United StatesChange Country, Oracle Worldwide Web Sites Communities I am a... I want to...
Bug ID: 6260293 fix set_ctrl() inconsistencies in loopopts
6260293 : fix set_ctrl() inconsistencies in loopopts

Details
Type:
Enhancement
Submit Date:
2005-04-22
Status:
Resolved
Updated Date:
2010-12-08
Project Name:
JDK
Resolved Date:
2005-08-26
Component:
hotspot
OS:
solaris,generic
Sub-Component:
compiler
CPU:
sparc,generic
Priority:
P4
Resolution:
Fixed
Affected Versions:
5.0u12,6
Fixed Versions:
6

Related Reports
Backport:
JDK-2174003 - fix set_ctrl() inconsistencies in loopopts
Backport:
JDK-2158032 - fix set_ctrl() inconsistencies in loopopts
Duplicate:
JDK-6644519 - 5u12 64bit fails with SIGBUS in interpreted code
Relates:
JDK-6659207 - access violation in CompilerThread0
Relates:
JDK-6435614 - code fails with impossible ArrayIndexOutOfBounds Exception
Relates:
JDK-6548082 - bug in C2 conditional-move - suggested fix is provided
Relates:
JDK-6233005 - Random javac failures on server VM

Sub Tasks

Description
In loopopts, we sometimes use set_ctrl() and friends to associate a node with a control that we wouldn't allow as in(0). For example, we may call set_ctrl() with an IfNode.  

Such sloppiness can cause surprising results downstream where we compare in(0) against get_ctrl() and do a particular transformation based on that comparison.

Such loose play lead to problems such as found in 6233005.
###@###.### 2005-04-22 21:40:23 GMT

Comments
SUGGESTED FIX

See PRT webrev:

http://analemma.sfbay.sun.com/net/prt-archiver.sfbay/data/archived_workspaces/main/c2_baseline/2005/20050815102757.rasbold.c2_baseline2/workspace/webrevs/webrev-2005.08.15/index.html
2005-08-15 
EVALUATION

Add assert in loopnode.hpp wherever we call _nodes.map.  Let the asserts be our
guide in finding bad ctrl settings.


###@###.### 2005-04-22 21:46:19 GMT
2005-04-22


Hardware and Software, Engineered to Work Together