We assume that we only need to write a CSV header into an empty file.
Thus, we can check whether the file already contains data.
If that's the case, we only write the metadata, but not the CSV header.
This PR also includes two other minor bits:
it fixes the rebench.conf file to work without error (test-vm1.py was changed in 2018 to take 5 arguments)
in some strange conditions, the return_code from running a benchmark command can be None, which breaks output, so, None needs to be handled separately
coverage: 83.079% (+0.05%) from 83.034% when pulling feebeb611f4409ebcba065a60e3393281a69159b on single-csv-header-per-file into cd302d5155c73575c2b158034afcde07d2403c6d on master.
This fixes #223.
We assume that we only need to write a CSV header into an empty file. Thus, we can check whether the file already contains data. If that's the case, we only write the metadata, but not the CSV header.
This PR also includes two other minor bits:
return_code
from running a benchmark command can beNone
, which breaks output, so,None
needs to be handled separately@naomiGrew please have a look, and review the PR