Description

jeffro256

opened on Sep 28, 2026

Issue body actions

Consider the following code snippet:

import git
repo = git.Repo('GitPython')
tag = repo.rev_parse('3.1.62')
commit = git.Commit(repo, tag.binsha)
index_file = git.IndexFile.new(repo, commit.tree)

If GitPython is a repo in the CWD, then upon running the above code, you will see a traceback like so:

Traceback (most recent call last):
  File "my_python_file.py", line 5, in <module>
    index_file = git.IndexFile.new(repo, commit.tree)
  File "/usr/lib/python3.14/site-packages/git/index/base.py", line 330, in new
    base_entries = aggressive_tree_merge(repo.odb, tree_sha_bytes)
  File "/usr/lib/python3.14/site-packages/git/index/fun.py", line 505, in aggressive_tree_merge
    for entry in traverse_tree_recursive(odb, tree_shas[-1], ""):
                 ~~~~~~~~~~~~~~~~~~~~~~~^^^^^^^^^^^^^^^^^^^^^^^^
  File "/usr/lib/python3.14/site-packages/git/objects/fun.py", line 271, in traverse_tree_recursive
    data = tree_entries_from_data(odb.stream(tree_sha).read())
  File "/usr/lib/python3.14/site-packages/git/objects/fun.py", line 111, in tree_entries_from_data
    while data[i] != 0:
          ~~~~^^^
IndexError: index out of range

This throws an out-of-bounds array index exception when trying to parse the tag's database stream as a tree, which should fail. However, the real user error occurred when trying to make a Commit object from a tag object's binsha: commit = git.Commit(repo, tag.binsha). The Commit constructor should raise an Exception with a message to this effect. For example, "Given sha value does not refer to a commit`.

Activity

1. Byron

   added a commit that references this issue on Sep 28, 2026

   fix: validate object types when loading binary SHAs (#2254)

   0610404
2. Byron commented on Sep 28, 2026

   

   Byron

   on Sep 28, 2026

   Member

   More actions

   I have a partial fix for it, which is in the realm of what can be done without affecting performance noticably.

   Codex

   The proposed fix catches the type mismatch before parsing: accessing Commit(repo, tag.binsha).tree for an annotated tag now raises a clear ValueError identifying the SHA, actual type, and expected type, instead of the unrelated IndexError.

   The audit also improved related paths:

   • The same validation covers blobs, trees, tags, and raw tree traversals used by index operations.
   • IndexFile.new() now preserves binary SHA inputs instead of converting them into Python string representations.
   • Submodule diffs now represent their commit IDs without incorrectly treating them as blobs.

   Commit(repo, tag.binsha) itself still succeeds. Constructors remain lazy because checking there would require a database lookup for every object created and interfere with representing placeholders or objects unavailable locally. Validation happens when metadata or data is requested.
3. Byron

   added 2 commits that reference this issue on Sep 28, 2026

   fix: validate size lookups and preserve binary SHAs (#2254)

   7ed5433

   fix: validate size lookups and preserve binary SHAs (#2254)

   6bf77c0
4. Byron

   closed this as completedin #2256on Sep 28, 2026

Issue actions