PATCH COMMITTED: PR 6493: Work around -D_FORTIFY_SOURCE

[Michael Hennebry]

On Fri, 9 May 2008, Ian Lance Taylor wrote:

> PR 6493 is about a warning in gold when compiling with
> -D_FORTIFY_SOURCE (at least, I assume they are using that -D option,
> as otherwise the bug report does not make sense).  In this case gold
> is calling write in an error case, so it is correctly ignoring the
> return value.  This patch adds some pointless uses of the return value
> to avoid the warning.

I thought that was what coercion to void was for.
It would also have the side-effect of being
more informative to subsequent readers.

> Index: gold.cc
> ===================================================================
> RCS file: /cvs/src/src/gold/gold.cc,v
> retrieving revision 1.55
> diff -p -u -r1.55 gold.cc
> --- gold.cc     7 May 2008 06:08:01 -0000       1.55
> +++ gold.cc     9 May 2008 14:12:14 -0000
> @@ -60,9 +60,18 @@ gold_nomem()
>    // We are out of memory, so try hard to print a reasonable message.
>    // Note that we don't try to translate this message, since the
>    // translation process itself will require memory.
> -  write(2, program_name, strlen(program_name));
> -  const char* const s = ": out of memory\n";
> -  write(2, s, strlen(s));
> +
> +  // LEN only exists to avoid a pointless warning when write is
> +  // declared with warn_use_result, as when compiling with
> +  // -D_USE_FORTIFY on GNU/Linux.  Casting to void does not appear to
> +  // work, at least not with gcc 4.3.0.
> +
> +  ssize_t len = write(2, program_name, strlen(program_name));
> +  if (len >= 0)
> +    {
> +      const char* const s = ": out of memory\n";
> +      len = write(2, s, strlen(s));

         (void)write(2, s, strlen(s));


> +    }
>    gold_exit(false);
>  }

-- 
Michael   hennebry@web.cs.ndsu.NoDak.edu
"Those parts of the system that you can hit with a hammer (not advised)
are called Hardware;  those program instructions that you can only
curse at are called Software."