Fails to load when cabal file contains non-ASCII characters

[chris-martin]

I have an example here of a project that crashes haskell-language-server because the cabal file contains an em dash. If I remove the offending character, then HLS works.

Couldn't figure out what GHC version the project is using: /home/chris/.config/Code/User/globalStorage/haskell.haskell/haskell-language-server-wrapper-1.2.0-linux --project-ghc-version exited with exit code 1: No 'hie.yaml' found. Try to discover the project type! haskell-language-server-wrapper-1.2.0-linux: /home/chris/environment-variables/environment-variables.cabal: hGetContents: invalid argument (invalid byte sequence)

• OS: NixOS 21.05
• lsp-client: VSCode
• no hie.yaml file

[jneira]

Thanks you for reporting the issue. Until the bug is fixed a possible workaround could be generate an explicit hie.yaml file using the utility gen-hie (you can get it with cabal install implicit-hie)
Maybe you will have to remove the unicode char to make it work, but once the file is generated you can add it again.

[chris-martin]

Thanks for the workaround, that does work.

[jneira]

Hmm, i cant reproduce with hls 1.1 in windows adding this line to my cabal file: description: With unicode char → —
Will try with 1.2

EDIT: hls-1.2.0 works for me too

D:\ws\haskell\cabal-test>haskell-language-server-wrapper
No 'hie.yaml' found. Try to discover the project type!
Run entered for haskell-language-server-wrapper(haskell-language-server-wrapper)
 Version 1.2.0.0, Git revision be2071e985cb417f984ab00a1aad76dee02d6d0b (dirty)
x86_64 ghc-8.10.4
Current directory: D:\ws\haskell\cabal-test
Operating system: mingw32
Arguments: []
Cradle directory: D:\ws\haskell\cabal-test
Cradle type: Cabal

Tool versions found on the $PATH
cabal:          3.4.0.0
stack:          2.6.0
ghc:            8.8.4


Consulting the cradle to get project GHC version...
Project GHC version: 8.8.4
haskell-language-server exe candidates: ["haskell-language-server-8.8.4.exe","haskell-language-server.exe"]
Launching haskell-language-server exe at:D:\bin\hls\haskell-language-server-8.8.4.exe
haskell-language-server version: 1.2.0.0 (GHC: 8.8.4) (PATH: D:\bin\hls\haskell-language-server-8.8.4.exe)
 ghcide setup tester in D:\ws\haskell\cabal-test.
Report bugs at https://github.com/haskell/haskell-language-server/issues

Step 1/4: Finding files to test in D:\ws\haskell\cabal-test
Found 6 files

Step 2/4: Looking for hie.yaml files that control setup
Found 1 cradle
  ()

Step 3/4: Initializing the IDE

Step 4/4: Type checking the files
............
Files that failed:
 * D:\ws\haskell\cabal-test\src\MyPrefix\MyModuleExe.hs

Completed (5 files worked, 1 file failed)
2021-06-17 11:05:51.066693 [ThreadId 178] INFO hls:     finish: GenerateCore (took 0.00s)
haskell-language-server-wrapper: callProcess: D:\bin\hls\haskell-language-server-8.8.4.exe (exit 1): failed

[ncfavier]

Can confirm having the same issue with the VSCode extension on NixOS.

[ncfavier]

Solved by installing nixpkgs' haskell-language-server instead of using the one managed by the extension.

[jneira]

It seems this is not reproduced with newer hls versions, feel free to reopen ion other case

[gusbicalho]

FYI, I just ran into this bug with hls 1.7.0. The log below shows the cabal file as text and hex dump.

It turned out that the problem was that the VM I was using had locales set to POSIX. Once I fixed things so that it got a UTF-8 locale, HLS worked again. I imagine this means hGetContents uses the system locale to pick an encoding, and if the locale is not UTF-8 it breaks down on multi-byte chars (?)

$ ls -la
total 28
drwxr-xr-x  2 gusbicalho gusbicalho  4096 Apr 29 19:16 .
drwxrwxrwt 69 root       root       20480 Apr 29 19:15 ..
-rw-r--r--  1 gusbicalho gusbicalho   157 Apr 29 19:15 repro-utf8.cabal

$ cat repro-utf8.cabal
cabal-version: 2.0
name: repro-utf8
build-type: Simple
version: 0.1.0.0
author: lön

library
    default-language: Haskell98
    build-depends: base == 4.*

$ od -t cx1 repro-utf8.cabal
0000000   c   a   b   a   l   -   v   e   r   s   i   o   n   :       2
         63  61  62  61  6c  2d  76  65  72  73  69  6f  6e  3a  20  32
0000020   .   0  \n   n   a   m   e   :       r   e   p   r   o   -   u
         2e  30  0a  6e  61  6d  65  3a  20  72  65  70  72  6f  2d  75
0000040   t   f   8  \n   b   u   i   l   d   -   t   y   p   e   :
         74  66  38  0a  62  75  69  6c  64  2d  74  79  70  65  3a  20
0000060   S   i   m   p   l   e  \n   v   e   r   s   i   o   n   :
         53  69  6d  70  6c  65  0a  76  65  72  73  69  6f  6e  3a  20
0000100   0   .   1   .   0   .   0  \n   a   u   t   h   o   r   :
         30  2e  31  2e  30  2e  30  0a  61  75  74  68  6f  72  3a  20
0000120   l 303 266   n  \n  \n   l   i   b   r   a   r   y  \n
         6c  c3  b6  6e  0a  0a  6c  69  62  72  61  72  79  0a  20  20
0000140           d   e   f   a   u   l   t   -   l   a   n   g   u   a
         20  20  64  65  66  61  75  6c  74  2d  6c  61  6e  67  75  61
0000160   g   e   :       H   a   s   k   e   l   l   9   8  \n
         67  65  3a  20  48  61  73  6b  65  6c  6c  39  38  0a  20  20
0000200           b   u   i   l   d   -   d   e   p   e   n   d   s   :
         20  20  62  75  69  6c  64  2d  64  65  70  65  6e  64  73  3a
0000220       b   a   s   e       =   =       4   .   *  \n
         20  62  61  73  65  20  3d  3d  20  34  2e  2a  0a
0000235

$ cabal build
Resolving dependencies...
Build profile: -w ghc-9.2.2 -O1
In order, the following will be built (use -v for more details):
 - repro-utf8-0.1.0.0 (lib) (first run)
Configuring library for repro-utf8-0.1.0.0..
Preprocessing library for repro-utf8-0.1.0.0..
Building library for repro-utf8-0.1.0.0..

$ haskell-language-server-wrapper-1.7.0.0
No 'hie.yaml' found. Try to discover the project type!
haskell-language-server-wrapper-1.7.0.0: /tmp/repro-utf8/repro-utf8.cabal: hGetContents: invalid argument (invalid byte sequence)

[isomorpheme]

I also seem to be having the same issue, having built 1.7.0.0. from source.

[istathar]

I'm hitting this. haskell-language-server 1.8.0.0's haskell-language-server works fine from command line but when run from within VSCode it results in

thing.cabal: hGetContents: invalid argument (invalid byte sequence)

It's possible that the problem traces to the builtin gen-hie code supplied by implict-hie (and possibly further to Cabal-syntax). It is POSSIBLE that an explicit hie.yaml file containing, in my case nothing more than

cradle:
  stack:

might act as a workaround suppressing the implicit hie code path. It's working? but if so, fragile. Unclear whether something Nix is doing might be triggering the localization issue; code developed on the underlying Fedora system works (or was working) fine.