| 31 Dec 2024 |
 sterni | staging-next-24.11 probably buys us another week before staging-next starts | 23:49:04 |
| 1 Jan 2025 |
 linj | To link a nixos test to a haskell package, I should edit pkgs/development/haskell-modules/configuration-nix.nix, right? | 01:40:28 |
 maralorn | In reply to @sternenseemann:systemli.org
staging-next-24.11 probably buys us another week before staging-next starts Meaning we could try to get another merge in this staging cycle? | 01:43:39 |
| sterni | yes e.g. with jailbreak-cabal for GHC 9.12 | 02:53:25 |
| sterni | to add to passthru.tests? yes | 02:53:37 |
 hellwolf | thank you, team.
and happy new year.
I will upgrade to 912 soon!! | 10:37:00 |
|  NixOS Moderation Botchanged room power levels. | 14:26:19 |
| linj | sterni: Do you think adding kmonad to the mergeable job is a good idea? | 18:08:00 |
| sterni | linj: isn't necessary, if the package has a maintainer it will already be added to maintained and the nixosTest will show up in reports of failing packages automatically (hydra-report.hs) | 18:39:39 |
| sterni | mergeable is for packages that may never be broken which kmonad isn't really | 18:40:05 |
| maralorn | Cabal started throwing errors when there is not local hackage db. Is there a way I can disable that it wants that? | 20:19:08 |
| sterni | https://github.com/MercuryTechnologies/cabal/commit/ec242b1cb21e450d5212efd5c445f0d70786092e | 23:00:58 |
| sterni | i.e. upgrade 3.14.1.1 | 23:01:39 |
| sterni | * i.e. upgrade to 3.14.1.1 not shipping in any nixpkgs near you | 23:03:46 |
| sterni | c.f. https://github.com/haskell/cabal/issues/10504 | 23:06:52 |
 emily | looks like that was merged upstream, https://github.com/haskell/cabal/commit/6bfdbfd33a5f4e1f2b1b614c9b40372c1f07c88f? | 23:33:58 |
| emily | so patch just needs backporting? | 23:34:08 |
|  scm left the room. | 23:53:34 |
| 2 Jan 2025 |
 jean-paul. | So enableSeparateBinOutput and justStaticExecutables have essentially the same effect on the closure size (in the former case, only for the bin output). I take this to mean they are making the same changes to how the final executable is linked?
It's a bit surprising that enableSeparateBinOutput changes the way the executable is linked, rather than merely creating a new derivation output that contains the executables, isn't it?
| 13:39:05 |
| sterni | jean-paul.: executables are always statically linked against their Haskell dependencies even without either override | 13:44:46 |
| sterni | justStaticExecutables means “just the static executables already included in the given package” | 13:45:20 |
| jean-paul. | Huh! Okay, I have misunderstood where the closure size reduction from those two flags comes from then. | 13:45:29 |
| sterni | if the package includes documentation and/or libraries those and/or their meta data files will reference all of the dependencies | 13:46:33 |
| jean-paul. | Ah | 13:46:59 |
| sterni | shared libraries are built so you have the option of linking dynamically. Also for static libs you need to know all of its dependencies in order to successfully link the resulting program | 13:47:47 |
| sterni | the closure size reduction for enableSeparateBinOutput stems from the fact that these things are shoved outside of bin and in justStaticExecutables they are deleted. | 13:48:34 |
| jean-paul. | Ah indeed, I see now. There are .so files in the nix lib output directory and those are dynamically linked against .so files from many other Haskell packages. | 13:48:37 |
| sterni | https://hackage.haskell.org/package/jailbreak-cabal-1.4.1 | 16:12:21 |
 Morgan (@numinit) | sterni - when you have a second, can you give https://github.com/NixOS/nixpkgs/pull/327052 another review? | 19:57:05 |
| 3 Jan 2025 |
| jean-paul. | https://haskell-haddock.readthedocs.io/latest/invoking.html#avoiding-recompilation says Haddock can build docs without recompiling the Haskell. Reading build logs of a nixpkgs Haskell package with Haddocks enabled, it doesn't look like nixpkgs takes advantage of this. During the haddockPhase I see, eg:
[ 1 of 103] Compiling Converge.Data ( src/Converge/Data.hs, nothing )
[ 2 of 103] Compiling Converge.Domain.NaturalKey ( src/Converge/Domain/NaturalKey.hs, /tmp/ghc4087862_0/ghc_72.o, /tmp/ghc4087862_0/ghc_72.dyn_o )
[ 3 of 103] Compiling Converge.Prelude ( src/Converge/Prelude.hs, /tmp/ghc4087862_0/ghc_82.o, /tmp/ghc4087862_0/ghc_82.dyn_o )
[ 4 of 103] Compiling Converge.Domain.Id ( src/Converge/Domain/Id.hs, /tmp/ghc4087862_0/ghc_70.o, /tmp/ghc4087862_0/ghc_70.dyn_o )
...
Are there known issues with getting "hi-haddock" to work in nixpkgs?
| 15:06:02 |
