Fix #20262: array_unique() with SORT_REGULAR misses duplicates

Zend/zend_operators.c

@@ -3424,9 +3424,18 @@ ZEND_API int ZEND_FASTCALL zendi_smart_strcmp(zend_string *s1, zend_string *s2)
 	int oflow1, oflow2;
 	zend_long lval1 = 0, lval2 = 0;
 	double dval1 = 0.0, dval2 = 0.0;
+	bool trailing1 = false, trailing2 = false;
 
-	if ((ret1 = is_numeric_string_ex(s1->val, s1->len, &lval1, &dval1, false, &oflow1, NULL)) &&
-		(ret2 = is_numeric_string_ex(s2->val, s2->len, &lval2, &dval2, false, &oflow2, NULL))) {
+	/* Use allow_errors=true to extract leading numeric values even with trailing data.
+	 * If BOTH strings have numeric prefixes, do numeric comparison to maintain transitivity.
+	 * This ensures the comparison function is transitive, which is required for
+	 * stable sorting and correct behavior in array_unique() with SORT_REGULAR.
+	 */
+	ret1 = is_numeric_string_ex(s1->val, s1->len, &lval1, &dval1, true, &oflow1, &trailing1);
+	ret2 = is_numeric_string_ex(s2->val, s2->len, &lval2, &dval2, true, &oflow2, &trailing2);
+
+	/* If BOTH strings have numeric prefixes, do numeric comparison */
+	if (ret1 && ret2) {
 #if ZEND_ULONG_MAX == 0xFFFFFFFF
 		if (oflow1 != 0 && oflow1 == oflow2 && dval1 - dval2 == 0. &&
 			((oflow1 == 1 && dval1 > 9007199254740991. /*0x1FFFFFFFFFFFFF*/)
@@ -3456,8 +3465,24 @@ ZEND_API int ZEND_FASTCALL zendi_smart_strcmp(zend_string *s1, zend_string *s2)
 				goto string_cmp;
 			}
 			dval1 = dval1 - dval2;
+			if (dval1 == 0.0) {
+				/* Numeric values are equal, check trailing data for tie-breaker */
+				if (!trailing1 && trailing2) {
+					return -1; /* No trailing data < has trailing data */
+				} else if (trailing1 && !trailing2) {
+					return 1;  /* Has trailing data > no trailing data */
+				}
+			}
 			return ZEND_NORMALIZE_BOOL(dval1);
 		} else { /* they both have to be long's */
+			if (lval1 == lval2) {
+				/* Numeric values are equal, check trailing data for tie-breaker */
+				if (!trailing1 && trailing2) {
+					return -1; /* No trailing data < has trailing data */
+				} else if (trailing1 && !trailing2) {
+					return 1;  /* Has trailing data > no trailing data */
+				}
+			}
 			return lval1 > lval2 ? 1 : (lval1 < lval2 ? -1 : 0);
 		}
 	} else {

ext/standard/tests/array/gh20262.phpt

@@ -0,0 +1,90 @@
+--TEST--
+GH-20262: array_unique() with SORT_REGULAR fails to identify duplicates with mixed numeric/alphanumeric strings
+--FILE--
+<?php
+// Original bug report: array_unique() with SORT_REGULAR doesn't properly
+// identify duplicates when array contains mixed numeric and alphanumeric strings
+$units = ['5', '10', '5', '3A', '5', '5'];
+$unique = array_unique($units, SORT_REGULAR);
+
+echo "Input array:\n";
+var_dump($units);
+
+echo "\nResult of array_unique():\n";
+var_dump($unique);
+
+echo "\nExpected: 3 unique values ('5', '10', '3A')\n";
+echo "Actual count: " . count($unique) . "\n";
+
+// Additional test: verify sort() groups equal values correctly
+$arr = ['5', '10', '5', '3A', '5', '5'];
+sort($arr);
+
+echo "\nSorted array (equal values should be grouped):\n";
+var_dump($arr);
+
+// Verify '5' values are consecutive
+$positions = [];
+foreach ($arr as $idx => $val) {
+    if ($val === '5') {
+        $positions[] = $idx;
+    }
+}
+
+$consecutive = true;
+for ($i = 0; $i < count($positions) - 1; $i++) {
+    if ($positions[$i] + 1 !== $positions[$i + 1]) {
+        $consecutive = false;
+        break;
+    }
+}
+
+echo "\nAll '5' values grouped together: " . ($consecutive ? "yes" : "no") . "\n";
+?>
+--EXPECT--
+Input array:
+array(6) {
+  [0]=>
+  string(1) "5"
+  [1]=>
+  string(2) "10"
+  [2]=>
+  string(1) "5"
+  [3]=>
+  string(2) "3A"
+  [4]=>
+  string(1) "5"
+  [5]=>
+  string(1) "5"
+}
+
+Result of array_unique():
+array(3) {
+  [0]=>
+  string(1) "5"
+  [1]=>
+  string(2) "10"
+  [3]=>
+  string(2) "3A"
+}
+
+Expected: 3 unique values ('5', '10', '3A')
+Actual count: 3
+
+Sorted array (equal values should be grouped):
+array(6) {
+  [0]=>
+  string(2) "3A"
+  [1]=>
+  string(1) "5"
+  [2]=>
+  string(1) "5"
+  [3]=>
+  string(1) "5"
+  [4]=>
+  string(1) "5"
+  [5]=>
+  string(2) "10"
+}
+
+All '5' values grouped together: yes

ext/standard/tests/dir/readdir_variation4.phpt

@@ -137,21 +137,21 @@ int(2)
 -- Call to readdir() --
 array(16) {
   [0]=>
-  string(9) "-10.5.tmp"
-  [1]=>
   string(9) "-2345.tmp"
+  [1]=>
+  string(9) "-10.5.tmp"
   [2]=>
   string(1) "."
   [3]=>
   string(2) ".."
   [4]=>
   string(4) ".tmp"
   [5]=>
-  string(7) "0.5.tmp"
-  [6]=>
   string(5) "0.tmp"
-  [7]=>
+  [6]=>
   string(17) "1.23456789E-9.tmp"
+  [7]=>
+  string(7) "0.5.tmp"
   [8]=>
   string(5) "1.tmp"
   [9]=>