bundler: hash mismatch of source

[rzetterberg]

When activating the gitlab service and rebuilding:

    services.gitlab = { 
      enable = true;
      databasePassword = "removed-for-issue-report";
    }; 

[zed@nixos-workstation:~]$ sudo nix-channel --update
[zed@nixos-workstation:~]$ sudo nixos-rebuild switch --upgrade --verbose

Fails with:

trying https://rubygems.org/downloads/mysql2-0.3.16.gem
  % Total    % Received % Xferd  Average Speed   Time    Time     Time  Current
                                 Dload  Upload   Total   Spent    Left  Speed
100   154  100   154    0     0    131      0  0:00:01  0:00:01 --:--:--   131Receiving objects:  77% (84150/109285), 16.73 MiB | 494.00 KiB/s   
  0     0    0     0    0     0      0      0 --:--:--  0:00:01 --:--:--     0scanning for references inside ‘/nix/store/p65pkq14sdjb6vq7clpwyv3qqj7x9718-bundler-a2343c9’
output path ‘/nix/store/p65pkq14sdjb6vq7clpwyv3qqj7x9718-bundler-a2343c9’ should have r:sha256 hash ‘1l4r55n1wzr817l225l6pm97li1mxg9icd8s51cpfihh91nkdz68’, instead has ‘0l9ghs6xg0pv3ssik807jcvai3f0w83cskx7ymq6kr8f1h8lqbp0’
building path(s) ‘/nix/store/34n6wbchzi8fvskxarsj8wd979j902zk-mysql2-pre-install’
cannot build derivation ‘/nix/store/l248881kizq0wm296syb0g0ck0rpfppg-ruby-2.1.3-p0-bundler-HEAD.drv’: 1 dependencies couldn't be built

System specs

[zed@nixos-workstation:~]$ sudo nix-channel --list
nixos https://nixos.org/channels/nixos-unstable
[zed@nixos-workstation:~]$ uname -a
Linux nixos-workstation 3.14.29 #1-NixOS SMP Thu Jan 1 00:00:01 UTC 1970 x86_64 GNU/Linux

Misc

For convenience here is the link to the latest commit to ruby-bundler: 7d8b2f0#diff-7cde46455d84d468d1a1b9c57fc96287

I does not seem like that commit introduced the problem. The hash was changed from 1f0isjrn4rwak3q6sbs6v6gqhwln32gv2dbd98r902nkg9i7y5i0 to 1l4r55n1wzr817l225l6pm97li1mxg9icd8s51cpfihh91nkdz68, but the actual output on my machine was 0l9ghs6xg0pv3ssik807jcvai3f0w83cskx7ymq6kr8f1h8lqbp0.

cc @cstrahan

I can provide more information on request. I'm still pretty new to NixOS so I'm not sure what kind of information is relevant for creating issues.

[rzetterberg]

An update on this issue, I just reproduced the bug on my machine. This time the output hash was different from the first time I tried:

output path ‘/nix/store/p65pkq14sdjb6vq7clpwyv3qqj7x9718-bundler-a2343c9’ should have r:sha256 hash ‘1l4r55n1wzr817l225l6pm97li1mxg9icd8s51cpfihh91nkdz68’, instead has ‘1fywz0m3bb0fmcikhqbw9iaw67k29srwi8dllq6ni1cbm1xfyj46’
building path(s) ‘/nix/store/04h4awa2x6v697kfpdzbbzkr3q2r45ga-gon-5.0.1.gem’
cannot build derivation ‘/nix/store/l248881kizq0wm296syb0g0ck0rpfppg-ruby-2.1.3-p0-bundler-HEAD.drv’: 1 dependencies couldn't be built

[cstrahan]

@rzetterberg Are you on Linux or Darwin?

EDIT: Oops! Just saw your "System specs" you posted earlier. I'll give this some thought...

[cstrahan]

@rzetterberg Do you have chroot builds enabled? (nix.useChroot = true; in /etc/nixos/configuration.nix)

I forget just how impure non-chroot builds are... but a hunch is that custom git hooks could be polluting the checkout. (I think that $HOME is set either way, so I sorta doubt this hunch of mine.)

[rzetterberg]

@cstrahan I have not set useChroot, so it's using the default value. Should I set it to true and try again?

I checked the bundler repo to see whether there are any git hooks used, and it seems there are no hooks (only samples) in the .git/hooks directory:

zed@shinji:~/tmp/bundler/.git/hooks$ ls -la
total 48
drwxr-xr-x 2 zed zed 4096 Feb 21 11:56 .
drwxr-xr-x 8 zed zed 4096 Feb 21 11:57 ..
-rwxr-xr-x 1 zed zed  452 Feb 21 11:55 applypatch-msg.sample
-rwxr-xr-x 1 zed zed  896 Feb 21 11:55 commit-msg.sample
-rwxr-xr-x 1 zed zed  189 Feb 21 11:55 post-update.sample
-rwxr-xr-x 1 zed zed  398 Feb 21 11:55 pre-applypatch.sample
-rwxr-xr-x 1 zed zed 1642 Feb 21 11:55 pre-commit.sample
-rwxr-xr-x 1 zed zed 1239 Feb 21 11:55 prepare-commit-msg.sample
-rwxr-xr-x 1 zed zed 1352 Feb 21 11:55 pre-push.sample
-rwxr-xr-x 1 zed zed 4898 Feb 21 11:55 pre-rebase.sample
-rwxr-xr-x 1 zed zed 3611 Feb 21 11:55 update.sample

[vcunat]

Chrooting is irrelevant, as AFAIK fixed-output derivations are never chrooted, and they have network access, etc. Better discuss this just on #6409, as there are some more relevant discussions about leaveDotGit.

[rzetterberg]

After c00d36f this is problem was resolved for me. Can I close this issue? @vcunat

[vcunat]

That commit probably wasn't affecting this. Anyway, if it works now...

[anderspapitto]

I'm seeing this same issue today, on the latest unstable

[vcunat]

That's what fetchgit with leaveDotGit currently does: #8567.