[Oorexx-bugs] [ oorexx-Bugs-1755232 ] 3.1.2 tokenized program crashes on 3.2.0 revsion 590

[SourceForge.net <no...@so...>]

Bugs item #1755232, was opened at 2007-07-16 21:28
Message generated for change (Settings changed) made by wdashley
You can respond by visiting: 
https://sourceforge.net/tracker/?func=detail&atid=684730&aid=1755232&group_id=119701

Please note that this message will contain a full copy of the comment thread,
including the initial issue submission, for this request,
not just the latest update.
Category: Interpreter
Group: 3.2.0
>Status: Closed
Resolution: Fixed
Priority: 5
Private: No
Submitted By: Mark Miesfeld (miesfeld)
Assigned to: Rick McGuire (bigrixx)
Summary: 3.1.2 tokenized program crashes on 3.2.0 revsion 590

Initial Comment:
There is an access violation on Windows running a tokenized program, tokenized with 3.1.2.

Unhandled exception at 0x6b0186e6 in rexx.exe: 0xC0000005: Access violation writing location 0xeacee9cc.

I'm attaching a small rexx program, source and the tokenized version.  (Source is TestArg.orx, tokenized version is TestArg.rex)

I just did the tokenizing a few minutes ago on a system wher I have the 3.1.2 release installed.  Then, when I execute it on a second system running 3.2.0 built from svn revision 590, it crashes. 

----------------------------------------------------------------------

Comment By: Rick McGuire (bigrixx)
Date: 2007-07-17 09:46

Message:
Logged In: YES 
user_id=1125291
Originator: NO

Committed revision 600.

----------------------------------------------------------------------

Comment By: Rick McGuire (bigrixx)
Date: 2007-07-17 06:29

Message:
Logged In: YES 
user_id=1125291
Originator: NO

Thanks for the program Mark.  I inadvertently introduced an image
compatibility with my trace enhancements.  I was storing the operator name
in a field inherited from the base class that I believed would be null in
older saved programs.  Boy, was I wrong :-).  This caused the object hash
code to get used as a reference pointer, resulting in a crash when the
program was reloaded. 

----------------------------------------------------------------------

Comment By: Mark Miesfeld (miesfeld)
Date: 2007-07-16 23:40

Message:
Logged In: YES 
user_id=191588
Originator: YES

Rick,

I did some investigation, but it will probably take your skills to explain
what the problem is.  <grin>

The access violation is in this code:

  inline void restoreObjectMark(RexxObject *markObject, RexxObject
**pMarkObject) {
                                         /* update the object reference   
   */
      markObject = (RexxObject *)((ULONG)markObject + objOffset);
      SetObjectLive(markObject);         /* Then Mark this object as live.
   */
      *pMarkObject = markObject;         /* now set this back again       
   */
  }

After markObject is set to markObject + objOffset, it is not a valid
pointer.  

If I just run the program and the debugger opens at the memory access, I
see that.  

But, if I run rexx under the debugger (devenv /debugexe rexx
.\TestArg.rex) and set a break point at the function, the break point is
never hit.  rexx runs and exits with -3.

And of course, if I retokenize the test program using the rexxc built at
the same same time as the interpreter, there is no crash, it runs fine.

----------------------------------------------------------------------

Comment By: Mark Miesfeld (miesfeld)
Date: 2007-07-16 21:28

Message:
Logged In: YES 
user_id=191588
Originator: YES

File Added: TestArg.rex

----------------------------------------------------------------------

You can respond by visiting: 
https://sourceforge.net/tracker/?func=detail&atid=684730&aid=1755232&group_id=119701