[jboss-jira] [JBoss JIRA] (JASSIST-158) Method calls with a parameter of long[] or double[] are compiled to incorrect bytecode resulting in errors such as java.lang.VerifyError: ... Inconsistent args_size for opc_invokeinterface

[Jens Deppe (JIRA)]

     [ https://issues.jboss.org/browse/JASSIST-158?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel ]

Jens Deppe updated JASSIST-158:
-------------------------------

    Description: 
I believe this patch fixes the problem:

Method calls with a parameter of long[] or double[] are compiled to incorrect bytecode. For example, the following piece of code...

{code}
long[] aLongArray;
PdxWriter writer;
...
writer.writeLongArray("aLongArray", aLongArray);
{code}

Produces code similar to this:

{code}
   13:  aload_1
   14:  ldc     #32; //String aLongArray
   16:  aload_0
   17:  getfield        #35; //Field aLongArray:[J
   20:  invokeinterface #39,  4; //InterfaceMethod com/gemstone/gemfire/pdx/PdxWriter.writeLongArray:(Ljava/lang/String;[J)Lcom/gemstone/gemfire/pdx/PdxWriter;
   25:  pop
{code}

Following the code, it looks like getfield calculates an incorrect stackdepth when applied to a long[] or double[], resulting in an incorrect number of args applied to invokeinterface.

The following patch seems to fix things. 

{code}
Index: src/main/javassist/compiler/MemberCodeGen.java
===================================================================
--- src/main/javassist/compiler/MemberCodeGen.java	(revision 624)
+++ src/main/javassist/compiler/MemberCodeGen.java	(working copy)
@@ -957,7 +957,7 @@
         else
             className = null;
 
-        boolean is2byte = (c == 'J' || c == 'D');
+        boolean is2byte = ((c == 'J' || c == 'D') && dim == 0);
         return is2byte;
     }
{code}

  was:
I believe this patch fixes the problem:

Method calls with a parameter of long[] or double[] are compiled to incorrect bytecode. For example, the following piece of code...

long[] aLongArray;
PdxWriter writer;
...
writer.writeLongArray("aLongArray", aLongArray);

Produces code similar to this:

   13:  aload_1
   14:  ldc     #32; //String aLongArray
   16:  aload_0
   17:  getfield        #35; //Field aLongArray:[J
   20:  invokeinterface #39,  4; //InterfaceMethod com/gemstone/gemfire/pdx/PdxWriter.writeLongArray:(Ljava/lang/String;[J)Lcom/gemstone/gemfire/pdx/PdxWriter;
   25:  pop

Following the code, it looks like getfield calculates an incorrect stackdepth when applied to a long[] or double[], resulting in an incorrect number of args applied to invokeinterface.

The following patch seems to fix things. 

Index: src/main/javassist/compiler/MemberCodeGen.java
===================================================================
--- src/main/javassist/compiler/MemberCodeGen.java	(revision 624)
+++ src/main/javassist/compiler/MemberCodeGen.java	(working copy)
@@ -957,7 +957,7 @@
         else
             className = null;
 
-        boolean is2byte = (c == 'J' || c == 'D');
+        boolean is2byte = ((c == 'J' || c == 'D') && dim == 0);
         return is2byte;
     }



    
> Method calls with a parameter of long[] or double[] are compiled to incorrect bytecode resulting in errors such as java.lang.VerifyError: ... Inconsistent args_size for opc_invokeinterface
> --------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------
>
>                 Key: JASSIST-158
>                 URL: https://issues.jboss.org/browse/JASSIST-158
>             Project: Javassist
>          Issue Type: Bug
>    Affects Versions: 3.16.1-GA
>            Reporter: Jens Deppe
>            Assignee: Shigeru Chiba
>
> I believe this patch fixes the problem:
> Method calls with a parameter of long[] or double[] are compiled to incorrect bytecode. For example, the following piece of code...
> {code}
> long[] aLongArray;
> PdxWriter writer;
> ...
> writer.writeLongArray("aLongArray", aLongArray);
> {code}
> Produces code similar to this:
> {code}
>    13:  aload_1
>    14:  ldc     #32; //String aLongArray
>    16:  aload_0
>    17:  getfield        #35; //Field aLongArray:[J
>    20:  invokeinterface #39,  4; //InterfaceMethod com/gemstone/gemfire/pdx/PdxWriter.writeLongArray:(Ljava/lang/String;[J)Lcom/gemstone/gemfire/pdx/PdxWriter;
>    25:  pop
> {code}
> Following the code, it looks like getfield calculates an incorrect stackdepth when applied to a long[] or double[], resulting in an incorrect number of args applied to invokeinterface.
> The following patch seems to fix things. 
> {code}
> Index: src/main/javassist/compiler/MemberCodeGen.java
> ===================================================================
> --- src/main/javassist/compiler/MemberCodeGen.java	(revision 624)
> +++ src/main/javassist/compiler/MemberCodeGen.java	(working copy)
> @@ -957,7 +957,7 @@
>          else
>              className = null;
>  
> -        boolean is2byte = (c == 'J' || c == 'D');
> +        boolean is2byte = ((c == 'J' || c == 'D') && dim == 0);
>          return is2byte;
>      }
> {code}

--
This message is automatically generated by JIRA.
If you think it was sent incorrectly, please contact your JIRA administrators: https://issues.jboss.org/secure/ContactAdministrators!default.jspa
For more information on JIRA, see: http://www.atlassian.com/software/jira