angerman changed the topic of #haskell.nix to: - alternative haskell infrastructure for nix; logs at
<tnks> I'm not sure what changed, but I feel like before, I'd be able to `nix show-derivation` to check a materialization, and it wasn't fast, but it didn't have to build a ton of Haskell. But now... that same action is building quite a lot.
<tnks> My gut feeling is that less things changed in code, but that your CI didn't cache something with recent changes.
<tnks> I just noticed what I think is `haskell-language-server` in the haskell.nix build. Are you building and caching the GHC versions in the table? Because if so, then I think my `nix-haskell-hls` project should align with the haskell-language-server in there to get a cache hit.
thongpv87 has quit [Remote host closed the connection]
<tnks> Man... cruising around the code, I had no idea that you could do "latest" for a hackage build. That is so convenient, because I was looking this stuff up all the time.
<tnks> Anyway, I appreciate help figuring out how to check this materialization without having to build so much.
<tnks> I suppose I could bisect haskell.nix to find out when this problem occurred. So much work, though, to do that.
<tnks> Okay, so I'm noticing that to check the materialization of what I'm actually trying to build, it's also checking the materialization of the dummy-ghc. But not sure the latter is why it's building all these uncached things.
<tnks> I mean... where in the world is hnix coming into play?
<tnks> ah man... am I building nix-tools?
<tnks> I think that's it. I think I'm getting a cache miss on nix-tools. Hopefully not GHC as well, though.
<tnks> I'm only trying to do a `show-derivation` to check the materialization. After that I was going to let GitHub Actions do the real build.
hekkaidekapus{ has joined #haskell.nix
hekkaidekapus} has quit [Ping timeout: 268 seconds]
stites[m] has quit [Quit: Idle for 30+ days]
<michaelpj> tnks: you're probably building nix-tools. Note that for cabal projects those also get built with your GHC and index-state to avoid having too many duplicate dependencies, so you may be less likely to get cache hits. You should make sure you build `project.roots` in your CI, because that pins all the stuff that's needed at eval-time
<tnks> michaelpj: I was relying on what IOHK is building. I turned out an update of Haskell.nix with a recent index-state got me hitting IOHK's cache again.
<tnks> For now, I'm just going to settle for what GHC versions IOHK is building, testing, and caching. I could support more with my own CI, but I kind of don't want to.
__monty__ has joined #haskell.nix
<michaelpj> yeah, building your own GHCs sucks
fendor has joined #haskell.nix
srk has quit [Remote host closed the connection]
srk has joined #haskell.nix
<tnks> Okay... I just noticed that the `updateMaterialized` script is now available as a passthru, so I can `nix run` it. This is great!
<__monty__> Ah, how do you call it? I've been piping to bash >.<
<tnks> `nix run --file . myPackageAttr.updateMaterialized --command updateMaterialized` is what I'm trying. But I seem to be building ghc this way. :(
<michaelpj> that seems odd
srk has quit [Remote host closed the connection]
proofofkeags has joined #haskell.nix
srk has joined #haskell.nix
srk has quit [Remote host closed the connection]
srk has joined #haskell.nix
<__monty__> Trying to build xmonad-contrib. It's failing because of a missing header from libXss. How do I provide this? Should haskell.nix pick up the right C dependencies from the cabal file or something?
<__monty__> Ah, looks like I'm in luck and the docs happened to have an example precisely for the X11 package.
fendor has quit [Remote host closed the connection]
fendor has joined #haskell.nix
<michaelpj> hmm, I thought we fixed a bunch of the missing X libs recently
<__monty__> My pin is probably over a month old.
__monty__ has quit [Quit: leaving]
fendor has quit [Read error: Connection reset by peer]