www.digitalmars.com         C & C++   DMDScript  

digitalmars.D.bugs - [Issue 5922] New: immutable and static variables usage in asm{}

reply d-bugmail puremagic.com writes:
http://d.puremagic.com/issues/show_bug.cgi?id=5922

           Summary: immutable and static variables usage in asm{}
           Product: D
           Version: D2
          Platform: x86
        OS/Version: Windows
            Status: NEW
          Severity: normal
          Priority: P2
         Component: DMD
        AssignedTo: nobody puremagic.com
        ReportedBy: bearophile_hugs eml.cc


--- Comment #0 from bearophile_hugs eml.cc 2011-05-03 03:04:41 PDT ---
I am not sure about the errors shown here (so no keywords). Please close this
bug report if things here are working as expected.

D2 code:


void main() {
    immutable size_t x = 10;
    asm {
        mov EDI, x;
    }
}


DMD 2.052 shows the error:
test.d(2): Error: Integer constant expression expected instead of x = 10u

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

D2 code, static (thread-local) variables can't be used in ASM:


void main() {
    static size_t x = 10;
    asm {
        mov EDI, x;
    }
}


It produces:
object.Error: Access Violation


They work using __gshared, this gives no errors:

void main() {
    __gshared static size_t x = 10;
    asm {
        mov EDI, x;
    }
}


My suggestion is to remove this source of errors. One solution may be to
disallow the direct access to static variables from asm code, avoiding this
bug.

-- 
Configure issuemail: http://d.puremagic.com/issues/userprefs.cgi?tab=email
------- You are receiving this mail because: -------
May 03 2011
next sibling parent d-bugmail puremagic.com writes:
http://d.puremagic.com/issues/show_bug.cgi?id=5922


Denis <verylonglogin.reg gmail.com> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
                 CC|                            |verylonglogin.reg gmail.com


--- Comment #1 from Denis <verylonglogin.reg gmail.com> 2012-03-02 17:37:11 MSK
---
The first one is definetly a bug with const/immutable local variables:
---
void f() {
    const size_t a = 1; // Error: Integer constant expression expected instead
of a = 1u
    version (D_InlineAsm_X86_64)
        asm { mov RAX, a; }
    else version (D_InlineAsm_X86)
        asm { mov EAX, a; }
    else
        static assert(1);
}
---

The second one looks like expected behaviour because with current inline
assembler for non-stack varable `someVar` instruction `mov EAX, someVar;` is
equal to `mov EAX, [someVar];` and `mov EAX, [n];` where `n` is a displacement
of `someVar` from some initial addres which depends on `someVar` type (global,
shared, or TLS).
Inline assembler behaves in this way but I've never seen any documentation and
reasons *why* does it behave this way because it's really unobvious.

-- 
Configure issuemail: http://d.puremagic.com/issues/userprefs.cgi?tab=email
------- You are receiving this mail because: -------
Mar 02 2012
prev sibling parent d-bugmail puremagic.com writes:
http://d.puremagic.com/issues/show_bug.cgi?id=5922


yebblies <yebblies gmail.com> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
                 CC|                            |yebblies gmail.com


--- Comment #2 from yebblies <yebblies gmail.com> 2013-01-17 00:04:57 EST ---
For some reason x's initializer is "x = 10" instead of just "10".  Probably a
side effect of some hack.

-- 
Configure issuemail: http://d.puremagic.com/issues/userprefs.cgi?tab=email
------- You are receiving this mail because: -------
Jan 16 2013