2025/09/07

Newest at the top

2025-09-07 02:11:02 +0200trickard_(~trickard@cpe-53-98-47-163.wireline.com.au)
2025-09-07 02:10:48 +0200trickard(~trickard@cpe-53-98-47-163.wireline.com.au) (Read error: Connection reset by peer)
2025-09-07 02:10:09 +0200Guest62(~Guest62@c-73-217-79-154.hsd1.co.comcast.net)
2025-09-07 02:06:53 +0200potato44(uid421314@id-421314.lymington.irccloud.com) potato44
2025-09-07 02:01:44 +0200merijn(~merijn@host-vr.cgnat-g.v4.dfn.nl) (Ping timeout: 256 seconds)
2025-09-07 02:00:48 +0200sprotte24(~sprotte24@p200300d16f1ae200cce55c66184a5d60.dip0.t-ipconnect.de) (Quit: Leaving)
2025-09-07 02:00:14 +0200 <slondr> yeah that's definitely not the case here. I'm in a blank directory with one .hs file
2025-09-07 02:00:06 +0200Tuplanolla(~Tuplanoll@91-159-69-59.elisa-laajakaista.fi) (Quit: Leaving.)
2025-09-07 01:59:54 +0200 <geekosaur> that said, nix wouldn't insert a ghc into your path unless you're in a directorry with a shell.nix and something knows enough about nix to see that file and load the derivations it specifies
2025-09-07 01:57:44 +0200 <geekosaur> shells cache paths, but something run in a subshell or by a program you run from that shell won;t see the cache and will find the nix one first in that case
2025-09-07 01:57:07 +0200 <slondr> though 'echo $PATH' sure does report that the /nix bin directories are above the ghcup bin directories, hmm
2025-09-07 01:56:44 +0200merijn(~merijn@host-vr.cgnat-g.v4.dfn.nl) merijn
2025-09-07 01:56:16 +0200 <slondr> 'which ghc' reports: $HOME/.ghcup/bin/ghc
2025-09-07 01:55:56 +0200 <geekosaur> the ghc from ghcup won't know about nix
2025-09-07 01:55:42 +0200 <geekosaur> make sure nix isn't substituting its own ghc for the one you intended to use
2025-09-07 01:54:47 +0200 <slondr> I'm not sure why it does
2025-09-07 01:54:42 +0200 <slondr> I don't want ghc to reference anything nix-related at all
2025-09-07 01:53:43 +0200 <geekosaur> nix really wants to be hermetic, but that can break things pretty thoroughly when it comes to stuff like glibc
2025-09-07 01:52:52 +0200 <geekosaur> https://github.com/NixOS/patchelf
2025-09-07 01:52:25 +0200 <slondr> which seems almost certainly wrong, I'd want ghc to reference the system glibc not some random nix derivation glibc
2025-09-07 01:52:00 +0200 <slondr> wait wait wait wait. the glibc error seems to be referencing a glibc version installed by nix
2025-09-07 01:51:40 +0200 <geekosaur> okay, if Arch changed their glibc recently then ghcup's release builds probably need to be adjusted
2025-09-07 01:50:42 +0200fp(~Thunderbi@89-27-10-140.bb.dnainternet.fi) (Ping timeout: 260 seconds)
2025-09-07 01:48:07 +0200 <slondr> I ran GHCup via the curl | bash command on the website
2025-09-07 01:47:54 +0200 <slondr> ah you see, i have already done that. my ghc is from GHCup.
2025-09-07 01:47:33 +0200 <geekosaur> this might be a question for maerwald, but the answer may again be "give up on the Arch ghc install" because it just causes too many problems for Haskell maintainers
2025-09-07 01:46:17 +0200merijn(~merijn@host-vr.cgnat-g.v4.dfn.nl) (Ping timeout: 258 seconds)
2025-09-07 01:45:43 +0200 <slondr> yeah I just have the normal default Arch Linux glibc
2025-09-07 01:45:30 +0200 <geekosaur> also I have 2.41 but it's Ubuntu's build
2025-09-07 01:45:18 +0200 <geekosaur> okay, so that's not it (same here)
2025-09-07 01:44:35 +0200 <slondr> libc ABIs: UNIQUE IFUNC ABSOLUTE
2025-09-07 01:44:21 +0200ljdarj1ljdarj
2025-09-07 01:44:21 +0200ljdarj(~Thunderbi@user/ljdarj) (Ping timeout: 250 seconds)
2025-09-07 01:44:18 +0200hiredman(~hiredman@frontier1.downey.family) hiredman
2025-09-07 01:44:12 +0200 <geekosaur> if you can find libc.so.6, can you run it as if it were a program and if so what does it say for "libc ABIs"?
2025-09-07 01:43:17 +0200hiredman(~hiredman@frontier1.downey.family) (Remote host closed the connection)
2025-09-07 01:42:13 +0200 <geekosaur> the question is whether or how your glibc supports linux's native posix threads library (the "nptl" in that symbol)
2025-09-07 01:41:46 +0200merijn(~merijn@host-vr.cgnat-g.v4.dfn.nl) merijn
2025-09-07 01:41:05 +0200 <slondr> I'm on glibc 2.42 should I expect hls to support that?
2025-09-07 01:40:55 +0200ljdarj1(~Thunderbi@user/ljdarj) ljdarj
2025-09-07 01:39:16 +0200 <geekosaur> the undefined symbol one sounds like much more of a problem
2025-09-07 01:39:01 +0200 <geekosaur> right, you need an HLS binary for each ghc version because the ghc-api it uses must match
2025-09-07 01:36:48 +0200 <slondr> I also see: libc.so.6: undefined symbol: __nptl_change_stack_perm, version GLIBC_PRIVATE
2025-09-07 01:36:16 +0200 <slondr> emacs lsp reports: ghc 9.6.7, cabal 3.12.1.0, stack 3.3.1
2025-09-07 01:33:38 +0200 <slondr> It looks like I also have haskell-language-server-9.6.7
2025-09-07 01:33:02 +0200 <slondr> my hls binary is named "haskell-language-server-9.4.8"
2025-09-07 01:32:51 +0200 <slondr> oh! that's a clue
2025-09-07 01:32:18 +0200 <geekosaur> also that sounds like an out of date ghcup, or at least ghcup metadata; current metadata has 9.6.7 as recommended
2025-09-07 01:31:30 +0200 <geekosaur> HLS versions don't go that high
2025-09-07 01:31:22 +0200 <geekosaur> you mean ghc version?