[Sbcl-devel] restoring debuggability on ppc

[Cyrus H. <ch...@bo...>]

Dear sbcl-devel,

In 0.9.5.76, we now test whether or not nodefer works using a  
SIGTRAP. While this works fine under normal circumstances on Mac OS,  
it breaks GDB. Gabor suggested the obvious fix of changing the signal  
under Darwin and this patch does that. It is a rather trivial and  
obvious patch, but it restores the ability to debug sbcl using gdb  
under Mac OS. I should offer the caveat that I really don't  
understand signals that well and don't know if this restores the  
intended behavior of the test, but it, at least, restores my ability  
to run sbcl using gdb.

Thank you for considering this patch,

Cyrus

Index: sbcl/src/runtime/interrupt.c
===================================================================
RCS file: /cvsroot/sbcl/sbcl/src/runtime/interrupt.c,v
retrieving revision 1.102
diff -u -r1.102 interrupt.c
--- sbcl/src/runtime/interrupt.c        17 Oct 2005 10:58:30  
-0000      1.102
+++ sbcl/src/runtime/interrupt.c        20 Oct 2005 03:09:18 -0000
@@ -1075,8 +1075,13 @@
      sa.sa_sigaction = sigaction_nodefer_test_handler;
      sigemptyset(&sa.sa_mask);
      sigaddset(&sa.sa_mask, SIGABRT);
+#ifdef LISP_FEATURE_DARWIN
+    sigaction(SIGUSR1, &sa, NULL);
+    kill(getpid(), SIGUSR1);
+#else
      sigaction(SIGTRAP, &sa, NULL);
      kill(getpid(), SIGTRAP);
+#endif
      while (sigaction_nodefer_works == -1);
}