Re: [gnugo-devel] hash collisions

[Gunnar Farnebäck]

Doug wrote:
> I think the intent is to see leading zeros (in particular,
> hashdata_to_string prints the hash values next to each other, so without
> leading zeros there's some ambiguity)

Correct.

> but %0lx doesn't do this.

Too bad.

> %08x works for me, but I don't know the right platform independent
> construct.

With some luck the patch below should solve the problem properly.

/Gunnar

Index: engine/hash.c
===================================================================
RCS file: /cvsroot/gnugo/gnugo/engine/hash.c,v
retrieving revision 1.28
diff -u -r1.28 hash.c
--- engine/hash.c       7 Jun 2004 17:15:52 -0000       1.28
+++ engine/hash.c       17 Jun 2004 00:10:48 -0000
@@ -168,8 +168,8 @@
 }
 
 
-#define BUFFER_SIZE (1 + NUM_HASHVALUES * (1 + (CHAR_BIT * SIZEOF_HASHVALUE \
-                                               - 1) / 4))
+#define HASHVALUE_NUM_DIGITS (1 + (CHAR_BIT * SIZEOF_HASHVALUE - 1) / 4)
+#define BUFFER_SIZE (1 + NUM_HASHVALUES * HASHVALUE_NUM_DIGITS)
 char *
 hashdata_to_string(Hash_data *hashdata)
 {
@@ -178,7 +178,8 @@
   int k;
   
   for (k = 0; k < NUM_HASHVALUES; k++) {
-    n += sprintf(buffer + n, HASHVALUE_PRINT_FORMAT, hashdata->hashval[k]);
+    n += sprintf(buffer + n, HASHVALUE_PRINT_FORMAT,
+                HASHVALUE_NUM_DIGITS, hashdata->hashval[k]);
     gg_assert(n < BUFFER_SIZE);
   }
 
Index: engine/hash.h
===================================================================
RCS file: /cvsroot/gnugo/gnugo/engine/hash.h,v
retrieving revision 1.29
diff -u -r1.29 hash.h
--- engine/hash.h       27 May 2004 00:08:47 -0000      1.29
+++ engine/hash.h       17 Jun 2004 00:10:48 -0000
@@ -52,7 +52,7 @@
  */
 typedef unsigned long Hashvalue;
 #define SIZEOF_HASHVALUE SIZEOF_LONG
-#define HASHVALUE_PRINT_FORMAT "%0lx"
+#define HASHVALUE_PRINT_FORMAT "%0*lx"
 
 /* for testing: Enables a lot of checks. */
 #define CHECK_HASHING 0