Truncated fatal error messages

Truncated fatal error messages

[Eli Zaretskii]

This change:

  @@ -2012,7 +2014,12 @@ shut_down_emacs (int sig, Lisp_Object st
	 {
	  reset_all_sys_modes ();
	  if (sig && sig != SIGTERM)
  -	  fprintf (stderr, "Fatal error %d: %s", sig, strsignal (sig));
  +	  {
  +	    char buf[100];
  +	    int buflen = snprintf (buf, sizeof buf, "Fatal error %d: %s",
  +				   sig, strsignal (sig));
  +	    ignore_value (write (STDERR_FILENO, buf, buflen));
  +	  }
	 }
     }
   #else

arbitrarily limits the signal description due to the 99-char limit of
the buffer.  Isn't it better to allocate the buffer (with alloca)
based on the length of the signal description?

Re: Truncated fatal error messages

[Andreas Schwab]

Eli Zaretskii <eliz@gnu.org> writes:

> This change:
>
>   @@ -2012,7 +2014,12 @@ shut_down_emacs (int sig, Lisp_Object st
> 	 {
> 	  reset_all_sys_modes ();
> 	  if (sig && sig != SIGTERM)
>   -	  fprintf (stderr, "Fatal error %d: %s", sig, strsignal (sig));
>   +	  {
>   +	    char buf[100];
>   +	    int buflen = snprintf (buf, sizeof buf, "Fatal error %d: %s",
>   +				   sig, strsignal (sig));
>   +	    ignore_value (write (STDERR_FILENO, buf, buflen));
>   +	  }
> 	 }
>      }
>    #else
>
> arbitrarily limits the signal description due to the 99-char limit of
> the buffer.  Isn't it better to allocate the buffer (with alloca)
> based on the length of the signal description?

There is no need to copy the signal description in the buffer at all.

Andreas.

-- 
Andreas Schwab, schwab@linux-m68k.org
GPG Key fingerprint = 58CA 54C7 6D53 942B 1756  01D3 44D5 214B 8276 4ED5
"And now for something completely different."

Re: Truncated fatal error messages

[Paul Eggert]

On 09/22/2012 01:08 AM, Eli Zaretskii wrote:
> Isn't it better to allocate the buffer (with alloca)

If we're worried about arbitrarily long output of strsignal,
alloca won't work, since alloca can't go over a few
kilobytes.  And we can't use malloc here, since signal
handlers can't malloc.

In practice I expect the old code was fine, since signal
descriptions are never that long.  But since the issue has
been raised, I modified the code to avoid the need to allocate
an arbitrarily long buffer.