| 10 Feb 2025 |
 alexfmpe | yeah, we're kind of in ripping off the band aid mode at this point | 22:35:25 |
 maralorn | I mean really we in this channel are really not responsible for any of the ecosystem developments we just try to keep it all working together. đŸ˜„ | 22:35:33 |
| alexfmpe | do you have a sense on how old their nixpkgs pin is? | 22:38:33 |
| alexfmpe | like, are they using recent nixpkgs or not anywhere near that | 22:39:31 |
 @dmjio:matrix.org | I'll try to get him in here to ask so I don't say the wrong thing | 22:39:52 |
| @dmjio:matrix.org | alright let me try this, ideally I could just update miso's nix scripts to this and then nixpkgs 25 and hope and pray for the best? | 22:41:16 |
 emily | (I think the decisions that lead to being on years-old Nixpkgs and years-old GHC and years-old […] are not uncorrelated…) | 22:41:42 |
| alexfmpe | well I don't know about the "linking" bits of closure compiler and whatnot, but haskell-wise that's building on haskell-updates | 22:41:46 |
| alexfmpe | well it might not matter for you if you're only getting libraries out of nixpkgs | 22:42:28 |
| emily | (which is fine, until you diverge from the pattern by also picking "new OS version" or "new Nixpkgs version" or … and then either something breaks or a whole lot of maintenance burden just got offloaded on someone else) | 22:42:36 |
| alexfmpe | it'll just be a problem for me getting miso-examples building and running in nixpkgs heh | 22:42:51 |
| emily | (and since it's Nixpkgs it's probably the responsibility-without-authority kind that has reliably burned people out for years…) | 22:43:09 |
 Tristan Ross | In reply to @emilazy:matrix.org
(and since it's Nixpkgs it's probably the responsibility-without-authority kind that has reliably burned people out for years…) Yeah, I think we're fixing the authority problem a bit with teams. Having people who are dedicated to working on specific things help. | 22:44:27 |
| alexfmpe | note you probably won't get much caching from nixos.org yet, especially on mac
not a problem for your downstream, but might be annoying now for you if you end up compiling 9.10 to compile 9.12 and such | 22:44:52 |
| alexfmpe | * note you probably won't get much caching from nixos.org yet, especially on mac
not a problem for your downstream since you have your own caching, but might be annoying now for you if you end up compiling 9.10 to compile 9.12 and such | 22:45:02 |
| emily | a team only means anything if it has defined responsibilities and authority and a clear path to resolution when teams have conflicting needs, which is not something Nixpkgs has historically been any good at | 22:45:38 |
| @dmjio:matrix.org | this tbh, got so bad I stopped doing version bounds in my cabal file, because you /had/ to use nix to work with ghcjs, stack dropped support, cabal-install was only used for building, nix was the resolver. Now stuff "works" again and the cabal people are upset that I'm making it hard for their constraint solver by underspeciying constraints | 22:46:57 |
| Tristan Ross | In reply to @emilazy:matrix.org
a team only means anything if it has defined responsibilities and authority and a clear path to resolution when teams have conflicting needs, which is not something Nixpkgs has historically been any good at Yeah, one of the things I've learned is if you want something done then you have to take authority because not many people are going to. | 22:47:09 |
| @dmjio:matrix.org | I think GHCJS just reads all the JS into memory from every dep. to concat it (aka JS linking) and it just blows the heap | 22:51:48 |
| @dmjio:matrix.org | a user said builds were taking 8GB resident and minutes to link | 22:53:14 |
| alexfmpe | I do remember 16GB ram being recommended minimum for ghcjs linking heh | 22:53:45 |
| alexfmpe | data point: I just checked and nix-build -A haskell.packages.ghc810.miso builds on a days-old master | 22:54:56 |
| alexfmpe | it doesn't on haskell-updates, but only like 4 deps fail, I don't mind fixing that | 22:55:05 |
| alexfmpe | so I guess the question becomes, is anyone using miso with ghcjs from today's nixpkgs that would also be driven into a corner if they didn't get to bump nixpkgs until js backend was feature par out of nixpkgs? | 22:55:11 |
| alexfmpe | (and that's assuming the ghcjs build works on that nixpkgs, but that build takes longer to have an answer for) | 22:56:50 |
| alexfmpe | * (and that's assuming that ghcjs 8.10 build works on this days-old nixpkgs, but that build takes longer to have an answer for) | 22:57:06 |
| @dmjio:matrix.org | that I'm not sure, there's one person in particular who would know Tysonzer0 and he's not online, but I've pinged to join. All of miso's examples and haskell-miso.org use the 6 year old hash. | 22:58:04 |
| @dmjio:matrix.org | I can make a PR to update my nix scripts and bump the nixpkgs hash to latest see how far it gets. | 23:04:18 |
| @dmjio:matrix.org | alexfmpe let me check the payload sizes again rn and compare with what you've shown | 23:04:27 |
| alexfmpe | I'd say compare with and without -dedupe, but not closure nor compression | 23:05:06 |
