Re: Another CFG problem.

[Feng QIAN <fqian@sable.mcgill.ca>]

A quick hack is to change two lines code, but it will trigger a bug in
jasmin parser. Changes in jasmin parser are necessary (see below).

> (Merge-Parents)
Index: 1.2.2.dev.41/src/soot/coffi/CFG.java
868c868,869
< 		Instruction newdefault = (Instruction)replacedInsns.get(switchinsn.default_inst);
---
> 		Instruction newdefault =
> 		    (Instruction)replacedInsns.get(switchinsn.default_inst);
874,875c875,876
<
< 		for (int i=0; i<switchinsn.match_insts.length; i++)
---
>
> 		for (int i=0; i<switchinsn.npairs; i++)
877c878,879
< 		    Instruction newtgt = (Instruction)replacedInsns.get(switchinsn.match_insts[i]);
---
> 		    Instruction newtgt =
> 			(Instruction)replacedInsns.get(switchinsn.match_insts[i]);
898c900
< 		for (int i=0; i<switchinsn.jump_insts.length; i++)
---
> 		for (int i=0; i<switchinsn.high-switchinsn.low; i++)
900c902,903
< 		    Instruction newtgt = (Instruction)replacedInsns.get(switchinsn.jump_insts[i]);
---
> 		    Instruction newtgt =
> 			(Instruction)replacedInsns.get(switchinsn.jump_insts[i]);

Temporarily fix-up in Jasmin parser :

jasmin/src/jasmin/parser.cup

lookup ::=
    lookup_args
    lookup_list_t
    lookup_default
  ;

lookup_list_t ::=
     | lookup_list
  ;


table ::=
    table_args
    table_list_t
    table_default
  ;

table_list_t ::=
     | table_list
  ;

====================
Feng Qian					fqian@sable.mcgill.ca

On Fri, 5 Oct 2001, Stephen Andrew Neuendorffer wrote:

> The code below:
>
> package ptolemy.copernicus.java.test;
>
> public class test {
>      public static int s[] = new int[5];
>
>      public static final int foo() {
>          int i = 1;
>          int startsAt = 0;
>
>          for (;;) {
>          MatchLoop:
>              do {
>                  switch(s[--i]) {
>                  default : break;
>                  }
>              } while(i != startsAt);
>          }
>      }
> }
>
> compiled into the attached class file throws an exception:
>
> Transforming ptolemy.copernicus.java.test... java.lang.NullPointerException
>          at soot.coffi.CFG.adjustBranchTargets(CFG.java:735)
>          at soot.coffi.CFG.eliminateJsrRets(CFG.java:456)
>          at soot.coffi.CFG.<init>(CFG.java:85)
>          at soot.coffi.CoffiMethodSource.getBody(CoffiMethodSource.java:87)
>          at soot.SootMethod.getBodyFromMethodSource(SootMethod.java:83)
>          at soot.SootMethod.retrieveActiveBody(SootMethod.java:276)
>          at soot.Main.handleClass(Main.java:1795)
>          at soot.Main.run(Main.java:1557)
>          at java.lang.Thread.run(Thread.java:484)
>
> It appears that the switch is not being handled properly?
> This is a stripped down example of some code generated using JavaCC using
> the standard soot 1.2.2 and jdk1.3.1
>
> Steve