On Fri, Jul 17, 2009 at 12:18:35PM -0400, Anurag S. Maskey wrote: > Did anyone notice that in the 7/13 build, when ENM scripts complete, > nwamd is terminated. svc.configd subsequently start nwamd again. > > Looking at the debug logs, I see that the signal handler catches > SIGCHLD. since we don't specify how to handle this signal, it > terminates nwamd with graceful_shutdown(). Previously, no such SIGCHLD > was caught by nwamd.
> Discussion on the IRC #onnv lead me to CR "6857047 sigwait malfunctions > for system-generated SIGCHLD signals". It seems that SIGCHLD was > improperly being ignored. There also a related CR "6861646 nwamd > terminates prematurely in snv_119" for Phase 0.5 since it has the > similar signal handling code as our gate. Yep, you've tracked all this down the hard way! This is the 0.5 bug Michael mentioned in our meeting yesterday. David Comay observed problems with nwam while working on the build 119-based opensolaris release. He filed a bug on it this morning, the one you found. I've coded up a fix that David has verified for the current onnv bits; I'll merge that into the phase 1 gate as well. If you're interested, check out (and comment on!) the onnv code review: file:///net/muskogee.eng/zspace/hg/bugs/webrev/index.html -renee > > What's the right thing to do here? Catch SIGCHLD in main.c:sighandler() > and then don't do anything?? > > Anurag > > http://monaco.sfbay/detail.jsf?cr=6857047 > http://monaco.sfbay/detail.jsf?cr=6861646 > > > _______________________________________________ > nwam-dev mailing list > nwam-dev at opensolaris.org > http://mail.opensolaris.org/mailman/listinfo/nwam-dev
