JDK-4377267 : JVM_handle_solaris_signal should not deinstall signal handler
  • Type: Bug
  • Component: hotspot
  • Sub-Component: runtime
  • Affected Version: 2.0
  • Priority: P4
  • Status: Closed
  • Resolution: Duplicate
  • OS: solaris_7
  • CPU: sparc
  • Submitted: 2000-10-06
  • Updated: 2001-03-29
  • Resolved: 2001-03-29
Related Reports
Duplicate :  

Name: cl74495			Date: 10/06/2000

If JVM_handle_solaris_signal decides that a signal
is unexpected, then it deinstalls the handler before
calling handle_unexpected_exception

If several threads hit the same signal-causing failure at the
nearby in time, then the second thread to execute JVM_handle_solaris_signal
will cause an immediate coredump since the handler has been
deinstalled.  So the ShowMessageBoxOnError logic won't work, etc.

In our builds we have disabled this code in
JVM_handle_solaris_signal :

  // Remove this handler now.  It is of no more use.
  os::Solaris::set_signal_handler(sig, false);

This code only applies to a debug build.
(Review ID: 110474) 

Closing as a duplicate of signal chaining bug. 4363638

gary.collins@East 2001-03-29

EVALUATION This looks to be the same problem as all the other signal chaining bug will fix. Asked Allen to run with -XX:+AllowUserSignalHandlers option If this works I will close this as a duplicate of bug 4363638 gary.collins@East 2001-03-20 Allen, Could you try running with this flag. We have signal handling fixed in later JDK's 1.3.1 and 1.4 java -XX:+AllowUserSignalHandlers yourProgram or add it to your createJVM c code. Let me know if this works for you. This is not a fix, rather a test and should not be used for prime time use. In later hotspot release this is fixed. I just wanted to make sure we are talking about the same issue. Regards, Gary Collins From ###@###.### Mon Mar 26 19:54:20 2001 From: Allen Otis <###@###.###> To: Gary Collins - SunSoft Engineering <###@###.###> Subject: Re: BugId 4377267 : (P4/S5) gary.collins@east, Now Responsible Engineer Mime-Version: 1.0 From the source code changes that I see associated with AllowUserSignalHandlers, I expect your changes will fix bug 4377267 and some other issues we had related to applications wanting to install handlers for SEGV, etc . It will be a week or two before I can run with -XX:+AllowUserSignalHandlers since I need to take out our changes to signal handling and merge in the AllowUserSignalHandlers code . Allen Awaiting to hear from Allen if this truely fixes his problem. gary.collins@East 2001-03-27