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

[aviruthen]

Description

Fix TypeError when using SourceCode without requirements in Pipeline

Fixes aws#5518

Problem

When creating a Pipeline with a TrainingStep that uses a SourceCode object without setting requirements, the pipeline build fails with:

TypeError: can only concatenate list (not 'NoneType') to list

This occurs because source_code.requirements is None when not explicitly set, and this None value gets passed as the dependencies parameter to get_training_code_hash(), which then attempts to use it in a list concatenation.

Solution

• In get_code_hash(): Normalize source_code.requirements using or None to ensure any falsy value (empty string, empty list, None) is consistently treated as None
• In get_training_code_hash(): Add a defensive dependencies = dependencies or None guard at the top of the function body to handle any falsy value passed by callers, ensuring the existing if dependencies: guards work correctly regardless of the input type

Testing

Added unit tests to verify:

• get_training_code_hash() works correctly when dependencies is None
• get_training_code_hash() works correctly when dependencies is an empty string
• get_training_code_hash() works correctly with both source_dir and entry_point paths when dependencies are not provided

Related Issue

Fixes aws#5518

Changes Made

The bug occurs in get_code_hash() in sagemaker-core/src/sagemaker/core/workflow/utilities.py. When a TrainingStep uses a SourceCode object without setting requirements, source_code.requirements is None. This None value is passed as the dependencies parameter to get_training_code_hash(). While the current code in the repo has if dependencies: guards before concatenating, the user's version (3.3.1) hits [source_dir] + dependencies where dependencies is None, causing TypeError: can only concatenate list (not 'NoneType') to list. The fix should add a defensive or None coercion in get_code_hash() when extracting requirements from source_code, and also add an explicit None/empty-list guard at the top of get_training_code_hash() to make it robust regardless of caller behavior.

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]

Nice work! Can you omit the change to requirements (source_code.requirements or None)? Our team is okay not having this change