Incomplete output
Following fd4e2721, on some architectures (depending on Python version?) the output is incomplete.
In vidjil
, this is the case for vidjil-h-examples.should-get
or stanford-k14.should-get
where the log outputs are not complete on some slaves (CentOS 6.3 amd64, Fedora 25 amd64, FreeBSD). Moreover those tests still timeout on those architectures.
Would a flush of the output be required?