Handling of CR as EOL in trailing-whitespace

AFulgens commented 5 months ago

Currently the trailing-whitespace hook (trailing_whitespace_fixer.py) handles the end-of-line (EOL) combinations of:

Windows flavor (CRLF)
Linux, modern MacOS flavor (LF)

However, files that use CR as EOL are handled incorrectly by the hook, and in combanation with EOF-Fixer (end_of_file_fixer.py) this makes it impossible to commit the file if both hooks are active, because:

EOF fixer adds a CR as last byte of the file
trailing-whitespace fixer removes the last CR

(Or the other way around, depending on order of the hooks.)

Current workarounds:

Change the EOL of the file (but what if I need it to have CR for testing some crazy edge-case?)
Ignore the hooks when committing this file (meh)

Question: Would it be OK to handle \r the same way \n and \r\n are handled in trailing_whitespace_fixer.py? If yes, I'll gladly prepare a PR.

asottile commented 5 months ago

it's been decades since anything used cr as eol. it is unsupported intentionally

AFulgens commented 4 months ago

I guess, we are the only team in the world then working on legacy software, which is generating XMLs with CR as EOL and we must parse it in modern codebase, and thus have tests for it. Nevertheless, I appreciate your answer, onto the exclude list with those test XMLs for this hook! 🏴‍☠️

asottile commented 4 months ago

if you have generated files you probably don't want to lint them anyway

pre-commit / pre-commit-hooks

Handling of CR as EOL in trailing-whitespace #1066