[arunimshukla]

Summary

Snapshot root-level accessors could panic after valid insertions reach MAX_DEPTH. The tree accepts that depth, but levels has only MAX_DEPTH slots; the root is also stored separately. Reading levels[depth] at maximum depth indexes beyond that array.

The fix serves root-level reads from the snapshot's existing root value. Empty trees still have zero nodes at level zero; non-empty roots have one. Invalid root indices return IndexOutOfRange. This includes maximum-depth trees before they reach full leaf capacity.

The diff remains limited to the two accessors and their tests.

Review changes

• Clarified why root-level reads use self.root, including that the branch applies at every depth.
• Removed proof generation checks from the accessor regression because proof generation does not call either accessor.
• Combined the two accessor tests into one regression with shared assertions.
• Removed duplicate above-root checks already covered by snapshot_get_node_and_level_len.
• Added regression/control descriptions and given/when/then annotations.

Regression coverage

The consolidated regression checks root length, root hash, and invalid root indices through the public snapshot interface. It covers sequential and batch insertion for binary depths one and three and ternary depth two, including every leaf count in each configuration's maximum-depth range. A separate control preserves empty and single-leaf behaviour.

With the revised tests and the upstream implementation at e5061c8, cargo test -p rotortree --test tree snapshot_ reports 1 failed, 3 passed. The maximum-depth regression fails with an index-out-of-bounds panic. Restoring the patch produces 4 passed, 0 failed.

Validation

Rechecked the revised code using Rust 1.95.0:

Check	Result
cargo test -p rotortree	107 passed
cargo test -p rotortree --all-features	180 passed
cargo test -p rotortree --no-default-features	107 passed
cargo clippy -p rotortree --all-features -- -D warnings	Passed
cargo +nightly fmt --all --check	Passed
git diff --check	Passed

Counts include doctests where applicable. The test count drops by one because the accessor regressions are consolidated. These are local results; the repository's complete CI feature powerset is checked separately by CI. The updated upstream CI run currently requires maintainer approval to run the fork workflow.

The revised PR contains one commit with a GitHub-verified signature and a sign-off.

[rymnc]

thank you for your contribution! please sign the commit so that we can merge after you address the comments :)