please check/test signal_handlers2 branch
- removed the sigsegv handler.. ;)
- it's rebased to the latest master
I'd like to test this bit more than other changes.. ;)
I'll make more tests end probably commit early next week
any feedback is appreciated
thanks,
jirka
On Tue, May 24, 2011 at 01:41:01PM +0200, Jiri Olsa wrote:
On Tue, May 24, 2011 at 02:31:10PM +0300, Akos PASZTORY wrote:
> On Tue, May 24, 2011 at 1:00 PM, Jiri Olsa <jolsa(a)redhat.com> wrote:
> >> 3. why the SIGSEGV handler?
> > just saw it in application I was stealing this from ;)
> > no special reason really, will need to look around
> > and think if we can do something smart, like display
> > bug report info or something.. ideas?
>
> It is usually not a good idea to do anything after a SEGV, I just
> wouldn't install a handler at all.
>
> > How about those 5 seconds to wait after SIGTERM is
> > sent to a child.. less/more/configurable?
>
> Hm, I wonder if that's needed at all... AFAIK if you hit ^C then
> everyone in the foreground process group gets a SIGINT.
well, if latrace is killed then it takes down the child
if the child terminates immediatelly there's no delay,
but if the child does not endup gracefully and handles
SIGTERM/SIGINT, within 5 seconds latrace sends SIGKILL
>
> > It has already started to annoy me in the test run ;)
this was related to the testing of those 5 seconds
explained above
jirka
>
> I haven't tested it but maybe killing the whole process group (kill
> -INT -$latrace_pid) would help.
> _______________________________________________
> latrace mailing list
> latrace(a)lists.fedorahosted.org
>
https://fedorahosted.org/mailman/listinfo/latrace
_______________________________________________
latrace mailing list
latrace(a)lists.fedorahosted.org
https://fedorahosted.org/mailman/listinfo/latrace