buildworld -j N issues

[David Rhodus]

On Mar 21, 2004, at 5:33 PM, David Cuthbert wrote:
Matthew Dillon wrote:
    Some issues are cropping up with what I believe are buildworld -j 
N
    bugs, due to the new scheduler ordering processes somewhat 
differently
    then the old scheduler.
Hm... I think the problem is more serious than that.  My builds are 
failing on any sleep statement (that is, /bin/sleep, the executable).

Indeed, '/bin/sleep 1' causes it to go off into lala-land, and truss 
is of no help.

The following test executable waits forever:

#include <unistd.h>
int main() { sleep(1); return 0; }
I am running on VMware, though.
David,

I think the problems you are seeing might be fixed in the
latest kernel as there was a small fall over still from the
timer changes. This was affecting the operational priority
values of programs because of time slice priories were
a little mixed up. Can you try and update to the latest kernel
build and see if that fixes the problem your seeing...
-DR