Review request: 8000230: Change os::print_location to be more descriptive when a location is pointing into an object

Stefan Karlsson stefan.karlsson at oracle.com
Fri Sep 28 13:49:52 UTC 2012


http://cr.openjdk.java.net/~stefank/8000230/webrev/

 From the bug:
---
When printing the "Register to memory mapping:" hs_err section, it's 
easy to get fooled when a register points into an object.

For example:
R13=0xfffffd7fcf900064 is an oop
[Ljava.lang.Object;
  - klass: 'java/lang/Object'[]
  - length: 327680

seems to indicate that an object starts at address 0xfffffd7fcf900064. 
In this case the real object start was at address 0xfffffd7fcf900000.

I propose that we change this line to:
R13=0xfffffd7fcf900064 is pointing into object: 0xfffffd7fcf900000
[Ljava.lang.Object;
  - klass: 'java/lang/Object'[]
  - length: 327680

If the address seems to be pointing at the start of an object, the old 
output can still be used.
---

StefanK



More information about the hotspot-gc-dev mailing list