Re: defect for the signal usage of SIGINT in Ruby 1.9 implementation

Hi,

At Thu, 6 Mar 2008 18:52:14 +0900,
Chirag Mistry wrote in [ruby-core:15809]:
> per our understanding, it may create problem when ruby interpreter is
> embedded in a process for a limited period of time. It should not install
> signal handler SIG_IGN for SIGINT signal.

It would be true, when embedded, but might not be true
otherwise.


Index: signal.c
===================================================================
--- signal.c	(revision 15720)
+++ signal.c	(working copy)
@@ -1004,8 +1004,25 @@ ruby_sig_finalize()
     sighandler_t oldfunc;
 
-    oldfunc = ruby_signal(SIGINT, SIG_IGN);
+#ifdef POSIX_SIGNAL
+    struct sigaction oldact;
+    MEMZERO(&oldact, struct sigaction, 1); 
+    sigaction(SIGINT, NULL, &oldact);
+#ifdef SA_SIGINFO
+    if (oldact.sa_flags & SA_SIGINFO)
+	oldfunc = (sighandler_t)oldact.sa_sigaction;
+    else
+#else
+	oldfunc = oldact.sa_handler;
+#endif
     if (oldfunc == sighandler) {
 	ruby_signal(SIGINT, SIG_DFL);
     }
+#else
+    oldfunc = ruby_signal(SIGINT, SIG_IGN);
+    if (oldfunc == sighandler) {
+	oldfunc = SIG_DFL;
+    }
+    ruby_signal(SIGINT, oldfunc);
+#endif
 }
 


-- 
Nobu Nakada