[Mingw-users] Re: Patch Request: Code Generation for main( ) function

[ranjit_mathew <rm...@ho...>]

--- In mingw-users@y..., Greg Chicares <chicares@m...> wrote:
>
> But main() is unique. It's C[++]'s interface to the loader.
> You can't call it.

I can't? Well, this program compiles and runs just 
fine with "gcc -Wall" on MinGW 1.1 (gcc 2.95.3-6):
------------------------ 8< ---------------------
#include <stdio.h>

void foo( void);

int main( int argc, char *argv[])
{
    if( argc == 23)
        printf( "Hello World!\n");
    else
        foo( );

    return 0;
}
void foo( void)
{
    main( 23, NULL);
}
------------------------ 8< ---------------------

:-P


> so code generation already has to distinguish it from other
> functions.

Nope. Code *compilation* has to distinguish it from
other functions, as well as the code *linked* with
it.

Nitpicking...


> > The call to __main( ) looks like a kludge to me - I
> > just cannot understand why #ifdefs in the C startup code
> > cannot achieve the same effect in the GCC source.
> 
> It's a common technique; here's how one vendor explains it
>   http://www.hh.se/stud/d98rolb/ansi/__main().html
> It's just an implementation choice.

From that page, it appears to be a part of the C startup
code and I have no problems at all with that. I'm 
uncomfortable when it generates a call to this from
within "my" main( ).


> Usually I want what _main() provides. But you can hack
> the sources to get something different. Maybe you can
> hack them enough to get something like this
> 
> > unsigned char main[] = {
> > 0x55,0x89,0xe5,0x56, 0x53,0x83,0xec,0x50,
> ...
> > 0x00,0x83,0xc4,0x50, 0x5b,0x5e,0x5d,0xc3
> > };
> 
> to do what you want, but all the standard requires mingw
> to do with that program is to emit at least one diagnostic.

Once again, even "gcc -Wall" does not give any warnings/errors
with the program above (again MinGW 1.1, gcc 2.95.3-6).

Ranjit.

http://groups.yahoo.com/group/mingw-users/
(mingw-users on Yahoo! Groups)