Very Computer

Is it possible to force a core dump in a program, and still have the program
continue running?  Forcing a core dump is easy with something like:

{
  char *dump=0;
  sprintf(dump, "This will segfault");

Quote:}

void
dump_core()
{
  switch(fork()) {
    case -1:
      exit(errno);
      break;
    case 0:
      {
        char *dump=0;
        sprintf(dump, "This will segfault");
        break;
      }
    default:
       return;
  }

Quote:}

Breif rational for trying this:
I want to write a sighandler that catches Seg-faults, bus errors, etc. and dumps
core so I can debug the program and find the error.  But I want the program to
re-exec itself also to attempt some sort of recovery from the error.

Eric

--------------------------------+----------------------------------------------
Eric Boyd (TSMA)                |       "It's easier to ask for
InterDimensions Corp.           |        forgiveness than for permission."
25 Ellery St.                   |
Cambridge Ma, 02138             |       "640K ought to be enough for anybody."
617-661-4200                    |               -- Bill Gates, 1981
                                |

Quote:

>Is it possible to force a core dump in a program, and still have the program
>continue running?  Forcing a core dump is easy with something like:

--

``Not only is UNIX dead, but it's starting to smell really bad.''  -- rob

 Eric> Is it possible to force a core dump in a program, and still
 Eric> have the program continue running?  Forcing a core dump is easy
 Eric> with something like:

 Eric> {
 Eric>   char *dump=0;
 Eric>   sprintf(dump, "This will segfault");
 Eric> }

Yuk!

 [code using fork() elided]

 Eric> This will cause the program to fork, and the child can then
 Eric> dump core.  The parent is then allowed to continue running.  Is
 Eric> there a better way to do this?

 Eric> Breif rational for trying this: I want to write a sighandler
 Eric> that catches Seg-faults, bus errors, etc. and dumps core so I
 Eric> can debug the program and find the error.  But I want the
 Eric> program to re-exec itself also to attempt some sort of recovery
 Eric> from the error.

I'd do it like this:

void dump_core()
{
    if (fork() == 0)
        raise(SIGABRT);

Quote:}

Your method would not work well if called from a handler for SIGSEGV,
unless that handler had been installed with "oldstyle" semantics.
Remember: use sigaction(), *NEVER* use signal(), even though on some
systems signal() gives you the behaviour you want. On a BSD system,
if you cause a segfault within a signal handler for SIGSEGV, the
probable result is an infinite loop trying to retry the faulting
instruction.

--
Andrew.

If you want to continue running try:

char buf[32];

sprintf(buf, "gcore %d", getpid());
system(buf);

The above code will show the top of stack in system(), but
everything below will be your original context.

Quote:>Breif rational for trying this:
>I want to write a sighandler that catches Seg-faults, bus errors, etc. and dumps
>core so I can debug the program and find the error.  But I want the program to
>re-exec itself also to attempt some sort of recovery from the error.

Shawn

Isn't gcore peculiar to Solaris?

--
Andy Knight

Quote:> If you want to continue running try:

> char buf[32];

> sprintf(buf, "gcore %d", getpid());
> system(buf);

> The above code will show the top of stack in system(), but
> everything below will be your original context.

It may be peculiar to Sun - it is present in both SunOS 4.x and Solaris.