spack: Installation issue: neovim doesn't concretize

Steps to reproduce the issue

$ spack spec -I neovim
==> Error:

Error message

Error message
The error message is empty as noted in the output of spack spec above

Information on your system

  • Spack: 0.22.0.dev0 (9e0720207afb5b0fed4035dc1dbf214a1eddfca1)
  • Python: 3.9.15
  • Platform: linux-ubuntu20.04-skylake
  • Concretizer: clingo

Additional information

Since the spec doesn’t even concretize, no build files were generated. I have confirmed this fails on releases/v0.21, but succeeds on releases/v0.20 (i.e. this spec stops concretizing after 0.20). @albestro @trws I am mentioning you as per the instructions in the issue template, but I’m actually unsure whether this is a bug in the neovim package or the concretizer.

General information

  • I have run spack debug report and reported the version of Spack/Python/Platform
  • I have run spack maintainers <name-of-the-package> and @mentioned any maintainers
  • I have uploaded the build log and environment files
  • I have searched the issues of this repo and believe this is not a duplicate

About this issue

  • Original URL
  • State: closed
  • Created 3 years ago
  • Reactions: 1
  • Comments: 16 (14 by maintainers)

Most upvoted comments

@albestro 👍 https://github.com/spack/spack/pull/42401 looks good to me.

I was really looking for a better error message from the concertizer, because it looked like this double-virtual ambiguity could happen somewhere else later. But better constraints in recipes is nice too for now.

If I intentionally look for the error, I can still get it using spack spec --fresh "neovim ^lua". But maybe the error is legit here (?), and there is a missing rule in the concretizer to properly report this ambiguity…

@Jordan474 With neovim ^ lua you end up with exactly the same problem that was previously internally triggered by the Lua build system in spack. Indeed, you’re not asking that lua to be the provider, but just that lua has to appear in the DAG.

It is probably under-constrained, but AFAIK there is nothing saying that it should be a conflict (even if it probably should be?!).

I understand it might not be expected, but I don’t have a better explanation for that. @alalazo surely has an answer to this too. What I can say is that I would express it as neovim ^[virtuals=lua-lang] lua.

Maybe we need a require("+no_luajit", "^lua") require("~no_luajit", "^luajit") or similar? I honestly wish we could make that go away and always require luajit, but there are a couple of platforms that still have issues with that occasionally.

@trws I have a couple of changes in the pipeline for neovim and related packages, where I’m going to be more specific about the fact that we are referring to the provider and not to a generic package. This should help the concretizer doing the right thing.