Java: Improve Regex flag parsing

Fixes:
- Flag `d` not being recognized
- Syntax for disabling flags (`-`) not being recognized
- Non-capturing group with flags erroneously containing `:` as literal

java/ql/lib/semmle/code/java/regex/regex.qll

@@ -479,7 +479,7 @@ abstract class RegexString extends StringLiteral {
   private predicate flagGroupStartNoModes(int start, int end) {
     this.isGroupStart(start) and
     this.getChar(start + 1) = "?" and
-    this.getChar(start + 2) in ["i", "m", "s", "u", "x", "U"] and
+    this.getChar(start + 2) in ["-", "i", "d", "m", "s", "u", "x", "U"] and
     end = start + 2
   }
 
@@ -491,7 +491,7 @@ abstract class RegexString extends StringLiteral {
     this.flagGroupStartNoModes(start, pos)
     or
     this.modeCharacter(start, pos - 1) and
-    this.getChar(pos) in ["i", "m", "s", "u", "x", "U"]
+    this.getChar(pos) in ["-", "i", "d", "m", "s", "u", "x", "U"]
   }
 
   /**
@@ -499,7 +499,10 @@ abstract class RegexString extends StringLiteral {
    */
   private predicate flagGroupStart(int start, int end) {
     this.flagGroupStartNoModes(start, _) and
-    end = max(int i | this.modeCharacter(start, i) | i + 1)
+    // Check if this is a capturing group with flags, and therefore the `:` should be excluded
+    exists(int maybeEnd | maybeEnd = max(int i | this.modeCharacter(start, i) | i + 1) |
+      if this.getChar(maybeEnd) = ":" then end = maybeEnd + 1 else end = maybeEnd
+    )
   }
 
   /**
@@ -510,9 +513,15 @@ abstract class RegexString extends StringLiteral {
    * ```
    */
   private predicate flag(string c) {
-    exists(int pos |
-      this.modeCharacter(_, pos) and
-      this.getChar(pos) = c
+    exists(int start, int pos |
+      this.modeCharacter(start, pos) and
+      this.getChar(pos) = c and
+      // Ignore if flag is disabled; use `<=` to also exclude `-` itself
+      // This does not properly handle the (contrived) case where a flag is both enabled and
+      // disabled, e.g. `(?i-i)a+`, in which case the flag seems to acts as if it was disabled
+      not exists(int minusPos |
+        this.modeCharacter(start, minusPos) and this.getChar(minusPos) = "-" and minusPos <= pos
+      )
     )
   }
 
@@ -524,6 +533,8 @@ abstract class RegexString extends StringLiteral {
     exists(string c | this.flag(c) |
       c = "i" and result = "IGNORECASE"
       or
+      c = "d" and result = "UNIXLINES"
+      or
       c = "m" and result = "MULTILINE"
       or
       c = "s" and result = "DOTALL"
@@ -930,13 +941,13 @@ class Regex extends RegexString {
 
   /**
    * Gets a mode (if any) of this regular expression. Can be any of:
-   * DEBUG
-   * IGNORECASE
-   * MULTILINE
-   * DOTALL
-   * UNICODE
-   * VERBOSE
-   * UNICODECLASS
+   * - IGNORECASE
+   * - UNIXLINES
+   * - MULTILINE
+   * - DOTALL
+   * - UNICODE
+   * - VERBOSE
+   * - UNICODECLASS
    */
   string getAMode() {
     result != "None" and
@@ -946,7 +957,7 @@ class Regex extends RegexString {
   }
 
   /**
-   *  Holds if this regex is used to match against a full string,
+   * Holds if this regex is used to match against a full string,
    * as though it was implicitly surrounded by ^ and $.
    */
   predicate matchesFullString() { matches_full_string = true }