• d-bugmail puremagic.com Dec 20 2006
• Sean Kelly <sean f4.ca> Dec 20 2006
• d-bugmail puremagic.com Jan 23 2007
• d-bugmail puremagic.com Nov 13 2008
• d-bugmail puremagic.com Nov 26 2010

d-bugmail puremagic.com writes:

http://d.puremagic.com/issues/show_bug.cgi?id=701

           Summary: Inline asm using incorrect offsets when used in inner
                    function
           Product: D
           Version: 0.177
          Platform: PC
        OS/Version: Windows
            Status: NEW
          Severity: normal
          Priority: P2
         Component: DMD
        AssignedTo: bugzilla digitalmars.com
        ReportedBy: sean f4.ca


I expect the following code:

    void main()
    {
        int i = 0;

        void fn()
        {
            asm
            {
                naked;
                lea EAX, i;
                mov [EAX], 42;
                ret;
            }
        }
        fn();
        printf( "i = %d\n", i );
    }

to print "42" but instead it prints "0".  This is because the assembler uses
the offset of 'i' that would be used within main() rather than adjusting for
the inner function.  Changing the code to this:

    void main()
    {
        int i = 0;

        void fn()
        {
            asm
            {
                naked;
                lea EAX, i;
                add EAX, 4;
                mov [EAX], 42;
                ret;
            }
        }
        fn();
        printf( "i = %d\n", i );
    }

Prints "42" as desired, but a manual adjustment of offsets should not be
necessary.  This is particulrly problematic in situations where "naked" is not
used, so the amount to adjust the offset by is not fixed.


-- 

Sean Kelly <sean f4.ca> writes:

Upon reflection, I'm not entirely sure what the correct behavior should 
be here.  However, I think it's misleading that the code currently 
complies and silently produces the incorrect result.  If nothing else, 
it would be nice if this worked with 'naked' not present.

d-bugmail puremagic.com writes:

http://d.puremagic.com/issues/show_bug.cgi?id=701





------- Comment #2 from thomas-dloop kuehne.cn  2007-01-23 06:05 -------
 mov [EAX], 42;


This should be 

 mov int ptr [EAX], 42;


I don't think there is a way to use a single "lea" to solve your problem, 
however lea seems to be broken:

# asm{
#    lea EAX, [EBP-24] + 1; 
#    lea EBX, 1 - [EBP-24];
# }

results in

 8d 45 e9 lea eax, [ebp-23]
 8d 5d 19 lea ebx, [ebp+25]


I'm not a master of all x86 addressing modes but it seems odd.


-- 

d-bugmail puremagic.com writes:

http://d.puremagic.com/issues/show_bug.cgi?id=701


clugdbug yahoo.com.au changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
            Summary|Inline asm using incorrect  |Inline naked asm uses
                   |offsets when used in inner  |incorrect offsets
                   |function                    |




------- Comment #3 from clugdbug yahoo.com.au  2008-11-13 12:50 -------
I'm changing the name of this issue, since it actually has nothing to do with
inner functions. It applies to _any_ use of 'naked'. Basically naked calculates
offsets assuming that a stack frame is present -- even though the main use of
naked is to avoid having a stack frame!


-- 

d-bugmail puremagic.com writes:

http://d.puremagic.com/issues/show_bug.cgi?id=701


Andrei Alexandrescu <andrei metalanguage.com> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
             Status|NEW                         |ASSIGNED
                 CC|                            |andrei metalanguage.com
         AssignedTo|nobody puremagic.com        |bugzilla digitalmars.com


-- 
Configure issuemail: http://d.puremagic.com/issues/userprefs.cgi?tab=email
------- You are receiving this mail because: -------