Summary

• Replaces the blanket strip of \?\ and //?/ byte sequences from all text data in text_replace() with targeted replacement that only matches extended-length prefixes immediately followed by the placeholder path
• Preserves legitimate source code referencing \?\ (e.g., huggingface_hub/file_download.py) while still correctly handling dangling extended-length prefixes left after core.py normalization
• Updates existing tests and adds regression tests for the corruption scenario

Root Cause

PR #432 (released in 0.9.0) fixed #398 by stripping \?\ and //?/ from all text file data on Windows to clean up dangling extended-length path prefixes:

# Previous approach - strips all occurrences unconditionally
data = data.replace(b'\\?\', b'')
data = data.replace(b'//?/', b'')

While this successfully addressed the original issue, it also inadvertently affects any .py file containing those byte sequences as legitimate source code (e.g., "\\?\\" in Python string literals), causing SyntaxError at import time.

Fix

Replace with targeted replacement that only strips extended-length prefixes when immediately followed by the placeholder:

# New approach - only matches prefix + placeholder
data = data.replace(b'\\?\' + placeholder_bytes, new_prefix_bytes)
data = data.replace(b'//?/' + placeholder_bytes, new_prefix_bytes)

This handles all three path cases without touching unrelated source code:

• \?\C:\envs\myenv -> replaced cleanly (extended-prefix + placeholder as unit)
• //?/C:\envs\myenv -> replaced cleanly (extended-prefix + placeholder as unit)
• C:\envs\myenv -> standard replacement
• Source code with \?\ (not adjacent to placeholder) -> preserved

Testing

• Updated test_windows_extended_length_path_cleanup to verify targeted replacement behavior
• Updated test_windows_extended_length_path_with_placeholder_replacement for targeted behavior
• Updated test_windows_multiple_extended_length_patterns to verify preservation of non-placeholder patterns
• Added test_windows_source_code_not_corrupted: regression test for issue conda-pack 0.9.0+ silently corrupts Python source files containing \\?\ on Windows #461 (huggingface_hub case)
• Added test_windows_targeted_extended_prefix_replacement: verifies both \?\ and //?/ prefix styles
• Added test_windows_mixed_extended_and_standard_paths: tests mixed extended and standard paths
• Verified all existing TestUpdatePrefix tests are unchanged and should still pass
• Verified issue Env with conda dist included results in conda-hook.ps1 content with extended path format on Windows? #398 fix is preserved (core.py normalization + update_prefix .ps1 handling untouched)

Fixes #461