when trying to improve top I was looking into ps's output when using optio=
n -S and found with the help of alexh that the numbers reported back or no=
t what is described in man page:
-S : Change the way the process time is calculated by summing all exited c=
hildren to their parent process
This seems due to the fact that for summing up childrens cpu time kinfo=5Fpr=
oc.kp=5Fcru is used in ps/print.c :
443 if (sumrusage) {
444 secs +=3D KI=5FPROC(k, cru).ru=5Futime.tv=5Fsec +
445 KI=5FPROC(k, cru).ru=5Fstime.tv=5Fsec;
446 psecs +=3D KI=5FPROC(k, cru).ru=5Futime.tv=5Fusec +
447 KI=5FPROC(k, cru).ru=5Fstime.tv=5Fusec;
448 }
scanning through DF sources in http://grok.x12.su gives me the impression =
that kinfo=5Fproc.kp=5Fcru is not filled anywhere.
Other BSDs use similar approach for ps -S (fbsd: ki=5Fchildtime, obsd: p=5Fuct=
ime=5Fsec), but these contain actual data (fbsd: kern=5Fproc.c,obsd: kvm=5Fproc.=
c).
So the question is, is this a ps bug and we should remove option -S (or ad=
d note that this doesn't work on DF) or should we actually will kp=5Fcru wit=
h meaningful data=3F
Jan=20
=5F=5F=5F=5F=5F=5F=5F=5F=5F=5F=5F=5F=5F=5F=5F=5F=5F=5F=5F=5F=5F=5F=5F=5F=5F=5F=5F=5F=5F=5F=5F=5F=5F=5F=5F=5F=5F=5F=5F=5F=5F=5F=5F=5F=5F=5F=5F=5F=5F=5F=5F=5F=5F=5F=5F=5F=5F=5F=5F
Preisknaller: WEB.DE DSL Flatrate f=FCr nur 16,99 Euro/mtl.!=20
http://produkte.web.de/go/02/