Re: [Nix-dev] Hydra: hashes change?

2013-09-21 Thread Eelco Dolstra
Hi, On 20/09/13 21:50, Peter Simons wrote: > Unfortunately, my theory was wrong: is > clearly *not* a Haskell issue: Indeed it isn't. It appears that the Hydra server was not shutdown cleanly when we rebooted it on Wednesday, so several store paths have be

Re: [Nix-dev] Hydra: hashes change?

2013-09-20 Thread Peter Simons
Unfortunately, my theory was wrong: is clearly *not* a Haskell issue: | these derivations will be built: | /nix/store/57xga0i378j0vmn89112pmngam1ghfvx-ghdl-wrapper-0.29.drv | /nix/store/l0iy63ynvb6jnhh59vsva1scg8l55jb2-gnat-wrapper-4.5.4.drv | /nix/

Re: [Nix-dev] Hydra: hashes change?

2013-09-20 Thread Vladimír Čunát
On 09/20/2013 07:19 PM, Peter Simons wrote: Maybe two build slaves ended up building GHC 7.6.3 independently from each other, and now we have two mutually incompatible GHC binaries lying around on the Hydra cluster? Ah, that didn't occur to me. I didn't even notice it was all haskell. You're

Re: [Nix-dev] Hydra: hashes change?

2013-09-20 Thread Peter Simons
Hi Vladimír, > I noticed that builders now often complain that hashes of some path > have changed, e.g.: https://hydra.nixos.org/build/6151693 I believe that this is an issue with GHC, not with Hydra. GHC cannot guarantee that two identical builds of itself yield identical results: http://gh

[Nix-dev] Hydra: hashes change?

2013-09-20 Thread Vladimír Čunát
Hi, I noticed that builders now often complain that hashes of some path have changed, e.g.: https://hydra.nixos.org/build/6151693 I'm not sure how these conditions are checked, but IMO it could be dangerous to have modified paths in Hydra's store. By my estimate, in the last 1-2 days there a