hGetContent says invalid byte sequence on Windows, non-English locale

[ice1000]

Before doing this readFile:

https://github.com/simonmar/happy/blob/27596ff0ce0171d485bf96d38943ffc760923c90/src/Main.lhs#L72-L74

we may do hSetEncoding h IO.utf8 before.

See agda/agda#4161 (comment)

[andreasabel]

You could submit a PR, it is probably easier for you to test the change as you have the right context (Windows, non-English locale).

[lehins]

This applies to non-Windows systems as well. If your locale is not set or set to something like LANG=ascii, reading files with unicode in them will result in this error. In fact we received this error on Linux machine as well with this happy file: https://github.com/erikd/language-javascript/blob/eef1887d430c18b108ff723479c3f1ef50c0e9b2/src/Language/JavaScript/Parser/Grammar7.y

I fixed an issue exactly like this one with hpc: https://gitlab.haskell.org/ghc/ghc/issues/17073

Same fix could be applied here. Haskell source files are always assumed to be encoded in utf-8, same principal could be applied to happy .y files.

[hdgarrood]

This has caused problems for people trying to build the purescript compiler from source too: eg purescript/purescript#3813, erikd/language-javascript#86. I think having happy always assume that .y files are UTF-8 encoded would indeed be a good option.

[thatwist]

struggled with this.. this combo worked for me

echo "LC_CTYPE=\"en_US.UTF-8\"" | sudo tee -a /etc/default/locale
echo "LC_ALL=\"en_US.UTF-8\"" | sudo tee -a /etc/default/locale
echo "LANG=\"en_US.UTF-8\"" | sudo tee -a /etc/default/locale
echo "LC_ALL=en_US.UTF-8" | sudo tee -a /etc/environment
echo "en_US.UTF-8 UTF-8" | sudo tee -a /etc/locale.gen
echo "LANG=en_US.UTF-8" | sudo tee -a /etc/locale.conf
sudo locale-gen en_US.UTF-8