JDK-4027933 : Native stack overflows not detected or handled correctly
  • Type: Bug
  • Component: hotspot
  • Sub-Component: runtime
  • Affected Version: 1.1
  • Priority: P5
  • Status: Closed
  • Resolution: Duplicate
  • OS: solaris_2.4
  • CPU: sparc
  • Submitted: 1997-01-24
  • Updated: 1999-04-09
  • Resolved: 1999-04-09
Related Reports
Duplicate :  
JDK-4217960 - [native stack overflow bug] reflection test causes crash
Description
The attached program causes a SEGV.  It appears that when the stack
overflows the exception is not handled properly.  In interpreter.c
at handle_exception it always expects an exception object.  However,
when its a stack overflow there is no object.

To reproduce:
% untar the sources
% javac -d . *.java
% rmic -d . TestRMI.TestRMIImpl
% rmiregistry &
% java TestRMI.TestRMIImpl &  # This is the one that SEGV's
% java TestRMI.TestRMIClient &
Comments
EVALUATION Downgrading because this is a native stack overflow, which we cannot fix in general without help from C. It is still the case that the interpreter fails to do something sensible with native stack overflows that it does catch. The other half of this problem is that RMI is doing deep Java->C->Java recursions in the example, which will chew up large amounts of native stack and leave itself open to native stack overflow. That seems an unfortunate way to implement things. timothy.lindholm@Eng 1997-01-27 I changed the Synopsis to one specifying the native stack overflow problem, and will report a second bug against RMI's recursive implementation. timothy.lindholm@Eng 1997-02-10
27-01-1997