Re: [RFA] pei386 dll: auto-import patch

[Charles Wilson <cwilson at ece dot gatech dot edu>]

DJ Delorie wrote:

 >> ret = snprintf (fixup_name, buffer_len, "__fu%d_%s", counter++,
 >> sym->name);
 >>
 >
 > Sorry, can't use snprintf.


Why not?

 > Use strlen every time (figure 20 for the counter) and compare to the
 >  max length.  Round up to a multiple of 100 to reduce heap
 > fragmentation.  Then use regular sprintf.

Like this?  ('course, you have to #include <math.h> for the ceil 
function.  Perhaps multiples of 128 or 256 would be better...bit 
shifting and masking is easy -- but not really portable...)

-----------------------
   static int counter;
   static char *fixup_name = NULL;
   static int buffer_len = 0;

   struct symbol_cache_entry *sym = *rel->sym_ptr_ptr;

   bfd *abfd = bfd_asymbol_bfd (sym);
   struct coff_link_hash_entry *myh = NULL;

   if (!fixup_name)
     {
       fixup_name = (char *) xmalloc (300);
       buffer_len = 300;
     }

   if (strlen (sym->name) + 25 > buffer_len)
   /* assume 25 chars for "__fu" + counter + "_".  If counter is
      bigger than 20 digits long, we've got worse problems than
      overflowing this buffer... */
     {
       free (fixup_name);
       /* new buffer size is length of symbol, plus 25, but then
          rounded up to the nearest multiple of 100 */
       buffer_len = (int) ((ceil (((double) ( strlen (sym->name) +
         25 )) / 100.0)) * 100.0);
       fixup_name = (char *) xmalloc (buffer_len);
     }

   sprintf (fixup_name, "__fu%d_%s", counter++, sym->name);

   ...

-------------------

FWIW, all of my changes so far (including this one) compile at least. 
:-)  I'll test actual operation <yawn> tomorrow.  (The version I 
reported at 11:26EDT 1Aug01, and for which I put binaries on my website 
at 1:07EDT 2Aug01, has been minimally tested in operation.  It's just 
these most recent nit-picky changes that haven't been tested beyond "it 
builds"...)

--Chuck