = ", ".join(affected_files)
+                print(
+                    f"WARNING: PR #{number} ({oid}) modifies files that were also changed by commit {missing_oid} ({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():

From 433e8e32e3778a657680f8ebe6994c7f398f8ff9 Mon Sep 17 00:00:00 2001
From: Tianlei Wu <tlwu@microsoft.com>
Date: Mon, 9 Mar 2026 21:13:56 -0700
Subject: [PATCH 09/11] review feedback

---
 tools/python/cherry_pick.py | 91 +++++++++++++++++++++----------------
 1 file changed, 53 insertions(+), 38 deletions(-)

diff --git a/tools/python/cherry_pick.py b/tools/python/cherry_pick.py
index a26cfb6d92ddd..7ac463b1ec274 100644
--- a/tools/python/cherry_pick.py
+++ b/tools/python/cherry_pick.py
@@ -24,8 +24,8 @@
 
 import argparse
 import json
+import os
 import sys
-from collections import defaultdict
 
 from cherry_pick_utils import (
     check_preflight,
@@ -72,6 +72,16 @@ def get_changed_files(oid):
     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):
     """Get the set of PR numbers already present in the target branch."""
     output = run_command(["git", "log", branch, "--oneline", "-n", "500"], silent=True)
@@ -118,11 +128,11 @@ def get_existing_pr_numbers(branch):
             if details:
                 # Collect sub-PRs from title, body, and commits
                 extracted_nums = []
-                extracted_nums.extend(extract_pr_numbers(details.get("title", "")))
-                extracted_nums.extend(extract_pr_numbers(details.get("body", "")))
+                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", "")))
+                    extracted_nums.extend(extract_pr_numbers(commit.get("messageHeadline", ""), strict=True))
 
                 for num in set(extracted_nums):
                     if num != pr_num_int:
@@ -155,8 +165,8 @@ def check_missing_dependencies(prs, branch):
 
         # 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, [list of affected files])
-        missing_commits = defaultdict(lambda: ("", []))
+        # missing_commits maps: missing_commit_oid -> {"title": ..., "files": [...]}
+        missing_commits = {}
 
         for filepath in files:
             # git log <cherry-pick-commit> --not <target-branch> -- <file>
@@ -173,19 +183,18 @@ def check_missing_dependencies(prs, branch):
                 if c == oid:
                     continue
                 if c not in cherry_pick_oids:
-                    existing_title, existing_files = missing_commits[c]
-                    if not existing_title:
-                        existing_title = title
-                    existing_files.append(filepath)
-                    missing_commits[c] = (existing_title, existing_files)
+                    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, (title, affected_files) in missing_commits.items():
-                files_str = ", ".join(affected_files)
+            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} ({title}), "
+                    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}"
                 )
 
@@ -205,7 +214,11 @@ def main():
     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="Wait limitation for PR fetching (default: 200)")
+    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
@@ -227,12 +240,28 @@ def main():
     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 = 0
-    skipped_count = 0
+    commit_count = len(cherry_pick_prs)
     with open(args.output, "w", encoding="utf-8") as f:
         if is_shell:
             f.write("#!/bin/bash\n")
@@ -244,41 +273,27 @@ def main():
             f.write(f"rem Cherry-pick {args.label} commits\n")
             f.write("rem Sorted by merge time (oldest first)\n\n")
 
-        for pr in prs:
+        for pr in cherry_pick_prs:
             number = pr["number"]
-            title = pr["title"]
-            safe_title = title.replace("\n", " ")
-
-            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
+            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")
-            commit_count += 1
 
     print(f"Generated {args.output} with {commit_count} commits ({skipped_count} skipped, already in branch).")
 
     # 3. Write PR Description Markdown (table format)
-    md_output = "cherry_pick_pr_description.md"
+    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 prs:
-            if not pr.get("mergeCommit"):
-                continue
+        for pr in cherry_pick_prs:
             number = pr["number"]
-            if number in existing_prs:
-                continue
-            title = pr["title"].replace("\n", " ")
+            title = escape_markdown_table_cell(pr["title"])
             oid = pr["mergeCommit"]["oid"]
             short_oid = oid[:10]
             f.write(f"| {short_oid} | #{number} | {title} |\n")
@@ -286,7 +301,7 @@ def main():
     print(f"Generated {md_output} with {commit_count} commits.")
 
     # 4. Dependency Check
-    check_missing_dependencies(prs, args.branch)
+    check_missing_dependencies(cherry_pick_prs, args.branch)
 
 
 if __name__ == "__main__":

From d7e922b8f482e9aa044871834a3234a55b2dd93a Mon Sep 17 00:00:00 2001
From: Tianlei Wu <tlwu@microsoft.com>
Date: Mon, 9 Mar 2026 21:35:46 -0700
Subject: [PATCH 10/11] address more feedback

---
 tools/python/cherry_pick.py       | 22 ++++++++++++++++------
 tools/python/cherry_pick_utils.py | 12 +++++++++---
 2 files changed, 25 insertions(+), 9 deletions(-)

diff --git a/tools/python/cherry_pick.py b/tools/python/cherry_pick.py
index 7ac463b1ec274..b90924a785680 100644
--- a/tools/python/cherry_pick.py
+++ b/tools/python/cherry_pick.py
@@ -82,9 +82,9 @@ def escape_markdown_table_cell(text):
     return sanitize_title(text).replace("|", "\\|")
 
 
-def get_existing_pr_numbers(branch):
+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", "500"], silent=True)
+    output = run_command(["git", "log", branch, "--oneline", "-n", str(log_limit)], silent=True)
     if not output:
         return set()
     pr_numbers = set()
@@ -115,7 +115,10 @@ def get_existing_pr_numbers(branch):
         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_out = run_command(["gh", "pr", "view", pr_num, "--json", "title,body,commits"], silent=True)
+                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)
@@ -145,11 +148,18 @@ def check_missing_dependencies(prs, branch):
     """Check for potential missing dependencies (conflicts)."""
     print("\nChecking for potential missing dependencies (conflicts)...")
 
-    # Collect OIDs being cherry-picked
+    # Collect OIDs being cherry-picked and all their ancestor commits
     cherry_pick_oids = set()
     for pr in prs:
         if pr.get("mergeCommit"):
-            cherry_pick_oids.add(pr["mergeCommit"]["oid"])
+            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:
@@ -236,7 +246,7 @@ def main():
     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)
+    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}'.")
 
diff --git a/tools/python/cherry_pick_utils.py b/tools/python/cherry_pick_utils.py
index cd58f3b589c35..7460761593a8c 100644
--- a/tools/python/cherry_pick_utils.py
+++ b/tools/python/cherry_pick_utils.py
@@ -52,9 +52,15 @@ def check_preflight():
         )
         return False
 
-    gh_auth = run_command(["gh", "auth", "status"], silent=True)
-    if not gh_auth:
-        print("Error: GitHub CLI not authenticated. Please run 'gh auth login'.", file=sys.stderr)
+    # gh auth status outputs to stderr, so run_command returns empty stdout even on success.
+    # Use subprocess directly to check the return code.
+    try:
+        auth_result = subprocess.run(["gh", "auth", "status"], capture_output=True, text=True, check=False)
+        if auth_result.returncode != 0:
+            print("Error: GitHub CLI not authenticated. Please run 'gh auth login'.", file=sys.stderr)
+            return False
+    except FileNotFoundError:
+        print("Error: GitHub CLI (gh) not found.", file=sys.stderr)
         return False
 
     return True

From 51e1db159f978e2680008a636c5a1a1d243c6a15 Mon Sep 17 00:00:00 2001
From: Tianlei Wu <tlwu@microsoft.com>
Date: Mon, 9 Mar 2026 22:09:52 -0700
Subject: [PATCH 11/11] final feedback

---
 tools/python/cherry_pick.py | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/tools/python/cherry_pick.py b/tools/python/cherry_pick.py
index b90924a785680..96a91c5e74c22 100644
--- a/tools/python/cherry_pick.py
+++ b/tools/python/cherry_pick.py
@@ -66,7 +66,7 @@ def get_merged_prs(repo, label, limit=200):
 
 def get_changed_files(oid):
     """Get list of files changed in a commit."""
-    output = run_command(["git", "diff-tree", "--no-commit-id", "--name-only", "-r", oid], silent=True)
+    output = run_command(["git", "diff-tree", "--no-commit-id", "--name-only", "-m", "-r", oid], silent=True)
     if output:
         return output.strip().splitlines()
     return []