[dajiaohuang]

Names printed by --list-hashes, such as RIPEMD-160, are not recognized as template directives and are emitted literally. Match hash names case-insensitively after removing dashes from both the input and hash-table name. Initialize the normalization buffer so legacy-directive comparisons cannot inspect unset bytes, and cast input bytes to unsigned char before case conversion.

Closes #311.

Validation: compiled the actual shared-library CLI on Ubuntu with GCC 13.3 and SHA-NI disabled. The added executable-level regression verifies all 32 public hash names against their dashless aliases, SHA-256 of abc, and existing urlname/mtime directives. The original implementation fails at RIPEMD-160, returning the directive literally; the fix passes. All 26 groups of the full existing CLI suite pass, and git diff --check passes. Run the new standalone regression with python3 tests/test_template_hash_names.py ./rhash; it is not automatically wired into the existing shell suite. Other platforms and SHA-NI-enabled builds were not exercised for this PR.

[rhash]

Commit looks good. The change requires testing.

[dajiaohuang]

The requested tests are complete. The Linux/GCC 13.3 shared build passes a compiled-CLI regression for 32 canonical hash names and their dashless aliases, a known SHA-256 value, and the legacy urlname/mtime directives. The original head fails the RIPEMD-160 canonical name case; this head passes. All 26 full CLI test groups pass with SHA-NI disabled. The CI jobs for Ubuntu, PowerPC 64 and Windows also pass at this head.

[rhash]

The fix is rewritten as PR #329

[rhash]

Accepted with changes as 7b2801d.