bug#32511: mariadb 10.1.35 fails to build on x86_64

[ludo@gnu.org (Ludovic Courtès)]

Hi Mark,

Mark H Weaver <mhw@netris.org> skribis:

> ludo@gnu.org (Ludovic Courtès) writes:
>
>> On berlin, mariadb 10.1.35 has one test failure (this is
>> /gnu/store/a5jm2hyalyblgjnxx5x3ly4fwvfivq5m-mariadb-10.1.35.drv, as
>> found in Guix commit b0cb92b2d43a2c4d5fa9b3f8c04c5732c60061e7, for
>> instance):
>
> FWIW, after staging was merged, I also ran into a problem building
> mariadb on my x86_64 GuixSD system, where I build everything locally.
> It failed several times in a row, always with just 1 test failure.  The
> problem turned out to be that I didn't have enough free disk space.
> Buried in the mountains of output, which I found quite difficult to
> navigate (searching for "[ fail ]" helped), I eventually found a message
> about unsufficient space.  5 GB of free space wasn't enough, but ~12 GB
> was.

Indeed, I’ve built it on berlin itself with --no-build-hook (that way, I
was sure there was enough space available), and it passed.

The lesson here is that we might not be GC’ing enough on the build
machines behind berlin.  Any idea, Ricardo?

Also, I wonder if we should report it upstream so that they arrange to
skip the test when insufficient space is available.  Unfortunately we
don’t have many clues in the build log to provide them.  :-/

Thoughts?

Ludo’.