Fixed bug in caseExactSubstringsIndexer() and caseIgnoreSubstringsIndexer().

UTF8 normalization must be done before we compute number of keys since
string length might increase.

diff --git a/servers/slapd/schema_init.c b/servers/slapd/schema_init.c
index b0a5b22d7ab4ce264a5cd374cca2b00528208202..e320a73ca8eb28d60b3a26f4a73306f87a8e41f9 100644 (file)
--- a/servers/slapd/schema_init.c
+++ b/servers/slapd/schema_init.c
@@ -1273,6 +1273,10 @@ int caseExactSubstringsIndexer(
        ber_len_t i, nkeys;
        size_t slen, mlen;
        struct berval **keys;
+#if UTF8MATCH
+        struct berval **nvalues;
+#endif
+
        HASH_CONTEXT   HASHcontext;
        unsigned char   HASHdigest[HASH_BYTES];
        struct berval digest;
@@ -1283,6 +1287,20 @@ int caseExactSubstringsIndexer(
        assert( values != NULL && values[0] != NULL );
 
        nkeys=0;
+
+#if UTF8MATCH
+        /* create normalized copy of values */
+        for( i=0; values[i] != NULL; i++ ) {
+                /* empty */
+        }
+        nvalues = ch_malloc( sizeof( struct berval * ) * (i+1) );
+        for( i=0; values[i] != NULL; i++ ) {
+                nvalues[i] = ber_bvstr( UTF8normalize( values[i]->bv_val, UTF8_NOCASEFOLD ) );
+        }
+        nvalues[i] = NULL;
+        values = nvalues;
+#endif
+
        for( i=0; values[i] != NULL; i++ ) {
                /* count number of indices to generate */
                if( values[i]->bv_len < SLAP_INDEX_SUBSTR_MINLEN ) {
@@ -1332,12 +1350,7 @@ int caseExactSubstringsIndexer(
 
                if( values[i]->bv_len < SLAP_INDEX_SUBSTR_MINLEN ) continue;
 
-#if UTF8MATCH
-               value = ber_bvstr( UTF8normalize( values[i]->bv_val,
-                       UTF8_NOCASEFOLD ) );
-#else
                value = values[i];
-#endif
 
                if( ( flags & SLAP_INDEX_SUBSTR_ANY ) &&
                        ( value->bv_len >= SLAP_INDEX_SUBSTR_MAXLEN ) )
@@ -1415,9 +1428,6 @@ int caseExactSubstringsIndexer(
 
                }
 
-#if UTF8MATCH
-               ber_bvfree( value );
-#endif
        }
 
        if( nkeys > 0 ) {
@@ -1428,6 +1438,10 @@ int caseExactSubstringsIndexer(
                *keysp = NULL;
        }
 
+#if UTF8MATCH
+       ber_bvecfree( nvalues );
+#endif
+
        return LDAP_SUCCESS;
 }
 
@@ -1929,6 +1943,10 @@ int caseIgnoreSubstringsIndexer(
        ber_len_t i, nkeys;
        size_t slen, mlen;
        struct berval **keys;
+#if UTF8MATCH
+       struct berval **nvalues;
+#endif
+
        HASH_CONTEXT   HASHcontext;
        unsigned char   HASHdigest[HASH_BYTES];
        struct berval digest;
@@ -1939,6 +1957,20 @@ int caseIgnoreSubstringsIndexer(
        assert( values != NULL && values[0] != NULL );
 
        nkeys=0;
+
+#if UTF8MATCH
+       /* create normalized copy of values */
+       for( i=0; values[i] != NULL; i++ ) {
+               /* empty */
+       }
+       nvalues = ch_malloc( sizeof( struct berval * ) * (i+1) );
+       for( i=0; values[i] != NULL; i++ ) {
+               nvalues[i] = ber_bvstr( UTF8normalize( values[i]->bv_val, UTF8_CASEFOLD ) );
+       }
+       nvalues[i] = NULL;
+       values = nvalues;
+#endif
+
        for( i=0; values[i] != NULL; i++ ) {
                /* count number of indices to generate */
                if( values[i]->bv_len < SLAP_INDEX_SUBSTR_MINLEN ) {
@@ -1989,7 +2021,7 @@ int caseIgnoreSubstringsIndexer(
                if( values[i]->bv_len < SLAP_INDEX_SUBSTR_MINLEN ) continue;
 
 #if UTF8MATCH
-               value = ber_bvstr( UTF8normalize( values[i]->bv_val, UTF8_CASEFOLD ) );
+               value = values[i];
 #else
                value = ber_bvdup( values[i] );
                ldap_pvt_str2upper( value->bv_val );
@@ -2070,8 +2102,9 @@ int caseIgnoreSubstringsIndexer(
                        }
 
                }
-
+#if !UTF8MATCH
                ber_bvfree( value );
+#endif
        }
 
        if( nkeys > 0 ) {
@@ -2082,6 +2115,9 @@ int caseIgnoreSubstringsIndexer(
                *keysp = NULL;
        }
 
+#if UTF8MATCH
+       ber_bvecfree( nvalues );
+#endif
        return LDAP_SUCCESS;
 }