Hi,
someone noticed that the stats option sometimes gives bogus results. I've checked it and found a bug in tv_sub: it added 1 extra second whenever the two arguments referred the same point in time. While it's possible to fix it, I think it's better to use the already existing timersub and timeradd macros, so I went for that.
The second patch is a bit questionable, it replaces 'nan' in the output with zero. You can decide if it makes sense...
Akos PASZTORY (2): stats: bugfix: use timersub() and timeradd() stats: report zero percents instead of NaN
src/stats.c | 44 ++++++++------------------------------------ 1 files changed, 8 insertions(+), 36 deletions(-)