On Thu, Jan 29, 2015 at 09:54:57PM +0100, tomazzi wrote: > To summarise: there are no bugs in core/main.c: > > 1. Ignoring return value from sigaction is not a problem, because > it's almost *unlikely* to fail. > In any case, kernel will do the job, so it doesn't even matter > whether the handlers are installed or not -> everything that can be > done in the handler is to just quit anyway. Actually the signal handler does sleep in an infinite loop, to not crash the system.
> > 2. SIGSEGV handler (and others): it's *unlikely* to happen that the > handler will be executed at the bottom of the main stack (with > insufficient stack space or when the stack is already overflowed) - > because stacks are huge, and in case of segfault see point 1. > > 3. SA_NODEFER is OK, cause it's *unlikely* that external signal gets > delivered. Even if the crash() handler will crash (f.e. because > it's not re-entrant), what (again) is *unlikely* to happen, see > point 1. OK, great to hear that what we have currently should work :) Zbyszek > > I've learned a lot, sorry to waste Your time. > > Regards. > > > > _______________________________________________ systemd-devel mailing list [email protected] http://lists.freedesktop.org/mailman/listinfo/systemd-devel
