Project whose path contains non-ASCII characters fails to build on Linux

[mhwombat]

General summary/comments (optional)

In latest Stack version, a project won't build if its pathname contains non-ASCII characters.

Steps to reproduce

1. stack new dúmmy # Note the accent on the ú
2. cd dúmmy/
3. stack build

Expected

I expect it to build

Actual

The build fails.

If you do the same steps, but using the project name "dummy" (no accent), the build succeeds.

$ stack build
Building all executables for `dúmmy' once. After a successful build of all of them, only specified executables will be rebuilt.
dúmmy-0.1.0.0: configure (lib + exe)
<stdout>: commitBuffer: invalid argument (invalid character)
Configuring d

--  While building package d\250mmy-0.1.0.0 using:
      /home/amy/.stack/setup-exe-cache/x86_64-linux-nix/Cabal-simple_mPHDZzAJ_2.2.0.1_ghc-8.4.4 --builddir=.stack-work/dist/x86_64-linux-nix/Cabal-2.2.0.1 configure --with-ghc=/nix/store/vrdzir6ql35x1an44ln0vir3a3pra75f-ghc-8.4.4/bin/ghc --with-ghc-pkg=/nix/store/vrdzir6ql35x1an44ln0vir3a3pra75f-ghc-8.4.4/bin/ghc-pkg --user --package-db=clear --package-db=global --package-db=/home/amy/.stack/snapshots/x86_64-linux-nix/lts-12.22/8.4.4/pkgdb --package-db=/home/amy/dúmmy/.stack-work/install/x86_64-linux-nix/lts-12.22/8.4.4/pkgdb --libdir=/home/amy/dúmmy/.stack-work/install/x86_64-linux-nix/lts-12.22/8.4.4/lib --bindir=/home/amy/dúmmy/.stack-work/install/x86_64-linux-nix/lts-12.22/8.4.4/bin --datadir=/home/amy/dúmmy/.stack-work/install/x86_64-linux-nix/lts-12.22/8.4.4/share --libexecdir=/home/amy/dúmmy/.stack-work/install/x86_64-linux-nix/lts-12.22/8.4.4/libexec --sysconfdir=/home/amy/dúmmy/.stack-work/install/x86_64-linux-nix/lts-12.22/8.4.4/etc --docdir=/home/amy/dúmmy/.stack-work/install/x86_64-linux-nix/lts-12.22/8.4.4/doc/dúmmy-0.1.0.0 --htmldir=/home/amy/dúmmy/.stack-work/install/x86_64-linux-nix/lts-12.22/8.4.4/doc/dúmmy-0.1.0.0 --haddockdir=/home/amy/dúmmy/.stack-work/install/x86_64-linux-nix/lts-12.22/8.4.4/doc/dúmmy-0.1.0.0 --dependency=base=base-4.11.1.0 --extra-include-dirs=/nix/store/3zal9zfqb41fsbmwl5ba7sxh3h2ngvxi-git-2.18.1/include --extra-include-dirs=/nix/store/ipba8r4418ls0j55p5y5gi4zmg2kna5n-gmp-6.1.2-dev/include --extra-include-dirs=/nix/store/iw94llkj05wgaz268mlzvgx8jkbi1ss0-gcc-wrapper-7.3.0/include --extra-include-dirs=/nix/store/vrdzir6ql35x1an44ln0vir3a3pra75f-ghc-8.4.4/include --extra-lib-dirs=/nix/store/3zal9zfqb41fsbmwl5ba7sxh3h2ngvxi-git-2.18.1/lib --extra-lib-dirs=/nix/store/iw94llkj05wgaz268mlzvgx8jkbi1ss0-gcc-wrapper-7.3.0/lib --extra-lib-dirs=/nix/store/rxaclc3vhxbzgfllv7601i85z854h48a-gmp-6.1.2/lib --extra-lib-dirs=/nix/store/vrdzir6ql35x1an44ln0vir3a3pra75f-ghc-8.4.4/lib --enable-tests --enable-benchmarks
    Process exited with code: ExitFailure 1

Output in verbose mode is shown in gist.

Stack version

$ stack --version
Version 1.9.3, Git revision 40cf7b37526b86d1676da82167ea8758a854953b (6211 commits) x86_64 hpack-0.31.1

Method of installation

This is on a NixOS, installed using NixOS package repo.

[mhwombat]

Forgot to add:

$ locale
LANG=en_IE.UTF-8
LC_CTYPE="en_IE.UTF-8"
LC_NUMERIC="en_IE.UTF-8"
LC_TIME="en_IE.UTF-8"
LC_COLLATE="en_IE.UTF-8"
LC_MONETARY="en_IE.UTF-8"
LC_MESSAGES="en_IE.UTF-8"
LC_PAPER="en_IE.UTF-8"
LC_NAME="en_IE.UTF-8"
LC_ADDRESS="en_IE.UTF-8"
LC_TELEPHONE="en_IE.UTF-8"
LC_MEASUREMENT="en_IE.UTF-8"
LC_IDENTIFICATION="en_IE.UTF-8"
LC_ALL=

[qrilka]

BTW @mhwombat did you try impure Nix mode in Stack with --no-nix-pure? By default nix mode is run in pure nix shell and that makes your locale listing inapplicable though as of #4294 en_US.UTF-8 should be used as a default.
Also running Stack in verbose mode could help a lot.

[mhwombat]

That gist I linked to includes the output when running in verbose mode.

I forgot to say that I tried --no-nix-pure, and got a similar message. Here's that gist with --no-nix-pure.

[mhwombat]

I was able to duplicate the problem on Debian. Here's the gist.

EDIT: Just to clarify, I don't have Nix installed on the Debian machine, so I don't think this problem is Nix-specific.

[dbaynard]

@mhwombat Thanks for the report. I've just had a flick through the related issues; would you please check whether this is still an issue for you with stack nightly?

Meanwhile, I'll try to repro with an integration test based on the one in #4294 (unless you want to do that?).

[mhwombat]

I'll try both of those and let you know.

Incidentally, this was working fine a few weeks ago. I have a hunch that it might be a cabal issue rather than a Stack issue. I'll try to test that theory as well.

[mhwombat]

It seems my hunch was correct; it's not a Stack issue per se. What's happening here is that cabal build automatically generates a module called Paths__projectname_, and that's what causes the problem.

$ cabal build
Config file path source is default config file.
Config file /home/amy/.cabal/config not found.
Writing default configuration to /home/amy/.cabal/config
Warning: The package list for 'hackage.haskell.org' does not exist. Run 'cabal
update' to download it.
Resolving dependencies...
Configuring dúmmy-0.1.0.0...
Preprocessing library for dúmmy-0.1.0.0..
Building library for dúmmy-0.1.0.0..
ghc: fd:23: hGetContents: invalid argument (invalid byte sequence)

In file included from <command-line>:7:0: error: 

dist/build/autogen/Paths_dúmmy.hs:4:15: error:
    lexical error (UTF-8 decoding error)
  |
4 | module Paths_d�mmy (
  |

[dbaynard]

@mhwombat Thanks for investigating — and please do comment here if you find a suitable workaround.

[qrilka]

BTW @mhwombat did you open a Cabal issue about this?

[mhwombat]

As it turns out, there is already a Cabal issue about it.