dotgnu-pnet
[Top][All Lists]
Advanced

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

[Pnet-developers] [Bug #2785] Verifier : ldobj , ldflda causing problems


From: nobody
Subject: [Pnet-developers] [Bug #2785] Verifier : ldobj , ldflda causing problems
Date: Thu, 27 Mar 2003 17:17:04 -0500

=================== BUG #2785: LATEST MODIFICATIONS ==================
http://savannah.gnu.org/bugs/?func=detailbug&bug_id=2785&group_id=353

Changes by: Rhys Weatherley <address@hidden>
Date: Thu 03/27/03 at 22:17 (GMT)

------------------ Additional Follow-up Comments ----------------------------
What was the resolution for this?  I still think that
this is a code generation bug in mcs.  Taking the address
of a field within the top-most value on the stack doesn't
seem correct - the value will be popped, leaving the
address pointing at invalid memory.  Can I close this?



=================== BUG #2785: FULL BUG SNAPSHOT ===================


Submitted by: t3rmin4t0r              Project: DotGNU Portable.NET          
Submitted on: Tue 03/11/03 at 16:08
Category:  None                       Severity:  5 - Major                  
Bug Group:  None                      Resolution:  None                     
Assigned to:  None                    Status:  Open                         

Summary:  Verifier : ldobj , ldflda causing problems 

Original Submission:  The attatched code is shredded down from SDL.Net code and
looks to be Ok.

        ldfld      valuetype 'SDL_Surface' * 'Surface'::'_surface'
        ldobj      valuetype 'SDL_Surface'
        ldflda     native int 'SDL_Surface'::'pixels'

The last line is creating problems because the ldobj makes
STK_UNARY== ILEngineType_MV and there's no case to
handle that.

PS: SDL.net is using MCS


Follow-up Comments
*******************

-------------------------------------------------------
Date: Thu 03/27/03 at 22:17         By: rweather
What was the resolution for this?  I still think that
this is a code generation bug in mcs.  Taking the address
of a field within the top-most value on the stack doesn't
seem correct - the value will be popped, leaving the
address pointing at invalid memory.  Can I close this?

-------------------------------------------------------
Date: Tue 03/11/03 at 21:04         By: rweather
I think this is a bug in mcs, but am not sure.  It isn't
legal to take the address of a field within a value type
on the stack, as this code is doing.  Leaving open for
a little while until more information can be collected.


CC list is empty


File Attachments
****************

-------------------------------------------------------
Date: Tue 03/11/03 at 16:08  Name: unsafe2.il  Size: 1KB   By: t3rmin4t0r
test case
http://savannah.gnu.org/bugs/download.php?group_id=353&amp;bug_id=2785&amp;bug_file_id=348


For detailed info, follow this link:
http://savannah.gnu.org/bugs/?func=detailbug&bug_id=2785&group_id=353


reply via email to

[Prev in Thread] Current Thread [Next in Thread]