spack: spack fails to include default variants as dependencies
Hi all,
There is an issue with spack where if I input the command spack install dealii ^trilinos@12.8.1 the command fails by saying Dealii does not depend on trilinos. However, Deal.II is compiled with Trilinos as one of the dependencies which are turned on by default.
On the other hand, if I point explicitly that trilinos should be enabled spack install dealii+mpi+trilinos ^trilinos@12.8.1, I don’t get the error anymore.
Is this a bug or a feature? 😃
About this issue
- Original URL
- State: closed
- Created 10 years ago
- Comments: 18 (18 by maintainers)
Instead of documenting things that are missing from the documentation, and creating elaborate posts containing the information that should be documented… why don’t we just add it to the documentation and close the issue? Yes, some missing documentation is contributor-knowledge specific (I’m looking at you @alalazo 😉) but a lot of these are pretty simple…
Can we add a “known issues” page to the docs to reduce the number of open issues? We have some things like this in the getting started guide, but not in easily referenced form like I’ve seen other projects do.
It could even reference original issues on github, and then when we close duplicates, we can reference the original as @adamjstewart did so that there’s a link to the new one.
Yes, this bug has been around since the early days of Spack. I believe #397 is the oldest known reference I’ve seen. @tgamblin has a lot of work planned for the concretizer, and I believe this is one of the issues we would all like addressed, just for the sake of avoiding confusion.
@citibeth:
I suspect, though I could be wrong, that a known issues page would incite considerably less controversy than the contributor guide. Does anyone disagree about which issues are known? I think we’re all pretty open about that. It’s much more clear cut than “what should every PR submitter have to do”.
The timescale we’re talking about here is long. I am not so concerned about the merge time because this is specifically for things we may not fix for a while. So the merge overhead seems small to me. If it’s something that’s easier to just fix we should do that (per @adamjstewart)
I don’t like the documentation requests issue, because we already have the
documentationlabel. You can just click on it to see all the open documentation requests. I would rather have a distributed labeling system than a system where we have to remember to edit and append to #2405.The idea was to get the known issues into the docs, so that new users can see them easily. I think most people will not want to wade through github issues.
I would definitely be in favor of linking to original issues and closing duplicates. I always cringe at spending time documenting bugs vs. just fixing them, but for complicated bugs like this that may take a while.
OK my mistake closing… I re-opened and added the “documentation” tag. Unless/until we fix, we should talk about this “gotcha” in the docs.