Hi Bill, Francois, Darin, Mike, David:

Thursday, January 30 1997, you wrote to me:

 BC> In order to get info into your "log" file when OS2 is
about to shut down
 BC> your "appli", setup and use the SIGNAL functions.

I had some trouble with signal. Consider the next program, which
deliberately causes a segment violation (trap 005):

=============================== snip =========================================
#include 
#include 
#include 

void emergency(int sig);

FILE *fuit;

main()
{
   long a[10];

   signal(SIGSEGV, emergency);

   if ((fuit=fopen("test.out", "w")) == (FILE*) NULL)
      {
      fprintf(stderr,"File test.out cannot be opened. Sri.\n");
      exit(-1);
      }

   a[9] = 10L;
   printf("%ld \n", a[9]);
   fprintf(fuit, "%ld \n", a[9]);

   printf("A statement preceeding the illegal assignment\n");
   fprintf(fuit, "A statement preceeding the illegal assignment\n");

   a[10000] = 10L;
   printf("%ld \n", a[10000]);
   fprintf(fuit, "%ld \n", a[10000]);

   return 0;
}

void emergency(int sig)
{
   exit(1);
}
=============================== snip =========================================

Nothing spectacular.

When i compile this program using Borland C++ version 1.0, the result is:

D:\he3>bcc test.c

Borland C++ Version 1.00 Copyright (c) 1993 Borland International
test.c:
Error test.c 13: Type mismatch in parameter '__func' in call to 'signal' in
function main
Warning test.c 39: Parameter 'sig' is never used in function emergency
*** 1 errors in Compile ***

bcc nags about the type of the function emergency. Renaming it into
whatever name, makes no difference, it is not a predefined name. Using
different types (int etc.) does not change the error message. I can't find
the definition of signal in the include files, so i can't check what's
wrong. Any idea? My test.c source is a copy of the example in Borlands
manual. Borlands example, literally copied from page 285, produces the same
error message.

IBM's icc compiles my source without problems.

The results are astonishing: the mere fact that a signal function is
invoked, is sufficient to fflush the buffers both of the redirected i/o as
well as the streamed file. The trap 005 is gone, the violation remains. So
far so good. Now back to the program i want to debug...


Thanks for your help,
 Henk

--- GoldED 2.50+