Re: egcs on UnixWare 2.1.2

[hjl at lucon dot org (H.J. Lu)]

> 
> Executive Summary: #undef HAVE_ATEXIT in i386/sys4.h prevents
> my hello world program from dumping core.
> 
> 
> The egcs distributions have been broken for me since late September
> on UnixWare 2.1.2.  A hello world progam would dump core after
> having been compiled with the stage1 compiler.
> Here is the source code:
> 
> (BTW, the _exit() call inhibited the core dump, when it was
> enabled.)
> 
> 
> #include <stdio.h>
> 
> main()
> {
> 	printf("Hello world!\n");
> 	/*_exit(0); */
> }
> 
> 
> Here's how I compile it:
> 
> : ~ Wed 19 0:26; ./egcs-971114/gcc/xgcc -B./egcs-971114/gcc/ -v -g x.c
> Reading specs from ./egcs-971114/gcc/specs
> gcc version egcs-2.90.17 971114 (gcc2-970802 experimental)
>  ./egcs-971114/gcc/cpp -lang-c -v -isystem ./egcs-971114/gcc/include -undef -D__GNUC__=2 -D__GNUC_MINOR__=90 -Di386 -Dunix -D__svr4__ -D__i386__ -D__unix__ -D__svr4__ -D__i386 -D__unix -Asystem(unix) -Asystem(svr4) -Acpu(i386) -Amachine(i386) -g x.c /var/tmp/cca001-h.i
> GNU CPP version egcs-2.90.17 971114 (gcc2-970802 experimental) (i386 System V Release 4)
> #include "..." search starts here:
> #include <...> search starts here:
>  egcs-971114/gcc/include
>  /usr/local/include
>  /usr/include
> End of search list.
>  ./egcs-971114/gcc/cc1 /var/tmp/cca001-h.i -quiet -dumpbase x.c -g -version -o /var/tmp/cca001-h.s
> GNU C version egcs-2.90.17 971114 (gcc2-970802 experimental) (i386-pc-sysv4.2uw2.1.2) compiled by GNU C version egcs-2.90.06 970907 (gcc2-970802 experimental).
>  ./egcs-971114/gcc/as -V -Qy -o /var/tmp/cca001-h1.o /var/tmp/cca001-h.s
> GNU assembler version 970731 (i386-pc-sysv4.2uw2.1.2), using BFD version linux-2.8.1.0.15
>  ./egcs-971114/gcc/ld -V -Y P,/usr/ccs/lib:/usr/lib -Qy /usr/ccs/lib/crt1.o /usr/ccs/lib/crti.o /usr/ccs/lib/values-Xa.o ./egcs-971114/gcc/crtbegin.o -L./egcs-971114/gcc -L/usr/ccs/bin -L/usr/ccs/lib /var/tmp/cca001-h1.o -lgcc -lc -lgcc ./egcs-971114/gcc/crtend.o /usr/ccs/lib/crtn.o
> UX:ld: INFO:  Optimizing C Compilation System  (CCS) 3.0  12/10/95 (eiger8)
> 
> 
> : ~ Wed 19 0:27; ./a.out 
> Hello world!
> zsh: IOT instruction (core dumped)  ./a.out
> 
> 
> Now, here's what happens when I run gdb on the executable:
> 
> : ~ Wed 19 0:28; gdb ./a.out
> GDB is free software and you are welcome to distribute copies of it
>  under certain conditions; type "show copying" to see the conditions.
> There is absolutely no warranty for GDB; type "show warranty" for details.
> GDB 4.15 (unixware2), Copyright 1995 Free Software Foundation, Inc...
> (gdb) b __do_global_dtors_aux
> Breakpoint 1 at 0x8048501
> (gdb) run
> Starting program: /home/acs/./a.out 
> 
> Program received signal SIGTRAP, Trace/breakpoint trap.
> [Switching to process 6533]
> Hello world!
> 
> Breakpoint 1, 0x8048501 in __do_global_dtors_aux ()
> (gdb) bt
> #0  0x8048501 in __do_global_dtors_aux ()
> #1  0x8049035 in _fini ()
> (gdb) cont
> Continuing.
> 
> Breakpoint 1, 0x8048501 in __do_global_dtors_aux ()
> (gdb) bt
> #0  0x8048501 in __do_global_dtors_aux ()
> #1  0x8049035 in _fini ()
> (gdb) cont
> Continuing.
> 
> Program received signal SIGABRT, Aborted.
> 0xbffb3783 in kill ()
> (gdb) q
> The program is running.  Quit anyway (and kill it)? (y or n) y
> 
> 
> As I wrote, the *second* time through __do_global_dtors_aux is when it 
> dies.  (Note, I never said that there were *nested* calls to
> __do_global_dtors_aux.)
> 

__do_global_dtors_aux () is called twice.

> 
> After I added #undef'ed HAVE_ATEXIT in i386/sys4.h, my hello world

If everything is setup right, HAVE_ATEXIT should have nothing
to do with __do_global_dtors_aux () on UnixWare.

> program no dumps core.  It also no longer calls
> __do_global_dtors_aux, as you can see here:

I don't know if it is good or bad, maybe bad.

If you can post crtbegin.s, crtend.s and __main.s, I will
take a look.



H.J.