New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
Update haskellPackages #122286
Update haskellPackages #122286
Conversation
fail due to GHC bug
A script used at test time in HTF assumes that you have GNU coreutils installed in a manner where all executables are prefixed with a `g`. Since we use GNU coreutils in the stdenv anyways, this is not necessary and needs to be patched out.
haskell-ci pins a specific version of ShellCheck as ShellCheck adds new checks with new minor versions which may break CI checks.
Okay. So we see ~5200 that I will mark broken again. I am confused about llvm. Apparently that package exists as a haskellPackage but is broken as such? Anyways the not working build will be disabled again and llvmPackages.llvm will be alright. So I think there is nothing to do there. |
Yes, we reexpose the LLVM version used by GHC. We should probably filter this out in the build report, seems like it has been broken in master for some time. |
Okay. This is going great! We have gained about 700 more working packages. 🎉.
Until this point I am certain. I looked into some of them and I have the feeling the biggest reason for this are packages that have license This is not a huge problem either way. In the worst case I just mark them manually broken. Any opinions? |
To be honest the correct way forward is to patch cabal2nix to emit |
This seems like a completely reasonable thing to do. I guess it would be great if we could report them upstream (or add new licenses to Nixpkgs), but that is quite a lot of annoying work. |
I think it’s not that easy. I have looked at a few of those packages. (I count 496 with LICENSE unknown.) Unknown means that hackage displays the license as Also consider this: https://hackage.haskell.org/upload
So IANAL but I assume we are pretty safe with assuming that all packages on Hackage are somehow Open Source licensed. I am curious about other opinions but (until I here new arguments) I vote for enabling builds of those packages. I am totally happy with maintaining a deny-list of packages where we know that we don‘t have the right to distribute them. But since we are processing packages in an automated fashion I can‘t imagine that we will get in a lot of trouble if we assume that it’s not the case. |
I'm no lawyer, so I am unsure whether this actually enough assurance for us. Hackage does not actively verify that this is the case as far as I am aware.
I also would think that it is an unreasonable amount of work to figure out unclear and inconsistent license identifiers. In cases where we care about it, I think it is more productive to convince upstream to finally switch to SPDX-license identifiers which we can process properly and are not ambiguous. |
The current state has been like this for years. So it will probably fine if we keep it for a while longer. Considering that we have 2 weeks left until branch-off now is not the time to make rush decisions breaking packages. So I suggest to let this rest until after branch-off and then we find a more principled solution. I fear there is a consensus that we somehow need to default to regard this packages as unfree but should find an easy solution to override them with their correct license, once we have figured the license out. |
See linked issue.
… to Maybe Int The buildstatus may be `null` from Hydra if the Build hasn't finished running yet.
Co-authored-by: Dennis Gosnell <cdep.illabout@gmail.com>
This commit has been generated by maintainers/scripts/haskell/mark-broken.sh
This commit has been generated by maintainers/scripts/haskell/mark-broken.sh
@thoughtpolice I tried fixing tamarin-prover, which hadn‘t built since January. The problem is that tamarin-prover 1.6.0 is not compatible with maude 3.1.0 which we have in nixpkgs now. I can get it to build by disabling the checks, but it looks like the program won‘t work correctly then. So I opted to err on the save side and marked the package broken. Please feel free to choose a different solution at any time. |
This Merge
This is the regularly happening merge of the
haskell-updates
branch intomaster
.This branch is being continually build and tested by hydra at https://hydra.nixos.org/jobset/nixpkgs/haskell-updates.
I will aim to merge this PR until 2021-05-14. If I can merge it earlier, there might be successor PRs in that time window. As part of our rotation @cdepillabout will continue these merges from 2021-05-15 to 2021-05-28.
haskellPackages Workflow Summary
Our workflow is currently described at #121140 (comment).
The short version is this:
haskell-updates
(normally at the beginning of a merge window).haskell-updates
intomaster
every two weeks.mergeable
job is succeeding on hydra.This is the follow-up to #122117.
Scan for unbroken packages
This pull request is special in the sense that I haven‘t bumped stackage or hackage at the moment. I don‘t want anything confounding happening right now.
I have thrown away the
broken
list and now we will try to let hydra build all hackage packages once. When this evaluation is over I will mark all failing builds broken again.After that we can go back to business as usual.
During this time merging fixes for broken packages is okay but we should try not to touch anything that is deep down in the dependency chain so that we don‘t case an unnecessary huge number of hydra builds.