Failed build messes up printing

[pv]

It appears sometimes a build failure results to clobbered print output:

git clone https://github.com/pv/scipy-bench.git scipy-bench
cd scipy-bench
git checkout origin/param
./run.py run -n master^! # to make benchmark discovery work
./run.py run -n e80cb04^^..e80cb04

results to the following output from asv: https://gist.github.com/pv/7b73f66f404f58703ba7

... huge  build log output ...
...
 error: Command "x86_64-linux-gnu-gcc -pthread -DNDEBUG -g -fwrapv -O2 -Wall [...] special/_cephesmodule.o" failed with exit status 1

[ 0.00%] ·· Skipping py2.7-Cython0.21.1-Tempita0.5.2-numpy1.8.2-six
[ 1.09%] ··· Benchmark fftpack_basic.RFft.time_rfft skipped
[ 2.17%] ··· Benchmark sparse.Construction.time_construction skipped 
...
[ 48.91%] ··· Benchmark fftpack_basic.Fftn.time_fftn skipped
[ 50.00%] ··· Benchmark sparse_linalg_expm.ExpmMultiply.time_expm_multiply skipped... failed failed failed failed failed failed failed failed failed 

The first build fails (as it should, buggy commit), but the output from the second build command becomes strange --- it's as if the asv logger stops producing output, after printing out the skipped benchmark messages. The only thing that is printed are apparently the log.add stuff which goes directly to stdout.

Didn't manage to reproduce this yet in a simpler case and smaller repository (sorry)...

[pv]

This was actually tested with gh-203 present, not against master

[pv]

Ok, this occurred only with gh-203, fixed by pv/asv@044b8d6f9c