Add option for stdio encoding, set stdin encoding

[lierdakil]

Relevant issue: atom-haskell-archive/haskell-ghc-mod#105

[lierdakil]

This is a pretty straightforward change. Only point of discussion I can see is whether optparse-applicative messages should use system encoding (in current version), or respect one set with --encoding. I'm betting on former, since those messages don't make much sense outside of console. Besides, changing stdout encoding before optparse-applicative does its thing would be reasonably convoluted.

@DanielG, I'd rather merge this sooner rather than later, if you don't mind, since it can cause interactive interface breakage on Windows machines if stdin encoding isn't set explicitly.

P.S. Also, might be a good idea to coach HLint into using encoding specified here, but I'll leave that for another PR, since it's not directly related.

[DanielG]

:)

[lierdakil]

Yes, that's been there for a while now. I thought that this is as good a time as any to remove it.

[DanielG]

Looks perfectly fine to me feel free to merge it. Do you need this released soon too?

[lierdakil]

If you plan on doing a minor release after all, I'd prefer it included there, I think. I can rebase on stable-5.5 and make a separate PR for that.

[DanielG]

Looks like this broke something on my end:

ghc-mod: <somewhere>/cabal-sandbox/x86_64-linux-ghc-7.10.3-packages.conf.d/package.cache: GHC.PackageDb.readPackageDb: inappropriate type (Not a valid Unicode code point!)

Probably shouldn't be forcing the encoding but rather only setting it when that's requested on the cmdline.

[DanielG]

See #758

[lierdakil]

This is weird. I only set stdin/stderr encoding (stdout was utf-8 anyway),
so unless readPackageDb gets its information from stdin, this shouldn't be
an issue, and I don't think it does? I think we're better off getting to
the bottom of this issue rather than reverting to using system encoding
unless specified. This will come up sooner or later if we don't.

See #758 #758

—
Reply to this email directly or view it on GitHub
#747 (comment).

[DanielG]

Maybe readProcess bases the encoding it uses on that of stdin/out or something like that? I did think that was weird too.

[DanielG]

Nevermind there is no readProcess in that code path GHC reads that directly it looks like.

[DanielG]

And of course I can't reproduce that now after rebuilding without the patch -.-