Zero Hydra Failures for 16.09

[domenkozar]

This is a call for testing of release-16.09 branch and fixing build failures.

Build failures of latest evaluation on http://hydra.nixos.org/jobset/nixos/release-16.09:

---

• [ ] nixos.tests.grsecurity.i686-linux
  • maintainers: @copumpkin, @joachifm
• [x] nixos.tests.mathics.i686-linux
  • maintainers: @benley
• [x] nixos.tests.mathics.x86_64-linux
  • maintainers: @benley
• [x] nixos.tests.networking.networkd.macvlan.i686-linux
  • maintainers: @wkennington
• [x] nixos.tests.networking.networkd.macvlan.x86_64-linux
  • maintainers: @wkennington
• [ ] nixos.tests.pumpio.i686-linux
  • maintainers: @rvl
• [ ] nixos.tests.virtualbox.net-hostonlyif
  • maintainers: @aszlig, @wkennington
• [ ] nixpkgs.amdappsdkFull.i686-linux
  • maintainers: @offlinehacker
• [ ] nixpkgs.amdappsdkFull.x86_64-linux
  • maintainers: @offlinehacker
• [ ] nixpkgs.linuxPackages_chromiumos_3_14.ena.x86_64-linux
  • maintainers: @edolstra
• [ ] nixpkgs.linuxPackages_chromiumos_3_14.virtualboxGuestAdditions.x86_64-linux
  • maintainers: @sander
• [x] nixpkgs.linuxPackages_chromiumos_3_18.lttng-modules.x86_64-linux
  • maintainers: @bjornfor
• [x] nixpkgs.linuxPackages_chromiumos_latest.lttng-modules.x86_64-linux
  • maintainers: @bjornfor
• [ ] nixpkgs.linuxPackages_grsec_nixos.virtualboxGuestAdditions.i686-linux
  • maintainers: @sander
• [ ] nixpkgs.linuxPackages_grsec_nixos.virtualboxGuestAdditions.x86_64-linux
  • maintainers: @sander
• [x] nixpkgs.mysqlWorkbench.x86_64-linux
  • maintainers: Karn
• [x] nixpkgs.x42-plugins.x86_64-linux
  • maintainers: @magnetophon

cc @NixOS/nixpkgs-committers

[dezgeg]

go-repo-root is fallout from my changes. I will tackle that.

[shlevy]

Fixed phc-intel stuff in https://github.com/NixOS/nixpkgs/commit/b3871e4b7bd64fd82f5353ad1e70c773eddfd42c

[kamilchm]

It looks like Mesos errors are caused by readdir_r deprecation in libc -> https://sourceware.org/ml/libc-alpha/2016-02/msg00093.html https://issues.apache.org/jira/browse/MESOS-6013

[dezgeg]

Can anyone locally reproduce these: pkgsi686Linux.librem pkgsi686Linux.sysklogd sysklogd? They all work for me.

[zimbatm]

^ got a patch for packer if interested

[kamilchm]

^ Mesos patches backported from 1.0

It may be worth to include latest bugfixes for Mesos and Marathon in 16.09:

• https://github.com/NixOS/nixpkgs/pull/18220
• https://github.com/NixOS/nixpkgs/pull/18221

[garbas]

for nixpkgs.linuxPackages_4_7.spl.* fixes are in their master but no idea if they plan any release soon.

should we include a latest master for kernels >= 4.7?

[shlevy]

@garbas Does the patch apply cleanly? If not IMO we should mark spl broken on >= 4.7 until the release

[obadz]

Fixed DBIxClass/hydra

[obadz]

@dezgeg: sysklogd build fine here too.

[shlevy]

I'll restart sysklogd build

[obadz]

@rushmorem: 940dc9b didn't seem to fix lizardfs

[rushmorem]

@obadz I'm still at work right now. Will work on it as soon as I'm free.

[obadz]

Fixed netatalk

[rushmorem]

lizardfs should be fixed now. I did not have nix.useSandbox enabled in my /etc/nixos/configuration.nix.

[shlevy]

Looks like there was a significant rebuild in the latest eval, I cancelled non-current builds to give hydra some breathing room.

[obadz]

Looks like minc_tools tests have never passed. cc @bcdarwin original contributor via 69c628e.

[domenkozar]

I've updated the issue with the new failing builds.

[rycee]

Fixed nixpkgs.trigger.i686-linux and nixpkgs.trigger.x86_64-linux in 8e3ea885e09a261b7d54c8a609f535e4dfc18290.

[rycee]

Also did nixpkgs.teeworlds in 69ac966f95ad117a263ef9084b5cad1c3cfccccc.

[lsix]

I have been to build pypy just fine locally from the release-16.09 branch, can someone try to relaunch the hydra build?

$ nix-build -A pypy
/nix/store/kakm6a50q46907w32ij4z3bda303mnh3-pypy-5.3.1
$ /nix/store/kakm6a50q46907w32ij4z3bda303mnh3-pypy-5.3.1/bin/pypy --version
Python 2.7.10 (7e8df3df96417c16c2d55b41352ec82c9c69c978, Sep 03 2016, 07:01:29)
[PyPy 5.3.1 with GCC 5.4.0]

[vcunat]

Restarted now, but over the long term, it would be nice if the tests were more reliable. Pypy tests tend to break relatively often, and that's no good as it takes so long to build.

[garbas]

@lancelotsix @vcunat do we know which tests from pypy are not reliable? i think this would be worth reporting upstream.

[FRidh]

I updated pypy and julia and both build

[vcunat]

x42-plugins has been repeatedly succeeding and failing for a long time, even on 16.03 http://hydra.nixos.org/job/nixos/release-16.03/nixpkgs.x42-plugins.i686-linux/all I never managed to find out how that could happen.

[garbas]

@domenkozar spl on kernels >= 4.7 are now marked as broken until new spl release is there. you want us to check off stuff from the list or you want to do it?

[vcunat]

I was check-boxing myself on that list, for stuff I fixed or verified to build.

[joachifm]

All the dpdk errors are the same. There's an update available, testing the build now.

[FRidh]

i686 version of julia still fails.

line 118: ErrorException("error compiling primes: could not load library "libopenlibm"

even though it is included.

[obadz]

nixpkgs.tarball broken by 5ce6cd0 cc @svanderburg

[domenkozar]

I've updated the issue with latest failing builds from http://hydra.nixos.org/eval/1290819

There is a new evaluation pending, which should fix some tests.

cc @peti for haskell packages.

[domenkozar]

Fixed tarball in ad771764743d9645084ac21730bb5a6cc11ce672 /cc @fpletz

[domenkozar]

Last blocker to get 16.09 channels is OVA generation, I'm suspecting the following change: https://github.com/NixOS/nixpkgs/pull/17822#issuecomment-244598190

[Mic92]

Regarding nixos.tests.ferm.i686-linux:

This is only broken on 32bit not x86_64: Is there a way to build a 32bit integration test on a 64bit installation?

[Mic92]

Is there a way to automatically inform all maintainers of these packages? I only saw this by a accident.

[domenkozar]

@Mic92 I'm working on a script to figure out when the build(s) started failing and who to really ping.

[domenkozar]

@Mic92 $ nix-build nixos/release.nix -A tests.ferm.i686-linux

[vcunat]

@domenkozar: doesn't Hydra do that with meta.maintainers already?

[domenkozar]

@vcunat if we had a strict rule that each module/pkgs would have meta.maintainers, that would work, yes. Otherwise I usually have to pin down the commit, just like mention bot does.

[vcunat]

Forcing people wouldn't help much, I think.

[domenkozar]

I fully agree, I'm just saying that meta.maintainers is partially useful here.

[Mic92]

ok, maybe I was not informed about this breakage by hydra, because I added the integration test after submitting the package.

[aristidb]

ispc should be fixed now. It's merged to 16.09, but not yet evaluated by Hydra.

[Mic92]

nixos.tests.ferm.i686-linux will be fixed with: https://github.com/NixOS/nixpkgs/pull/18288

[lsix]

nixpkgs.salut_a_toi fixed in c47eb4a43a9ab338c26b6b05d59ad0cb90507ed1

[domenkozar]

I've opened a separate issue to address slow webkitgtk24 build: https://github.com/NixOS/nixpkgs/issues/18312

[obadz]

gnome.gnomepanel no longer builds, probably due to staging merge 8c4aeb1, http://hydra.nixos.org/build/39585120/nixlog/1#line-2564 cc @dezgeg

[lsix]

nixpkgs.netcdf-mpi fixed in 396ffac28f948ebe4864551c67a01d1828e22ba6

[domenkozar]

I've updated the failing builds, they now include maintainers.

[shlevy]

@domenkozar Poor @eelco getting saddled with @edolstra's packages!