Void-packages: No setup information found for GHC

[alexjacobson95]

System

• xuname:
  Void 4.19.15_1 x86_64 GenuineIntel uptodate rmFFFFF
• package:
  stack-1.9.3_1

Expected behavior

When running stack setup, I expect it to download ghc and other haskel tools.

Actual behavior

$ stack setup
No setup information found for ghc-8.6.3 on your platform.
This probably means a GHC bindist has not yet been added for OS key 'linux64-ncurses6'.
Supported versions: ghc-7.10.3, ghc-8.0.1, ghc-8.0.2, ghc-8.2.1, ghc-8.2.2

Steps to reproduce the behavior

stack new test-project
cd test-project
stack setup

[alexjacobson95]

This also seems related:
https://github.com/commercialhaskell/stack/issues/4034

[rektrex]

I too have the same problem. I have tried the potential solutions, including the correct solution in the link in the previous comment.

[lambdadog]

Also running into this issue

[lambdadog]

Linking /usr/lib/libncurses.so.6.1 to /usr/lib/libtinfo.so.6 seems to have fixed it for me

reading more, seems to possibly be an issue with ncurses packaging since tinfo (AFAICT) seems to be just the low-level interface to ncurses so the symlink is safe and valid?

EDIT: Not necessarily an ncurses packaging issue, although the issue and those like it could be fixed by adding a compile flag to ncurses it looks like. So, while it's strictly a GHC issue for not providing a version linked to ncurses6 alongside tinfo6, it's a complicated problem

unsure of why providing a release linked to tinfo6 over ncurses6 was chosen however, since ncurses6 should be more portable i believe

more info here: https://bugs.launchpad.net/ubuntu/+source/ncurses/+bug/259139

[vinnyA3]

Thanks @beta-phenylethylamine for the helpful link.
In addition to linking /usr/lib/libncurses.so.6.1 to /usr/lib/libtinfo.so.6, I also had to install gmp-devel to get ghc to work properly.

[CameronNemo]

@vinnyA3 sounds like a separate issue. please file a report.

[CameronNemo]

@beta-phenylethylamine is there a ticket with haskell / ghc sources?

[lambdadog]

@CameronNemo sorry, not that I'm aware of. I'm currently poking around in different sources here and there to see if I can figure out whether the fault lies in stack or ghc itself. Will update when I sort it out

[lambdadog]

Seems to be at the stackage level, in regards to generating bindists?

The response is "most distros have standardised on either providing or symlinking libtinfo.so.6, so there's no bindists that link to ncurses6", so it seems the only solution for stackage compatibility is to change our ncurses packaging as suggested earlier, or somehow convince them to build generic ncurses6 bindists.

https://github.com/commercialhaskell/stack/issues/4711

[egormalyutin]

Any progress on this?

[egormalyutin]

If someone will need this: you can try to run stack build --system-ghc, on my machine it works

[mihaimaruseac]

That workaround keeps you using the GHC that void linux provides, but that might not be the newest one

[fuszenecker]

Same happens on Raspberry Pi 3:

fuszenecker@raspberrypi:~ $ stack build No setup information found for ghc-8.6.5 on your platform. This probably means a GHC bindist has not yet been added for OS key 'linux-armv7', 'linux-armv7-ncurses6', 'linux-armv7-tinfo6'. Supported versions: ghc-7.10.2, ghc-7.10.3, ghc-8.0.1, ghc-8.0.2, ghc-8.2.1, ghc-8.2.2, ghInstalling GHC ...

[lambdadog]

@fuszenecker This is a separate issue of Stack not having a GHC 8.6.5 bindist for armv7, I believe. Stack seems to be detecting that you do indeed have libtinfo as indicated by linux-armv7-tinfo6 being listed in the OS keys being searched for.

[MarleDK]

This should be fixed by installing ncurses-libtinfo-libs now, I will also look into making it a prerequisite for stack. If it does not break something else.
See #21409