Support ChangeLog for the new branch model [skip ci] by NvTimLiu · Pull Request #13687 · NVIDIA/cudf-spark

NvTimLiu · 2025-10-29T03:16:59Z

1, Retrieve commit hashes for a release after enabling the new branch model.

2, Create a query to fetch pull request (PR) information from GitHub using commit hashes.

3, Support retrieving PR ChangeLogs for both the old and new branch models.

1, Retrieve commit hashes for a release after enabling the new branch model. 2, Create a query to fetch pull request (PR) information from GitHub using commit hashes. 3, Support retrieving PR changelogs for both the old and new branch models. Signed-off-by: timl <timl@nvidia.com>

greptile-apps

Greptile Overview

Greptile Summary

This PR modernizes the changelog generation script to accommodate the repository's branch naming transition from branch-YY.MM to release/YY.MM, starting at version 25.12. The implementation adds a commit-based PR lookup path using git log and GitHub's GraphQL API, while preserving backward compatibility with the legacy branch-name query method. The logic pivots on a hardcoded cutoff (FROM_RELEASE = '25.12') and handles three scenarios: both releases after the cutoff (commit-based), mixed mode at the cutoff, and both before the cutoff (branch-based). The script now sorts releases in descending order to ensure git range syntax works correctly, since get_commits() expects the newest release first for constructing from_branch..to_branch ranges.

Important Files Changed

Filename	Score	Overview
scripts/generate-changelog	3/5	Adds commit-based PR retrieval for new branch model, with hybrid logic handling three release scenarios and git log parsing.

Confidence score: 3/5

This PR introduces significant control-flow branching and external process calls (git, GraphQL), increasing the surface area for runtime errors and edge cases.
Score reflects three concerns: line 325 incorrectly treats a sorted list as a set (will always take first element), line 250 silently takes only the first associated PR per commit (losing multi-PR commits), and line 255uses a bare except that swallows all exceptions including keyboard interrupts and network failures.
Pay close attention to the process_pr() function (lines 323-345) and get_pr_via_commits() (lines 244-258), as these contain the problematic set/list confusion, silent exception handling, and assumptions about commit-to-PR cardinality.

Sequence Diagram

sequenceDiagram
    participant User
    participant main
    participant process_changelog
    participant process_pr
    participant get_commits
    participant get_pr_via_commits
    participant fetch
    participant post
    participant GitHub API
    participant form_changelog

    User->>main: "Execute script with --releases, --token, --path"
    main->>main: "Parse arguments and validate GITHUB_TOKEN"
    main->>process_changelog: "Process PRs for releases"
    process_changelog->>process_pr: "Get PRs based on releases"
    
    alt Current version > FROM_RELEASE (25.12)
        process_pr->>get_commits: "Get commit hashes between release tags"
        get_commits->>get_commits: "Execute git log for each release pair"
        get_commits-->>process_pr: "Return ver_commits dict"
        process_pr->>get_pr_via_commits: "Get PRs from commit hashes"
        loop For each commit SHA
            get_pr_via_commits->>post: "Query PR by commit SHA"
            post->>GitHub API: "POST query_pr_by_commit with SHA"
            GitHub API-->>post: "Return PR data"
            post-->>get_pr_via_commits: "Return response"
            get_pr_via_commits->>get_pr_via_commits: "Add PR to list if merged and unique"
        end
        get_pr_via_commits-->>process_pr: "Return PR list"
    else Current version == FROM_RELEASE
        process_pr->>get_commits: "Get commits for FROM_RELEASE"
        get_commits-->>process_pr: "Return ver_commits"
        process_pr->>get_pr_via_commits: "Get PRs from commits"
        get_pr_via_commits-->>process_pr: "Return PR list"
        process_pr->>fetch: "Fetch PRs for previous branch"
        fetch->>post: "Query PRs with baseRefName"
        post->>GitHub API: "POST query_pr"
        GitHub API-->>post: "Return PR data"
        post-->>fetch: "Return response"
        fetch-->>process_pr: "Return additional PRs"
    else Current version < FROM_RELEASE
        loop For each release
            process_pr->>fetch: "Fetch PRs for branch-{release}"
            fetch->>post: "Query PRs with baseRefName"
            post->>GitHub API: "POST query_pr"
            GitHub API-->>post: "Return PR data"
            post-->>fetch: "Return response"
            fetch-->>process_pr: "Extend PR list"
        end
    end
    
    process_pr-->>process_changelog: "Return all PRs"
    process_changelog->>process_changelog: "Categorize PRs by project and labels"
    process_changelog-->>main: "Update changelog dict with PRs"
    
    main->>process_changelog: "Process Issues for releases"
    process_changelog->>process_issue: "Get issues from last 3*N months"
    process_issue->>fetch: "Fetch issues with since timestamp"
    loop While hasNextPage
        fetch->>post: "Query issues"
        post->>GitHub API: "POST query_issue"
        GitHub API-->>post: "Return issue data"
        post-->>fetch: "Return response"
        fetch->>fetch: "Append issues and update cursor"
    end
    fetch-->>process_issue: "Return all issues"
    process_issue-->>process_changelog: "Return issues"
    process_changelog->>process_changelog: "Categorize issues by project and labels"
    process_changelog-->>main: "Update changelog dict with issues"
    
    main->>form_changelog: "Generate markdown from changelog dict"
    form_changelog->>form_changelog: "Sort projects, format sections"
    form_changelog->>form_changelog: "Write to file"
    form_changelog-->>main: "Changelog file created"
    main-->>User: "Done"

_{1 file reviewed, 3 comments}

_{Edit Code Review Agent Settings | Greptile}

NvTimLiu · 2025-10-29T03:18:39Z

Build PASS:

Copilot

Pull Request Overview

This PR modifies the changelog generation script to support a transition from branch-based releases to release-branch-based releases starting with version 25.12. The script now uses git commit history to find PRs for releases after 25.12, while continuing to use the GitHub API for earlier releases.

Key changes:

Added logic to fetch PRs via git commit history for releases >= 25.12
Implemented version-based branching strategy detection (branch-YY.MM vs release/YY.MM)
Maintained backward compatibility for releases before 25.12

💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.

Signed-off-by: timl <timl@nvidia.com>

pxLi · 2025-10-31T02:12:08Z

some follow updates could be taken care later:

Add NOTE for compatibility support
Concurrently send requests to query PR by commit
Dynamically detect local git remote ref

NvTimLiu · 2025-10-31T07:04:40Z

build

greptile-apps

Greptile Overview

Greptile Summary

Extended changelog generation to support the new branch model (release/YY.MM) starting from version 25.12, while maintaining backward compatibility with the old model (branch-YY.MM). Adds commit-based PR retrieval using GraphQL queries, and implements logic to determine which branch naming convention to use based on release versions.

Key changes:

Added get_commits() to retrieve commit hashes between release branches using git log
Added get_pr_via_commits() to fetch PR metadata from GitHub API using commit SHAs
Modified process_pr() to route to old vs new branch model based on FROM_RELEASE threshold (25.12)
Added get_prev_release_version() utility to calculate previous release version

Critical issues found:

Logic bug in process_pr() line 330: assumes all releases use new model when highest version > FROM_RELEASE, but list can contain older versions requiring old model
Bug in get_commits() line 214: always uses origin/release/{to_rel} prefix, even for releases before FROM_RELEASE
Type hint mismatches between function signatures and actual parameters
Cherry-picked commit handling may select wrong PR (existing comment)

Confidence Score: 2/5

This PR has critical logic bugs that will cause failures when generating changelogs for mixed old/new release versions
Score reflects two critical logic bugs: (1) process_pr() incorrectly routes all releases to new branch model when only the highest version exceeds FROM_RELEASE, causing git failures for older releases, and (2) get_commits() always uses release/ prefix for to_branch regardless of version. These bugs will break changelog generation for common scenarios like releases=['26.02', '25.10']. Additionally, there are type hint inconsistencies and the existing cherry-pick issue remains unresolved.
Critical attention needed for scripts/generate-changelog lines 214, 330-332 to fix branch model routing logic

Important Files Changed

File Analysis

Filename	Score	Overview
scripts/generate-changelog	2/5	Added support for new release branch model (release/YY.MM), but has critical logic bug in process_pr that incorrectly assumes all releases use new model when highest version > FROM_RELEASE, and get_commits always uses release/ prefix for to_branch

Sequence Diagram

sequenceDiagram
    participant User
    participant main
    participant process_pr
    participant get_commits
    participant get_pr_via_commits
    participant GitHub API

    User->>main: --releases=26.02,25.12
    main->>main: Sort releases descending → [26.02, 25.12]
    main->>process_pr: releases=[26.02, 25.12]
    
    alt current_ver > FROM_RELEASE (26.02 > 25.12)
        process_pr->>get_commits: All releases
        loop For each release
            get_commits->>get_commits: Determine from/to branches
            Note over get_commits: to_branch = origin/release/{ver}<br/>from_branch = origin/release/ or branch-
            get_commits->>get_commits: git log from_branch..to_branch
        end
        get_commits-->>process_pr: ver_commits dict
        process_pr->>get_pr_via_commits: ver_commits
        loop For each commit
            get_pr_via_commits->>GitHub API: query_pr_by_commit(sha)
            GitHub API-->>get_pr_via_commits: PR info
            get_pr_via_commits->>get_pr_via_commits: Deduplicate & filter
        end
        get_pr_via_commits-->>process_pr: PR list
    else current_ver == FROM_RELEASE (25.12)
        process_pr->>get_commits: {FROM_RELEASE}
        process_pr->>get_pr_via_commits: commits for 25.12
        process_pr->>GitHub API: fetch PRs for branch-{prev_ver}
        GitHub API-->>process_pr: Old model PRs
    else current_ver < FROM_RELEASE
        loop For each release
            process_pr->>GitHub API: fetch(baseRefName=branch-{rel})
            GitHub API-->>process_pr: PRs
        end
    end
    
    process_pr-->>main: Combined PR list
    main->>main: Generate changelog markdown

_{1 file reviewed, 5 comments}

_{Edit Code Review Agent Settings | Greptile}

greptile-apps · 2025-10-31T07:08:26Z

+    if current_ver_float > based_rel:
+        ver_commits = get_commits(releases)
+        pr = get_pr_via_commits(ver_commits, token)


logic: Logic assumes all releases use new branch model when current_ver > FROM_RELEASE, but releases list can contain versions ≤ FROM_RELEASE. Example: if releases=['26.02', '25.10'], it calls get_commits which tries origin/release/25.10 (doesn't exist, should be origin/branch-25.10). Consider filtering releases or checking each version individually.

greptile-apps · 2025-10-31T07:08:26Z

+    ver_commits = {}
+    count = len(rel_list)  # descending version order assured
+    for i, to_rel in enumerate(rel_list):
+        to_branch = f"origin/release/{to_rel}"


logic: Always uses origin/release/{to_rel} even for releases before FROM_RELEASE. Should check if float(to_rel) >= float(FROM_RELEASE) before deciding branch name, similar to from_branch logic on lines 221-224.

greptile-apps · 2025-10-31T07:08:27Z

+# Get the commit hashes between two branches or release tags.
+# param releases: set of release versions, e.g. {'YY.MM2', 'YY.MM1'}
+# return: dict of commit hashes, e.g. {YY.MM2: [sha1, sha2, ...], YY.MM1: [shaX, shaY, ...]}
+def get_commits(releases: set):


style: Type hint says set but receives sorted list from line 435. Update to def get_commits(releases: list): or keep as set and handle sorting internally.

greptile-apps · 2025-10-31T07:08:28Z

+# param ver_commits, e.g. {v1: [sha1, sha2, ...], v2: [shaX, shaY, ...]}
+# param token: the token for the API
+# return: list of PRs associated with the commit hashes, e.g. [{PR1 info}, {PR2 info}, ...]
+def get_pr_via_commits(ver_commits: set, token: str):


syntax: Type hint says set but receives dict from get_commits. Should be def get_pr_via_commits(ver_commits: dict, token: str):

greptile-apps · 2025-10-31T07:08:29Z

+        ]
+
+        # Use check=True to raise exception if git fails, making errors explicit
+        result = subprocess.run(git_log_args, capture_output=True, text=True, check=True)


style: Git command will fail silently returning empty commits if branches don't exist. Consider checking branch existence first or handling the specific error case separately from other exceptions.

… the changelog Signed-off-by: timl <timl@nvidia.com>

NvTimLiu · 2025-10-31T08:29:55Z

build

greptile-apps

Greptile Overview

Greptile Summary

Extends changelog generation to support the new branch model where release branches transition from branch-YY.MM to release/YY.MM starting from version 25.12.

Key Changes:

Adds commit-hash-based PR retrieval using get_commits() and get_pr_via_commits() for releases ≥ 25.12
Implements hybrid approach in process_pr() to handle old model (baseRefName queries), new model (commit-based), and transition release (25.12)
Adds FROM_RELEASE constant (25.12) to mark the transition point
Introduces get_prev_release_version() to calculate 2-month-prior releases

Critical Issues Found:

Line 214: to_branch always uses origin/release/{to_rel} format, which will fail for mixed-version lists containing releases < 25.12
Line 210: Converting sorted list back to unsorted list loses the descending order required by loop logic
Line 325: Assumes releases (typed as set) is ordered, but works only due to line 436's conversion to sorted list
Type inconsistencies between function signatures (expecting set) and actual usage (passing sorted list)

Confidence Score: 2/5

PR has critical logic errors that will cause failures with mixed-version release lists
The implementation has fundamental issues: (1) to_branch determination on line 214 doesn't check if to_rel uses old/new model, causing git commands to fail for pre-25.12 releases, (2) type inconsistencies where functions expect sets but receive lists, (3) ordering dependencies that break when converting sorted lists back to unordered structures. These will cause runtime failures when generating changelogs for mixed-version scenarios
scripts/generate-changelog requires fixes to branch name logic, type consistency, and ordering preservation

Important Files Changed

File Analysis

Filename	Score	Overview
scripts/generate-changelog	2/5	Adds new branch model support with commit-based PR retrieval, but has critical logic errors in branch name determination, type inconsistencies, and ordering assumptions

Sequence Diagram

sequenceDiagram
    participant User
    participant main
    participant process_pr
    participant get_commits
    participant get_pr_via_commits
    participant GitHub API

    User->>main: Call with releases (e.g., "26.02,25.12")
    main->>main: Parse and sort releases descending
    main->>process_pr: Pass sorted releases
    
    alt current_ver > FROM_RELEASE (25.12)
        process_pr->>get_commits: Get commits for all releases
        loop For each release
            get_commits->>get_commits: Determine to_branch (origin/release/YY.MM)
            get_commits->>get_commits: Determine from_branch (check < FROM_RELEASE)
            get_commits->>GitHub API: git log from_branch..to_branch
            GitHub API-->>get_commits: Return commit hashes
        end
        get_commits-->>process_pr: Return ver_commits dict
        process_pr->>get_pr_via_commits: Pass ver_commits and token
        loop For each commit in each version
            get_pr_via_commits->>GitHub API: Query PR by commit SHA
            GitHub API-->>get_pr_via_commits: Return associated PRs
            get_pr_via_commits->>get_pr_via_commits: Filter and deduplicate
        end
        get_pr_via_commits-->>process_pr: Return PR list
    else current_ver == FROM_RELEASE
        process_pr->>get_commits: Get commits for FROM_RELEASE only
        process_pr->>get_pr_via_commits: Get PRs via commits
        process_pr->>GitHub API: Fetch PRs for previous release (old model)
        GitHub API-->>process_pr: Return PRs
    else current_ver < FROM_RELEASE
        loop For each release
            process_pr->>GitHub API: Fetch PRs by baseRefName (branch-YY.MM)
            GitHub API-->>process_pr: Return PRs
        end
    end
    
    process_pr-->>main: Return all PRs
    main->>main: Process and generate changelog

_{1 file reviewed, 3 comments}

_{Edit Code Review Agent Settings | Greptile}

greptile-apps · 2025-10-31T08:31:23Z

+    ver_commits = {}
+    count = len(rel_list)  # descending version order assured
+    for i, to_rel in enumerate(rel_list):
+        to_branch = f"origin/release/{to_rel}"


logic: to_branch assumes new branch model for all releases in the list, but mixed release versions (e.g., ['26.02', '25.10']) will fail for older releases

Suggested change

to_branch = f"origin/release/{to_rel}"

to_rel_float = float(to_rel)

if to_rel_float < float(FROM_RELEASE):

to_branch = f"origin/branch-{to_rel}"

else:

to_branch = f"origin/release/{to_rel}"

greptile-apps · 2025-10-31T08:31:24Z

+# param releases: set of release versions, e.g. {'YY.MM2', 'YY.MM1'}
+# return: dict of commit hashes, e.g. {YY.MM2: [sha1, sha2, ...], YY.MM1: [shaX, shaY, ...]}
+def get_commits(releases: set):
+    rel_list = list(releases)


logic: Converting sorted list to unsorted list loses the ordering that line 436 established. Ordering matters for the loop logic on lines 216-219

Suggested change

rel_list = list(releases)

rel_list = releases # Already sorted in descending order

greptile-apps · 2025-10-31T08:31:25Z

 def process_pr(releases: set, token: str):
    pr = []
-    for rel in releases:
+    current_ver = list(releases)[0]


logic: list(releases)[0] assumes ordering but releases is typed as set (unordered). Works due to line 436 conversion but type hint is incorrect

Suggested change

current_ver = list(releases)[0]

current_ver = releases[0] # releases is already a sorted list

pxLi · 2025-10-31T09:39:16Z

bad base ref

NvTimLiu self-assigned this Oct 29, 2025

Copilot AI review requested due to automatic review settings October 29, 2025 03:17

NvTimLiu requested a review from a team as a code owner October 29, 2025 03:17

NvTimLiu added the build Related to CI / CD or cleanly building label Oct 29, 2025

NvTimLiu marked this pull request as draft October 29, 2025 03:17

greptile-apps Bot reviewed Oct 29, 2025

View reviewed changes

Comment thread scripts/generate-changelog

Comment thread scripts/generate-changelog

Comment thread scripts/generate-changelog Outdated

NvTimLiu changed the title ~~Support changelog for the new branch model:~~ Support ChangeLog for the new branch model [skip ci] Oct 29, 2025

Copilot AI reviewed Oct 29, 2025

View reviewed changes

Comment thread scripts/generate-changelog Outdated

Comment thread scripts/generate-changelog

Comment thread scripts/generate-changelog

Comment thread scripts/generate-changelog Outdated

Comment thread scripts/generate-changelog

Comment thread scripts/generate-changelog

Update for the 'Copilot' review suggestions

ec8f8bf

Signed-off-by: timl <timl@nvidia.com>

pxLi previously approved these changes Oct 31, 2025

View reviewed changes

NvTimLiu marked this pull request as ready for review October 31, 2025 07:04

greptile-apps Bot reviewed Oct 31, 2025

View reviewed changes

Add a note indicating that only the last two releases are included in…

10e1f7a

… the changelog Signed-off-by: timl <timl@nvidia.com>

NvTimLiu dismissed pxLi’s stale review via 10e1f7a October 31, 2025 08:28

greptile-apps Bot reviewed Oct 31, 2025

View reviewed changes

NvTimLiu merged commit f34ab9c into NVIDIA:branch-25.12 Oct 31, 2025
60 checks passed

-        to_branch = f"origin/release/{to_rel}"
+        to_rel_float = float(to_rel)
+        if to_rel_float < float(FROM_RELEASE):
+            to_branch = f"origin/branch-{to_rel}"
+        else:
+            to_branch = f"origin/release/{to_rel}"

	rel_list = list(releases)
	rel_list = releases # Already sorted in descending order

	current_ver = list(releases)[0]
	current_ver = releases[0] # releases is already a sorted list

Uh oh!

Conversation

NvTimLiu commented Oct 29, 2025

Uh oh!

greptile-apps Bot left a comment

Choose a reason for hiding this comment

Greptile Overview

Greptile Summary

Important Files Changed

Confidence score: 3/5

Sequence Diagram

Uh oh!

Uh oh!

Uh oh!

Uh oh!

NvTimLiu commented Oct 29, 2025

Uh oh!

Copilot AI left a comment

Choose a reason for hiding this comment

Pull Request Overview

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

pxLi commented Oct 31, 2025 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

NvTimLiu commented Oct 31, 2025

Uh oh!

greptile-apps Bot left a comment

Choose a reason for hiding this comment

Greptile Overview

Greptile Summary

Confidence Score: 2/5

Important Files Changed

Sequence Diagram

Uh oh!

greptile-apps Bot Oct 31, 2025

Choose a reason for hiding this comment

Uh oh!

greptile-apps Bot Oct 31, 2025

Choose a reason for hiding this comment

Uh oh!

greptile-apps Bot Oct 31, 2025

Choose a reason for hiding this comment

Uh oh!

greptile-apps Bot Oct 31, 2025

Choose a reason for hiding this comment

Uh oh!

greptile-apps Bot Oct 31, 2025

Choose a reason for hiding this comment

Uh oh!

NvTimLiu commented Oct 31, 2025

Uh oh!

greptile-apps Bot left a comment

Choose a reason for hiding this comment

Greptile Overview

Greptile Summary

Confidence Score: 2/5

Important Files Changed

Sequence Diagram

Uh oh!

greptile-apps Bot Oct 31, 2025

Choose a reason for hiding this comment

Uh oh!

greptile-apps Bot Oct 31, 2025

Choose a reason for hiding this comment

Uh oh!

greptile-apps Bot Oct 31, 2025

Choose a reason for hiding this comment

Uh oh!

Uh oh!

pxLi commented Oct 31, 2025

Uh oh!

Reviewers

Assignees

Labels

pxLi commented Oct 31, 2025 •

edited

Loading