| 19 Dec 2024 |
 Alex | In reply to @b:chreekat.net
I have run into a "Hash mismatch in fixed output derivation '/nix/store/png58zkagrplhcvxvfnlk8saz2nvgbmc-source.drv'". Since the path is an anonymous "-source", I know it's probably a Haskell package. But how do I find out which? The trace just shows it's a dependency of a package that needs to get built, but there are a lot of those... nix why-depends --derivation is also an option | 13:15:57 |
| 20 Dec 2024 |
 jean-paul. | I want to have a Haskell package with a justStaticExecutables out and also a doc alongside it with Haddocks and such. Is there a good way to compose such a thing? | 21:57:12 |
| 21 Dec 2024 |
|  @stablejoy:matrix.org left the room. | 05:08:28 |
|  @dmiskovic:matrix.org left the room. | 05:14:07 |
| @stablejoy:matrix.org joined the room. | 06:43:44 |
 cdepillabout | sterni: thanks for fixing up stack | 09:57:53 |
 sterni (he/him) | jean-paul.: justStaticExecutables is pointless with haddocks since haddocks also have all the references on other packages. You probably want to use enableSeparateBinOutput which at least creates a bin output that won't pull in the world | 19:36:21 |
| jean-paul. | sterni: Does enableSeparateBinOutput make a bin output that's as small as the one justStaticExecutables makes? | 19:39:40 |
| jean-paul. | (If it does I wonder what the point of justStaticExecutables is) | 19:39:57 |
| jean-paul. | I want a small closure to deploy the software in production and a doc output to deploy elsewhere. I don't even care if everything has to get built separately for each of them, as long as the bin output is small and I can have the docs. | 19:40:41 |
 maralorn | Yeah, I wonder about that at times, too. I think it is mainly because separate bin is newer. | 19:47:34 |
| jean-paul. | Okay ... yes, enableSeparateBinOutput seems to also produce a bin output that is quite small ... in my case, 5.3 GiB for the combined output or 118 MiB for the separate bin output. That's quite a bit easier than all the hacking I was trying to do around justStaticExecutables. | 19:59:51 |
| sterni (he/him) | Alex: didn't we have this problem before?
error: In file included from /nix/store/23kvzbcicd0a8vhb3iszl529r7cdjaxk-libffi-riscv64-unknow
n-linux-gnu-3.4.6-dev/include/ffi.h:84,
from FFI.hsc:33:
/nix/store/23kvzbcicd0a8vhb3iszl529r7cdjaxk-libffi-riscv64-unknown-linux-gnu-3.4.6-dev/include
/ffitarget.h:36:2: error: #error "libffi was configured for a RISC-V target but this does not
appear to be a RISC-V compiler."
36 | #error "libffi was configured for a RISC-V target but this does not appear to be a RIS
C-V compiler."
| ^~~~~
| 20:18:05 |
| Alex | In reply to @sternenseemann:systemli.org
Alex: didn't we have this problem before?
error: In file included from /nix/store/23kvzbcicd0a8vhb3iszl529r7cdjaxk-libffi-riscv64-unknow
n-linux-gnu-3.4.6-dev/include/ffi.h:84,
from FFI.hsc:33:
/nix/store/23kvzbcicd0a8vhb3iszl529r7cdjaxk-libffi-riscv64-unknown-linux-gnu-3.4.6-dev/include
/ffitarget.h:36:2: error: #error "libffi was configured for a RISC-V target but this does not
appear to be a RISC-V compiler."
36 | #error "libffi was configured for a RISC-V target but this does not appear to be a RIS
C-V compiler."
| ^~~~~
Yes, but I still can't recall which fix was for this. I'll take a look tomorrow. | 20:22:56 |
| sterni (he/him) | Alex: we trigger the issue in a different way, it happens when building a cross compiler for riscv64-linux on https://github.com/NixOS/nixpkgs/pull/367194. Wanted to try the native backend… | 20:23:52 |
|  @sleepymonad:matrix.org joined the room. | 21:14:53 |
| @sleepymonad:matrix.org set a profile picture. | 21:56:33 |
| 22 Dec 2024 |
| cdepillabout | The new automatic merge strategy for haskell-updates seems to make it so that it is really hard to read through the history/timeline on the PR on github: https://github.com/NixOS/nixpkgs/pull/354270 | 02:29:31 |
| cdepillabout | * The new automatic merge strategy for haskell-updates (where master gets merged in, but the branch is based on staging) seems to make it so that it is really hard to read through the history/timeline on the PR on github: https://github.com/NixOS/nixpkgs/pull/354270 | 02:30:04 |
| cdepillabout | Any recent conversation seems like it just gets hidden in the github "hidden items" thing. | 02:31:03 |
| cdepillabout | * The new automatic merge strategy for haskell-updates (where master gets merged in automatically, but the branch is based on staging) seems to make it so that it is really hard to read through the history/timeline on the PR on github: https://github.com/NixOS/nixpkgs/pull/354270 | 02:31:18 |
| maralorn | Huh, that shouldn't be, should it? | 07:53:50 |
| maralorn | I think this is the result of something not going according to plan . | 07:55:36 |
| maralorn | Either an erroneous manual merge or the automatic merges are not correct yet. Or githubs diffing sucks. | 07:57:19 |
| Alex | In reply to @sternenseemann:systemli.org
Alex: we trigger the issue in a different way, it happens when building a cross compiler for riscv64-linux on https://github.com/NixOS/nixpkgs/pull/367194. Wanted to try the native backend…
- Changing to 9.10.1 progresses further
- Changing to aarch64-multiplatform progresses further
What an odd regression. I suspect it's upstream in GHC itself.
It is for some reason trying to use riscv64 libffi on an x86_64 build.
Maybe there's some arch-specific code that prevents that from happening on aarch64?
I tried diffing the build logs, but could not find anything of value. | 11:30:00 |
| sterni (he/him) | Alex: IIRC this #error is only an issue with riscv64 headers, so it may not happen at all with aarch64. The problem as I remember it is that something relating to libffi is also built with the stage0 compiler (so on x86_64-linux), but the wrong libffi used | 11:48:57 |
| sterni (he/him) | nothing has changed yet, it's just from targeting staging and merging in master/staging-next occasionally, so the branch gets ahead of staging sometimes | 11:49:30 |
| sterni (he/him) | also the web UI gets out of sync with the HEAD of staging so it sometimes even lists commits that already have been merged into staging | 11:49:56 |
| sterni (he/him) | See https://github.com/NixOS/nixpkgs/issues/361143 | 11:50:17 |
| @stablejoy:matrix.org left the room. | 13:25:36 |
