Enable commit-specific git fetch for compatibility with external forks.

[leander-dsouza]

Basic Info

Info	Please fill out this column
Ticket(s) this addresses	resolves #8
Primary OS tested on	Ubuntu
Is this a breaking change?	No
Does this PR contain AI generated software?	No

---

Description of contribution in a few bullet points

• Enabled commit-specific fetching of origin on non-shallow clones containing hashes.
• This allows us to clone PRs from repositories not tied to the source branch for history (external forks).

Description of how this change was tested

1. Testing against vcs2l old pull requests

• First, clone the codebase:

  git clone git@github.com:ros-infrastructure/vcs2l.git vcs2l-ext-fork

• Next, use the following repos.yaml to get the commit hash of the following external fork:

  ---
  repositories:
    vcs2l-ext-fork:
      type: git
      url: 'https://github.com/ros-infrastructure/vcs2l.git'
      version: '4ad2dd14b0033c275e5faff7935c4a7d7581fee4'

• Finally, update the existing repository:

  vcs import --input repos.yaml

  This will successfully checkout the repository to the specific hash.
  This step fails without this contribution with the following error message:

  === ./vcs2l-ext-fork (git) ===
  Could not checkout ref '4ad2dd14b0033c275e5faff7935c4a7d7581fee4': fatal: reference is not a tree:    4ad2dd14b0033c275e5faff7935c4a7d7581fee4

2. Test against the reproduction step in the parent issue - dirk-thomas/vcstool#258

• Reproduction measures specified by @timor-raiman

  git clone git@github.com:timor-raiman/action-ros-ci.git -b vcs-issue && cd action-ros-ci
  git fetch origin +refs/pull/3/*:refs/remotes/origin/pr/3/*
  export PRMERGE_SHA=`git rev-parse origin/pr/3/merge`
  rm -rf new-path/ && mkdir -p new-path/
  cat test-vcs-issue.repo | envsubst | vcs import --force --recursive new-path/

[codecov-commenter]

Codecov Report

❌ Patch coverage is 0% with 14 lines in your changes missing coverage. Please review.
✅ Project coverage is 26.76%. Comparing base (fbd0321) to head (f24838b).

Files with missing lines	Patch %	Lines
vcs2l/clients/git.py	0.00%	14 Missing ⚠️

Additional details and impacted files

@@           Coverage Diff           @@
##             main      #83   +/-   ##
=======================================
  Coverage   26.76%   26.76%           
=======================================
  Files          31       31           
  Lines        2298     2298           
  Branches      405      408    +3     
=======================================
  Hits          615      615           
  Misses       1623     1623           
  Partials       60       60           

☔ View full report in Codecov by Harness.
📢 Have feedback on the report? Share it here.

🚀 New features to boost your workflow:

• ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.

[Copilot]

Pull Request Overview

This pull request adds support for fetching specific commit hashes in non-shallow Git import mode, and improves test skipping for VCS tools that may not be installed.

• Refactored version type determination to work for both shallow and non-shallow modes
• Added explicit fetch of commit hashes for non-shallow repository imports
• Added test skipping when SVN or Mercurial tools are not available

Reviewed Changes

Copilot reviewed 4 out of 4 changed files in this pull request and generated no comments.

File	Description
vcs2l/clients/git.py	Refactored to determine version type earlier and fetch specific commit hashes in non-shallow mode for both existing and new repositories
test/test_commands.py	Added utility variables and skip decorator for tests requiring SVN and Mercurial
test/reimport_force.txt	Updated expected output to include fetch messages for commit hash imports
test/import.txt	Updated expected output to include fetch messages for commit hash imports

[cottsay]

The checks added here are part of the fixture setup in __init__.py. What's the motivation to duplicate them here?

[leander-dsouza]

They were initially duplicated as Subversion, and Mercurial ignores were flaky in Windows 2025 in previous workflow runs.

However, they currently work as expected, so I have dropped the additional commit.

[KmoM88]

Overall LGTM.

I have left a suggestion regarding the execution workflow.

Also as a side note: What do you think about adding a dedicated test case that requests a commit from an unmerged, orphaned branch (simulating a real external fork) to prevent any future regressions?

Nice work, waiting to hear your thoughts.

[KmoM88]

For a non-shallow commit-specific clone this is doing a full git clone and then does a git fetch for the target hash. For large repositories, this fetches a lot of unnecessary data.
What would you think about bypassing git clone entirely and using the git init + git fetch <hash> path for all precise tags and hashes (both shallow and non-shallow)?

Adjusting the conditional to redirect any hash or tag to the else block and toggling the --depth 1 flag on the fetch command dynamically if command.shallow is True.

This could save significant bandwidth for repositories with large default branches.

What do you think?

[leander-dsouza]

Yes, I agree.
Now, hash/tag imports go through git init + git fetch <ref> for both shallow and non-shallow clones, with the --depth 1 added for shallow ones.

One thing I ran into in the hash case is that git fetch origin <hash> does not pull any tags, so git describe --tags breaks. Therefore, I have added the --tags to the fetch command to preserve it, and now subsequent expect .txt files have been updated to match it.

[leander-dsouza]

Nice work

Thank you <3

Also as a side note: What do you think about adding a dedicated test case that requests a commit from an unmerged, orphaned branch (simulating a real external fork) to prevent any future regressions?

All tests from #72 have been designed to run without an internet connection to ensure compatibility across various operating systems.

So, I do not think we can add a real external fork to validate this behaviour.

[KmoM88]

Good job on the contribution, the proposed changes look correct. The logic for fetching commit-specific clones seems correct. Overall LGTM! Thank you!