From:	Martin Str|mberg <ams AT ludd DOT luth DOT se>
Message-Id:	<200009131724.TAA23635@father.ludd.luth.se>
Subject:	(fwd) startup-code
To:	djgpp-workers AT delorie DOT com (DJGPP-WORKERS)
Date:	Wed, 13 Sep 2000 19:24:55 +0200 (MET DST)
X-Mailer:	ELM [version 2.4ME+ PL54 (25)]
MIME-Version:	1.0
Reply-To:	djgpp-workers AT delorie DOT com

Christer Sandberg <christer DOT sandberg AT mdh DOT se> said on c.o.m.d:
> Declaring a global identifier named 'end' will clatch with some 
> startup/exit function or data. Internal data should have underscore as prefix.
> Program built from the following code will crash on exit.
> int end[1000];
> int main(void)
> {
>    memset(end, 0, sizeof(int)*1000)
>     return 0;
> }

I can confirm the problem. Digging in I find this in the output of "nm
-a libc.a":

dpmiexcp.o:
00001078 b .bss
00001014 d .data
00000000 t .text
         U __8087
...
00001628 b _timer_ori
00000000 ? dpmiexcp.c
         U end
         U exception_stack
00000000 t gcc2_compiled.

But I can't find any symbol "end" defined/declared (I don't know the
assembly term) globally. How can this link at all?


I also find an interesting tidbit in crt0.S:

/*      .lcomm  __what_size_app_thinks_it_is, 4 */
.data
__what_size_app_thinks_it_is:
        .long   end

Coincidence? (I don't think so.)

So I try uncommenting this. No go. The symbol can't be found while
linking symify. I've also tried substituting
__what_size_app_thinks_it_is for end and vice versa. Looking for
similar cases in the code of dpmiexcp.c I see it uses
__djstack_limit. But I can't figure out where __djstack_limit is
defined. Neither can I figure out why __djstack_limit can be found
while __what_size_app_thinks_it_is can't.


Assistance required.


Right,

						MartinS

- Raw text -