Move warning about too many pids to error path #834
Conversation
| kill -s KILL $pid 2>/dev/null | ||
| return 2 | ||
| pid_to_kill=$pid | ||
| # else |
There was a problem hiding this comment.
Maybe add an explanation of why there are other processess running?
There was a problem hiding this comment.
I don't think we actually know, but we will emit a message on line 298 if we don't find a pid to kill.
portante
force-pushed
the
fix-tool-warnings
branch
from
29da9a9 to
9727034
Compare
| return 2 | ||
| fi | ||
|
|
||
| if [[ $len > 4 || ($len > 3 && $tool != "turbostat") ]] ;then |
There was a problem hiding this comment.
Should there be a comment here that explain why there is a special case for turbostat?
There was a problem hiding this comment.
Yes there should be. Does anybody know the history of that check?
There was a problem hiding this comment.
Sorry - I don't remember. I'll try to instrument safer_kill and see why turbostat seems to behave differently from everything else.
There was a problem hiding this comment.
I get 3 pids from pidof -x turbostat and also 3 pids from pidof -x iostat. The only difference is that there is a turbostat-datalog between the pbench-tool turbostat call and the "real" one, but pidof does not count the datalog process. So far, there does not seem to be a reason for treating turbostat differently.
There was a problem hiding this comment.
Okay, can we track this as a separate piece of work in issue #887, and not make further changes here, since this PR did not change that behavior?
3 participants
No description provided.