#42 linux-abi-2.6.17-git11
Hi All,
on the way to 2.6.18, something weird happens between
2.6.17-git12 and 2.6.17-git13, probably related to
rework of scheduling policyes and mutex-futex-rt_mutex.
Most of legacy code still runs, but - at least in my
environment - fork-exec-wait cycles terminate the
father process which receives a signal 11.
Earlier, as I realized now, 2.6.17 was much less
stable in running something like
loop ;
system "somecommand" ;
forever ;
which lasts ~ 70-80 cicles before exiting with "cannot
create child" error. Again, in 2.6.17 the scheduler
got reworked. As soon as I find some time, I'm trying
to figure out something out of the trace logs.
Please, take some time to give a try 2.6.17, 2.6.17-
git12, 2.6.17-git13 and 2.6.18-rc (patch follows), to
check whether the same behaviour occurs in your
environments.
Regards
Mauro Tassinari
linux-abi-2.6.17-git11.diff.bz2
Logged In: YES
user_id=80373
forgot: same cicle is absolutely stable in 2.6.16