I think I have tracked the problem to libev. Kudu uses libev for its reactor, and libev installs a signal handler for SIGCHLD (http://pod.tst.eu/http://cvs.schmorp.de/libev/ev.pod): > Libev grabs SIGCHLD as soon as the default event ...cout (Paul Brannan)
As far as I can tell looking at the kudu source, none of the threads created by the kudu library installs a SIGCHLD handler. But this certainly feels related, because `system` behaves as expected if I ignore SIGCHLD prior to calling it....cout (Paul Brannan)
I have a very simple client for kudu (https://kudu.apache.org/) implement as a wrapper that I invoke with FFI. After creating a kudu client with `KuduClientBuilder::Build()`, subsequent calls to `system` cause the ruby interpreter to bl...cout (Paul Brannan)
On Wed, 2012-04-25 at 05:24 +0900, Roger Pack wrote: > >> It always confuses me how that one defies the rule. > > > > There’s no non-dangerous version of Array#shift, so it doesn’t defy the rules. > > True, I just always reac...cout (Paul Brannan)
On Fri, 2012-04-13 at 22:58 +0900, mame (Yusnuke Endoh) wrote: > In the same manner, a "nested" method definition you said: > > def define_foo > def foo > end > end > > should define a method named "define_...cout (Paul Brannan)