Logs: liberachat/#haskell
| 2026-02-11 17:36:38 | → | perryprog joins (~perryprog@wikipedia/perryprog) |
| 2026-02-11 17:37:25 | <haskellbridge> | <sm> __monty__ I think that's implied, isn't it ? Shake's dependency engine will do that by default |
| 2026-02-11 17:37:52 | <haskellbridge> | <sm> if by <file> you mean a shake target |
| 2026-02-11 17:39:47 | <perryprog> | I'm having some trouble troubleshooting a build failure—I'm getting "The build process segfaulted" from cabal and I'm not seeing much help from cabal build -v (https://paste.tomsmeding.com/aGOsdU5i). The project is https://github.com/pdobsan/oama. There's reports of this sort of build failure from a few years ago also on macOS, but that was from an older macOS version and also seems to have been fixed... |
| 2026-02-11 17:40:17 | × | karenw quits (~karenw@user/karenw) (Ping timeout: 250 seconds) |
| 2026-02-11 17:41:18 | → | tzh joins (~tzh@c-76-115-131-146.hsd1.or.comcast.net) |
| 2026-02-11 17:43:46 | <haskellbridge> | <sm> are you running out of memory perryprog ? watch with a fast-updating top , eg |
| 2026-02-11 17:44:11 | <perryprog> | nope, I got loads |
| 2026-02-11 17:44:49 | <Guest2> | Hey, sorry for the stupid question, but is there a built-in function to "interleave" two lists? Like f [1,2,3] [11,12,13] = [1,11,2,12,3,13]. I for some reason thought there was one, but I can't find it now, did I dream it up or something? |
| 2026-02-11 17:45:04 | <haskellbridge> | <sm> I see Cabal-7125, but as usual it's not in the haskell error index |
| 2026-02-11 17:45:51 | <haskellbridge> | <sm> are you sure it's not eating up your loads of memory due to a compiler bug ? I would make sure |
| 2026-02-11 17:46:08 | × | bliminse_ quits (~bliminse@user/bliminse) (Server closed connection) |
| 2026-02-11 17:46:21 | <__monty__> | sm: Maybe I was adding my dependencies wrong. I refactored a bit and now I'm getting the expected behavior. |
| 2026-02-11 17:46:24 | → | bliminse joins (~bliminse@user/bliminse) |
| 2026-02-11 17:46:34 | <perryprog> | sm, I'm sure—I just checked with htop -d 1. 10 GB free the whole time. |
| 2026-02-11 17:46:35 | <haskellbridge> | <sm> \o/ |
| 2026-02-11 17:47:02 | <haskellbridge> | <sm> perryprog good, just making sure |
| 2026-02-11 17:47:04 | <perryprog> | nw |
| 2026-02-11 17:47:25 | <haskellbridge> | <sm> Guest2 you're thinking of zip maybe. It would be a start |
| 2026-02-11 17:48:20 | <Guest2> | Wouldn't that give me a list of pairs? I waned to combine two lists into one of the same type |
| 2026-02-11 17:48:31 | <probie> | > concat $ zipWith (\x y -> [x,y]) [1,2,3] [11,12,13] |
| 2026-02-11 17:48:32 | <lambdabot> | [1,11,2,12,3,13] |
| 2026-02-11 17:48:59 | <Guest2> | Oh that's clever |
| 2026-02-11 17:49:04 | <Guest2> | Thanks |
| 2026-02-11 17:49:13 | <probie> | Or use `foldr`. `foldr` solves everything |
| 2026-02-11 17:50:08 | <geekosaur> | perryprog: if I read that log correctly, it's ghc that segfaulted so you may want to file an issue on gitlab |
| 2026-02-11 17:50:29 | <probie> | > foldr (\x k ys -> case ys of { [] -> []; (y:ys') -> x:y:k ys' }) (const []) [1, 2, 3] [11, 12, 13] -- to clarify, this is a joke |
| 2026-02-11 17:50:30 | <lambdabot> | [1,11,2,12,3,13] |
| 2026-02-11 17:50:31 | <perryprog> | but I don't want to file an issue, I want my build to work ;-; /hj |
| 2026-02-11 17:51:09 | <haskellbridge> | <sm> I was going to say ask #hackage:matrix.org (https://matrix.to/#/#hackage:matrix.org) . It looks like a very straightforward package, clearly there's a tool bug |
| 2026-02-11 17:51:25 | <geekosaur> | alternatively it might be https://github.com/haskell/cabal/issues/11465, but as it's not custom `Setup` and it reached running ghc that seems pretty unlikely |
| 2026-02-11 17:51:48 | <geekosaur> | (because in that ticket it's `Setup` that segfaults, so it wouldn't reach running ghc) |
| 2026-02-11 17:52:05 | <haskellbridge> | <sm> you could also try building with a different ghc version, cabal-install version, and/or stack |
| 2026-02-11 17:52:37 | <haskellbridge> | <sm> I might even try building in a fresh clone, to rule out ghc environment files |
| 2026-02-11 17:52:47 | × | Guest2 quits (~Guest2@client-8-178.eduroam.oxuni.org.uk) (Quit: Client closed) |
| 2026-02-11 17:52:59 | <perryprog> | I tried with a git clean -fxd which should be same thing |
| 2026-02-11 17:54:11 | <perryprog> | the project needs 9.12.1 or higher, so I kinda doubt it needs 9.12.1 and not 9.12.2 unless it relied on inaccurate division somehow |
| 2026-02-11 17:54:13 | <perryprog> | (/s) |
| 2026-02-11 17:55:28 | × | rncwnd quits (~quassel@2a01:4f8:221:27c6::1) (Server closed connection) |
| 2026-02-11 17:55:40 | → | rncwnd joins (~quassel@2a01:4f8:221:27c6::1) |
| 2026-02-11 17:56:15 | <geekosaur> | might still be worth trying 9.12.1 |
| 2026-02-11 17:56:27 | × | sord937_ quits (~sord937@gateway/tor-sasl/sord937) (Quit: sord937_) |
| 2026-02-11 17:56:30 | <geekosaur> | in case a GC bug crept into .2 |
| 2026-02-11 17:56:50 | haskellbridge | sm tries 9.12.2 on mac |
| 2026-02-11 17:57:43 | <perryprog> | failed on cabal 3.14.2.0 |
| 2026-02-11 17:58:21 | <perryprog> | my stack version /shouldn't/ matter if it's just a cabal project, right?? |
| 2026-02-11 17:59:21 | <haskellbridge> | <sm> stack version doesn't matter much, no |
| 2026-02-11 18:02:52 | <haskellbridge> | <sm> it built normally for me with cabal 3.16.1.0 and ghc 9.12.2, on mac |
| 2026-02-11 18:03:14 | <haskellbridge> | <sm> that module 6 did take an unusually long time |
| 2026-02-11 18:03:53 | <haskellbridge> | <sm> ok, not so bad the second time |
| 2026-02-11 18:04:34 | <perryprog> | hmmm. What version of xcode command line tools and what macOS version? |
| 2026-02-11 18:05:18 | <perryprog> | and I still failed to build with 9.12.1 :( |
| 2026-02-11 18:08:12 | <tomsmeding> | perryprog: regardless of whether you submit a bug report to GHC or not, it might be worth minimising a little; there's a bunch of TemplateHaskell in the failing file, what if you remove all of that? |
| 2026-02-11 18:08:20 | <haskellbridge> | <sm> macos 26.2, xcode-select version 2416. |
| 2026-02-11 18:08:22 | <perryprog> | got an lldb backtrace: https://paste.tomsmeding.com/eDJZhqvX |
| 2026-02-11 18:08:31 | <tomsmeding> | (oh actually, just one usage of TH) |
| 2026-02-11 18:08:48 | × | tureba quits (tureba@tureba.org) (Server closed connection) |
| 2026-02-11 18:09:01 | <tomsmeding> | that runInteractiveProcess, in combination with the TH, makes me suspect external-interpreter somehow |
| 2026-02-11 18:09:05 | → | tureba joins (tureba@tureba.org) |
| 2026-02-11 18:09:06 | <tomsmeding> | definitely try removing the TH from that module |
| 2026-02-11 18:09:22 | <perryprog> | sm, xcode-select --version is a prank, that's just the installer's version. If you have xcode installed you can open it > preferences > components > then whatever's under platform support |
| 2026-02-11 18:10:18 | <tomsmeding> | % System.Process.system "echo --version" |
| 2026-02-11 18:10:18 | <yahb2> | --version ; ExitSuccess |
| 2026-02-11 18:10:21 | <perryprog> | hahaha tomsmeding you're correct |
| 2026-02-11 18:10:27 | <perryprog> | removing that fixes it |
| 2026-02-11 18:10:38 | <tomsmeding> | ok then there's definitely something off either with GHC or with your installation |
| 2026-02-11 18:10:43 | <perryprog> | that checks out |
| 2026-02-11 18:10:49 | <perryprog> | this was my first time updating it in a while |
| 2026-02-11 18:11:11 | <tomsmeding> | if you want to debug this, then my first suggestion would be to see if _any_ TH at all works -- i.e. `cabal new`, add some TH, build it |
| 2026-02-11 18:11:31 | <haskellbridge> | <sm> it might be an issue known to cabal devs |
| 2026-02-11 18:11:35 | <tomsmeding> | but this may also give you enough to devise a workaround if you just want oama to work |
| 2026-02-11 18:11:53 | <perryprog> | well I have to figure out the cause out of spite now, surely :) |
| 2026-02-11 18:11:55 | <haskellbridge> | <sm> ok. That's too much work :) |
| 2026-02-11 18:12:00 | <tomsmeding> | sm: SIGSEGV when running TH sounds like a GHC / platform toolchain problem, not a cabal problem, but I can't know for sure |
| 2026-02-11 18:12:02 | <haskellbridge> | <sm> could it be a mac permissions issue, like your terminal doesn't have full disk access or something ? |
| 2026-02-11 18:12:22 | → | prdak joins (~Thunderbi@user/prdak) |
| 2026-02-11 18:12:39 | × | prdak quits (~Thunderbi@user/prdak) (Read error: Connection reset by peer) |
| 2026-02-11 18:12:40 | <perryprog> | sm, nah, it's an address boundary error, so almost definitely not. And thank you for your help! |
| 2026-02-11 18:12:50 | → | prdak1 joins (~Thunderbi@user/prdak) |
| 2026-02-11 18:13:20 | <tomsmeding> | well, it's a null pointer dereference; who knows, perhaps the GHC RTS assumes that some file will be present, opens it, receives NULL, then proceeds to read from that? |
| 2026-02-11 18:13:44 | <tomsmeding> | it would be rather stupid and unlike the GHC RTS developers to do something like that (if it's that simple), but who knows |
| 2026-02-11 18:13:54 | <perryprog> | eh fair |
| 2026-02-11 18:14:11 | <haskellbridge> | <sm> claude says it's the process library's posix_spawn wrapper crashing when GHC tries to shell out to an external tool during compilation, and suggests to force a rebuild or upgrade of that package |
| 2026-02-11 18:14:51 | <tomsmeding> | at that point `rm -rf ~/.cabal/store`, and uninstall the relevant GHC, then reinstall the relevant GHC with ghcup and rebuild |
| 2026-02-11 18:15:06 | <tomsmeding> | (because 'process' may well from from the bootlibs distributed with GHC) |
| 2026-02-11 18:15:08 | prdak1 | is now known as prdak |
| 2026-02-11 18:15:30 | <tomsmeding> | (the ~/.cabal/store nuke can be more targeted by only removing the relevant directory inside) |
| 2026-02-11 18:15:32 | <haskellbridge> | <sm> I suspect it was installed before a macos upgrade, maybe |
| 2026-02-11 18:15:44 | <perryprog> | me too sm |
| 2026-02-11 18:16:10 | <tomsmeding> | if that's the problem that still points to a bug in ghc/rts/`process`/whatever, but it's worth trying |
| 2026-02-11 18:20:58 | <perryprog> | belated but $[|1+2|] works |
| 2026-02-11 18:21:20 | × | chele quits (~chele@user/chele) (Remote host closed the connection) |
| 2026-02-11 18:23:22 | <tomsmeding> | and $(liftIO $ LitE . StringL <$> readProcess "echo" ["hi"] "") ? |
| 2026-02-11 18:23:33 | <tomsmeding> | (import Control.Monad.IO.Class, Sstem.Process, Language.Haskell.TH) |
| 2026-02-11 18:23:37 | <tomsmeding> | *System.Process |
| 2026-02-11 18:24:48 | × | Ging_ quits (46fea76d80@2001:bc8:1210:2cd8::470) (Server closed connection) |
| 2026-02-11 18:24:54 | <tomsmeding> | if that works then the problem is somewhere between "process execution in TH" and "'git' execution in TH in this cabal project" :p |
| 2026-02-11 18:24:55 | → | Ging_ joins (46fea76d80@2001:bc8:1210:2cd8::470) |
| 2026-02-11 18:25:26 | <perryprog> | that works as well, and blowing out ~/.cabal/store did not fix oama building |
| 2026-02-11 18:25:52 | <tomsmeding> | that this point it's binary search in reproducer complexity then |
All times are in UTC.