<typetetris>
Is the documentation on pinning hackage.nix / stackage.nix still recent?
tchouri has joined #haskell.nix
<typetetris>
I did this `sourceOverrides` stuff but it still downloads `hackage.nix` from github.
<typetetris>
`sourcesOverride` instead of `sourceOverrides`, but still, downloads hackage.nix from github.
<typetetris>
sourcesOverride isn't mentioned in all of hackage.nix ... :(
<typetetris>
Also I have the iohk nix substituter configured, but it still compiles ghc-8.4.4 for me ...
<typetetris>
Trying to sue ghc8102
o1lo01ol1o has joined #haskell.nix
o1lo01ol1o has quit [Ping timeout: 264 seconds]
tchouri has quit [Ping timeout: 240 seconds]
tchouri has joined #haskell.nix
fendor has joined #haskell.nix
<hexagoxel>
typetetris: ghc-8.4 is not in the iohk cache, so if you do have a build plan that uses it (old stackage snapshot?) then you will start bootstrapping it locally.
tchouri has quit [Remote host closed the connection]
tchouri has joined #haskell.nix
<hexagoxel>
but if your buildplan only uses ghc8102 then it should _not_ do that, it should get the cached 8102 only.
o1lo01ol1o has joined #haskell.nix
o1lo01ol1o has quit [Ping timeout: 246 seconds]
__monty__ has joined #haskell.nix
<typetetris>
its a cabal only build, as far as i can tell (just haskell-language-server, 0.4.0, cabal build, hackage.nix, stackage.nix from today)
mariatsji has quit [*.net *.split]
wedens[m] has quit [*.net *.split]
jonge[m] has quit [*.net *.split]
ptival[m] has quit [*.net *.split]
Poscat[m] has quit [*.net *.split]
wedens[m] has joined #haskell.nix
Poscat[m] has joined #haskell.nix
ptival[m] has joined #haskell.nix
jonge[m] has joined #haskell.nix
mariatsji has joined #haskell.nix
<typetetris>
seem like haskell-language-server has a fixed index state from `2020-08-07T11:45:57Z` could that influence how/which ghc is built?
<typetetris>
Okay now it fails, cause diagrams-svg doesn't like base 4.14, how can I relax that bound? Or do I have to vendor diagrams-svg?
<typetetris>
I wouldn't know where to place that option. Inside the set given to `haskell-nix.project` ? Do I need to create a package set first? Maybe a few examples would be really cool.
<typetetris>
Now I tried just to get a nix-shell with ghc 8.10.2 through haskell.nix, but the chapter points to `nixpkgs.haskell-nix.haskell.haskellPackages.ghcWithPackages` which is 8.6.5
<__monty__>
The default haskellPackages is a set based on an LTS stackage snapshot. You need a haskellPackages that comes with a different GHC version.
<typetetris>
__monty__: `nix-build -A diagrams-svg.components.setup.doExactConfig` with my plain scaffolding (changed compiler to ghc8102) default.nix (taken from `input-output-hk.github.io/haskell.nix/tutorials`) in haskell-language-server checkout yields the same error of diagrams-svg not liking base-4.14.
<typetetris>
__monty__ and how can I access a different haskellPackages?
<__monty__>
I'm not sure what doExactConfig does tbh. Does it force the package's config or the haskell.nix config? Maybe angerman can enlighten both of us : )
<__monty__>
You make one using mkStackPkgSet or mkCabalProjectPkgSet.
<__monty__>
Or mkPkgSet which is a bit deeper down.
<typetetris>
Thanks. Had a big knot in my head. Your pointer to `mkCabalProjectPkgSet` helped. I read about it, then about `project` calling `cabalProject` and then what `cabalProject` does and it states: `Then feeding its result into mkCabalProjectPkgSet passing also pkg-def-extras, extra-hackages and modules arguments.` So will try that now.
<typetetris>
hmm still stuck, what to put into `modules` the tests seem to do strange stuff with it.
<__monty__>
Try a nightly stackage snapshot? `haskell-nix.snapshots."nightly-2020-09-06".ghcWithPackages
<__monty__>
I think that'll be the easiest way to get a package set with a recent GHC.
<__monty__>
I'm not sure you need to specify a predefined snapshot or whether any nightly works though.
<__monty__>
And you tried diagrams-svg.components.setup.doExactConfig?
<typetetris>
If I change `components.library.doExactConfig` to `components.setup.doExactConfig` in my posted gist sadly nothing changes.
<__monty__>
Let's wait for Deus Ex Machina. The team does this as a job so weekends are usually pretty quiet.
<typetetris>
For me its the other way round :(
<typetetris>
I probably stick to standard nixpkgs for now, there I know my way around.
<__monty__>
Sure. I don't know how to do the equivalent of doJailbreak tbh.
<__monty__>
I think you're supposed to specify the package bounds to override in your cabal.project or something.
<__monty__>
The haskell.nix philosophy is everything from cabal and stack should just work rather than be reimplemented (sometimes poorly).
<julm>
using allow-newer: $pkg:$dep in cabal.project works well for me
<__monty__>
That sounds like the proper way to "jailbreak."
o1lo01ol1o has joined #haskell.nix
julm_ has joined #haskell.nix
julm_ has quit [Client Quit]
julm_ has joined #haskell.nix
<ocharles>
Anyone have any idea how to get `build-tool-depends: tasty-discover:tasty-discover` with shellFor to work? I can't seem to get a shell where this executable is in the tasty-discover package in the ghc-pkg database. Only the library
<__monty__>
Hmm, does specifying it in the tools attribute for shellFor work?
<ocharles>
Nope, tried that too
<ocharles>
I think that just gives you something on $PATH, but build-tool-depends needs it in ghc-pkg