From mboxrd@z Thu Jan  1 00:00:00 1970
From: Mark Klein <mklein@dis.com>
To: law@cygnus.com
Cc: gcc@gcc.gnu.org
Subject: Re: MPE Port 
Date: Sun, 31 Oct 1999 23:35:00 -0000
Message-ID: <4.2.0.58.19991014070452.00aecef0@209.60.244.235>
References: <4.2.0.58.19991009200725.00c5a100@garfield.dis.com> <30257.939898754@upchuck.cygnus.com>
X-SW-Source: 1999-10n/msg00293.html
Message-ID: <19991031233500.-j8iaVASYZfHhTBgo4YaaUnBOjMpdhNuoYOXcKbYr94@z>

At 04:59 AM 10/14/99 -0600, Jeffrey A Law wrote:
    ^^^^^
Late night, huh? :-)

>I can't stress enough that it is totally wrong for us to try and do this
>within GCC.

I don't see how this is any different than the following code that is
already in GCC:

/* We want __gcc_plt_call to appear in every program built by
    gcc, so we make a reference to it out of __main.
    We use the asm statement to fool the optimizer into not
    removing the dead (but important) initialization of
    REFERENCE.  */

#define DO_GLOBAL_DTORS_BODY                    \
do {                                            \
   extern void __gcc_plt_call ();                \
   void (*reference)() = &__gcc_plt_call;        \
   func_ptr *p;                                  \
   __asm__ ("" : : "r" (reference));             \
   for (p = __DTOR_LIST__ + 1; *p; )             \
     (*p++) ();                                  \
} while (0)

If you change "__gcc_plt_call" to "__MPE_arm_subsystem_break", what's
the rationale for doing one, but not the other?

gcc provides it's own __main and by so doing, eliminates certain
functionality expected by the system __main. Why is it not
appropriate for gcc to replace functionality that it eliminates?


Regards,


M.


--
Mark Klein                                    DIS International, Ltd.
http://www.dis.com                            415-892-8400
PGP Public Key Available
--