On Wed, Dec 28, 2016 at 10:17:31AM +0100, John Darrington wrote:
> On Tue, Dec 27, 2016 at 11:13:05PM -0500, Leo Famulari wrote:
>      On Tue, Dec 27, 2016 at 06:56:44PM -0500, Leo Famulari wrote:
>      > On Fri, Dec 23, 2016 at 12:23:39AM -0500, Leo Famulari wrote:
>      > > The last staging branch was merged into the master branch on December
>      > > 11.
>      > > 
>      > > According to the 3 week staging cycle we are trying to adhere to [0],
>      > > this would mean we freeze the current branch on December 25 and try to
>      > > merge it on January 1.
>      > > 
>      > > These dates are inconvenient for many people. I propose we freeze the
>      > > branch and start building on December 27, and try to merge on January 3.
>      > 
>      > And the build failures are rolling in :)
>      > 
>      > https://hydra.gnu.org/eval/109410?full=1&compare=master#tabs-now-fail
>      
>      The 'qemu-image' build failed due to a lack of disk space, presumably on
>      the VM image file:
>      
>      ------
>      registering closures...
>      error: setting journal mode: database or disk is full
>      ERROR: In procedure scm-error:
>      ERROR: failed to register store items "/xchg/system"
>      ------
>      https://hydra.gnu.org/build/1741220
>      
>      Any advice? I'm not sure where to start looking in the code.
> 
> For what it's worth, I find that building *anything* related to qemu fails
> for different reasons in about 2/3 attempts. So I think there is a 
> fundamental, yet to be diagnosed problem.

I've seen this error before when I fail to create a large enough blank
QEMU image. I think that's the problem here, but I'm not sure where this
qemu-image derivation comes from in the Guix code base, so I'm not sure
what to try adjusting or even how to replicate the failure.