fix: Pipeline TypeError: can only concatenate list (not "NoneType") to list Using Sou (#5518)

[aviruthen]

Description

Fix TypeError in Pipeline.upsert() when SourceCode.requirements is None

Fixes aws#5518

Problem

When calling pipeline.upsert() with a TrainingStep whose SourceCode object has no requirements set (defaults to None), a TypeError is raised because None is concatenated with a list in the code hashing utilities.

Specifically, get_processing_code_hash() has unguarded + dependencies list concatenation that fails when dependencies is None. While get_training_code_hash() already has if dependencies: guards, the processing code hash function did not have this protection.

Solution

Added defensive None-to-empty-list coercion (dependencies = dependencies or []) at the start of get_processing_code_hash() to prevent TypeError when dependencies is None and gets concatenated with a list.

The get_training_code_hash() function already handles None correctly via its existing if dependencies: conditional checks.

Testing

Added test cases to verify:

• get_processing_code_hash() works correctly when dependencies=None (with and without source_dir)
• get_training_code_hash() works correctly when dependencies=None (with source_dir and with entry_point only)

Related Issue

Fixes aws#5518

Changes Made

The bug occurs when pipeline.upsert() is called with a TrainingStep whose SourceCode object has no requirements set (defaults to None). In get_code_hash() within sagemaker-core/src/sagemaker/core/workflow/utilities.py, source_code.requirements (which is None) is passed as the dependencies parameter to get_training_code_hash(). While the current get_training_code_hash() has an if dependencies: guard, the issue traceback suggests an older version had hash_files_or_dirs([source_dir] + dependencies) without that guard. Additionally, get_processing_code_hash() still has unguarded + dependencies concatenation that would fail with None. The fix should ensure both functions defensively handle None dependencies, and that the caller in get_code_hash() normalizes None to a safe default before passing it along.

AI-Generated PR

This PR was automatically generated by the PySDK Issue Agent.

• Confidence score: 92%
• Classification: bug
• SDK version target: V3

Merge Checklist

• Changes are backward compatible
• Commit message follows prefix: description format
• Unit tests added/updated
• Integration tests added (if applicable)
• Documentation updated (if applicable)

[aviruthen]

Can you remove only the comment you left on the PR explaining why you added the code dependencies = dependences or []?