<dhess>
michaelpj: ahh ok, so that also works for already-supported GHCs that just happen to have new materializations because of, say, a bumped nixpkgs pin?
<dhess>
I wonder if ci.nix needs to be tweaked for nixpkgs master...
<dhess>
there is some special-casing in there for different nixpkgs branches
<michaelpj>
I mean, that's not the compiler that failed, that's something part way through the build for a test
<dhess>
michaelpj: I suppose so, but to me, all that matters is I can't build the ghc865 target using ci.nix and an updated nixpkgs pin :)
<michaelpj>
sure, but "the compiler build doesn't work" would be a lot worse than "some of the tests don't work". especially since this looks like it's the static build stuff, which is always finicky
<michaelpj>
whereas it looks like the 8.10 one failed on the compiler itself
<dhess>
in fact, the ghc8102 target failed while building ghc 8.6.5, which is weird.
__monty__ has joined #haskell.nix
<angerman>
I might get around setting up some CI in the next few weeks.
<dhess>
angerman: anything obvious from those build logs? Anything you've seen before?
<michaelpj>
dhess: I think I've seen something like those failures before sometimes when trying to re-build lib:ghc. I don't know why that would happen, though
<michaelpj>
no idea what's going on with the cross/static failure