[Henglie]

Closes #3207

Root cause

When an ELF has no section headers, pyelftools resolves dynamic symbol names through _DynamicStringTable.get_string, which decodes strictly as UTF-8. A single invalid byte raises UnicodeDecodeError, and capa's extractor did not catch it, so the whole run aborted.

Fix

Dynamic symbol reads now go through iter_dynamic_symbols, which catches UnicodeDecodeError from both num_symbols() and get_symbol() and skips unreadable entries instead of terminating. It yields the original symbol index alongside each symbol so import-to-symbol mapping remains correct after a skip. get_relocation_tables() is guarded the same way because DT_NEEDED-family tags use the same strict decode path.

Validation

• python -m pytest tests/test_elffile_features.py tests/test_elf_os_detection.py -q -> 173 passed
• New regression test_elffile_invalid_utf8_dynamic_symbol_name synthesizes a section-header-less ELF with a 0xff symbol name followed by a valid good symbol, and asserts extraction still completes and yields the format plus the readable export.
• The regression fails on current master with the reported UnicodeDecodeError and passes with this branch.

[x] No CHANGELOG update needed

[github-actions]

Please add bug fixes, new features, breaking changes and anything else you think is worthwhile mentioning to the master (unreleased) section of CHANGELOG.md. If no CHANGELOG update is needed add the following to the PR description: [x] No CHANGELOG update needed

CHANGELOG updated or no update needed, thanks! 😄

[williballenthin]

thanks @Henglie!