From: Douglas Bagnall <douglas.bagnall@catalyst.net.nz>
Date: Tue, 9 Apr 2024 22:54:41 +0000 (+1200)
Subject: ldb:attrib_handlers: make ldb_comparison_Boolean more consistent
X-Git-Tag: ldb-2.8.1~7
X-Git-Url: http://git.ipfire.org/gitweb/gitweb.cgi?a=commitdiff_plain;h=ca905ccbf1e426dd04debb14c10e40db4ee4e42f;p=thirdparty%2Fsamba.git

ldb:attrib_handlers: make ldb_comparison_Boolean more consistent

This isn't supposed to be used for sorting, but it is hard to say it
won't be, so we might as well make it sort properly.

Following long-standing behaviour, we try to sort "FALSE" > "TRUE", by
length, then switch to using strncasecmp().

strncasecmp would sort the other way, so we swap the operands. This is
to make e.g. "TRUE\0" sort the same as "TRUE".

BUG: https://bugzilla.samba.org/show_bug.cgi?id=15625

Signed-off-by: Douglas Bagnall <douglas.bagnall@catalyst.net.nz>
Reviewed-by: Andrew Bartlett <abartlet@samba.org>
(cherry picked from commit a75c98ad688415aec8afc617a759ba90cfd9f23b)
---

diff --git a/lib/ldb/common/attrib_handlers.c b/lib/ldb/common/attrib_handlers.c
index baccf193f88..c01477331f0 100644
--- a/lib/ldb/common/attrib_handlers.c
+++ b/lib/ldb/common/attrib_handlers.c
@@ -281,15 +281,36 @@ static int ldb_canonicalise_Boolean(struct ldb_context *ldb, void *mem_ctx,
 }
 
 /*
-  compare two Booleans
-*/
+ * compare two Booleans.
+ *
+ * According to RFC4517 4.2.2, "the booleanMatch rule is an equality matching
+ * rule", meaning it isn't used for ordering.
+ *
+ * However, it seems conceivable that Samba could be coerced into sorting on a
+ * field with Boolean syntax, so we might as well have consistent behaviour in
+ * that case.
+ *
+ * The most probably values are {"FALSE", 5} and {"TRUE", 4}. To save time we
+ * compare first by length, which makes FALSE > TRUE. This is somewhat
+ * contrary to convention, but is how Samba has worked forever.
+ *
+ * If somehow we are comparing incompletely normalised values where the length
+ * is the same (for example {"false", 5} and {"TRUE\0", 5}), the length is the
+ * same, and we fall back to a strncasecmp. In this case, since "FALSE" is
+ * alphabetically lower, we swap the order, so that "TRUE\0" again comes
+ * before "FALSE".
+ *
+ * ldb_canonicalise_Boolean (just above) gives us a clue as to what we might
+ * expect to cope with by way of invalid values.
+ */
 static int ldb_comparison_Boolean(struct ldb_context *ldb, void *mem_ctx,
 			   const struct ldb_val *v1, const struct ldb_val *v2)
 {
 	if (v1->length != v2->length) {
-		return NUMERIC_CMP(v1->length, v2->length);
+		return NUMERIC_CMP(v2->length, v1->length);
 	}
-	return strncasecmp((char *)v1->data, (char *)v2->data, v1->length);
+	/* reversed, see long comment above */
+	return strncasecmp((char *)v2->data, (char *)v1->data, v1->length);
 }