------- Comment #3 from dg dot recrutement31 at gmail dot com 2010-05-27 09:03
-------
(In reply to comment #2)
> (In reply to comment #1)
> > I think your inline-asm is totally broken, the constraints you have don't
> > mean
> > the extended (32bit) registers.
> Humm... strange answer. Certainly, you've read too fast and too late in the
> evening.
> You "think" but it's better to demonstrate.
> How to explain the bug disappears when -g or -O0 turned on ?
To notice:
the bug disappears also if function toto is removed and its code moved in the
main function like this:
#include "far_pointers.h"
char u;
/*void toto(far_pointer<unsigned char> pt)
{
u = *pt;
}*/
int main(int argc, char **argv)
{
unsigned short dataSeg = 0;
asm (
" \
.intel_syntax noprefix; \n \
mov %0, cs; \n \
"
: "=r"(dataSeg)
);
far_pointer<unsigned char> pt(dataSeg, (unsigned char*) &main);
//toto(pt);
u = *pt;
return u;
}
The assembly is correctly generated:
.file "main.cpp"
.intel_syntax noprefix
.def ___main; .scl 2; .type 32; .endef
.text
.globl _main
.def _main; .scl 2; .type 32; .endef
_main:
LFB30:
push ebp
LCFI0:
mov ebp, esp
LCFI1:
and esp, -16
LCFI2:
call ___main
/APP
# 19 "main.cpp" 1
.intel_syntax
noprefix;
mov ax, cs;
# 0 "" 2
# 59 "far_pointers.h" 1
.intel_syntax noprefix;
mov gs, ax;
mov al, gs:[OFFSET FLAT:_main];
# 0 "" 2
/NO_APP
mov BYTE PTR _u, al
movsx eax, al
leave
LCFI3:
ret
LFE30:
.globl _u
.bss
_u:
.space 1
All these tests seem to demonstrate there is a bug.
--
http://gcc.gnu.org/bugzilla/show_bug.cgi?id=44288
