Python: Understand multiple parse mode flags specified in a regular expression string

python/ql/lib/change-notes/2023-07-20-regex-parse-modes.md

@@ -0,0 +1,4 @@
+---
+category: minorAnalysis
+---
+* Regular expressions containing multiple parse mode flags are now interpretted correctly. For example `"(?is)abc.*"` with both the `i` and `s` flags.

python/ql/lib/semmle/python/regexp/internal/ParseRegExp.qll

@@ -617,7 +617,7 @@ class RegExp extends Expr instanceof StrConst {
   private predicate group_start(int start, int end) {
     this.non_capturing_group_start(start, end)
     or
-    this.flag_group_start(start, end, _)
+    this.flag_group_start(start, end)
     or
     this.named_group_start(start, end)
     or
@@ -679,20 +679,37 @@ class RegExp extends Expr instanceof StrConst {
     end = min(int i | i > start + 4 and this.getChar(i) = "?")
   }
 
-  private predicate flag_group_start(int start, int end, string c) {
+  /**
+   * Holds if a parse mode starts between `start` and `end`.
+   */
+  private predicate flag_group_start(int start, int end) {
     this.isGroupStart(start) and
     this.getChar(start + 1) = "?" and
-    end = start + 3 and
-    c = this.getChar(start + 2) and
-    c in ["i", "L", "m", "s", "u", "x"]
+    this.getChar(start + 2) in ["i", "L", "m", "s", "u", "x"] and
+    end = start + 2
+  }
+
+  /**
+   * Holds if a parse mode group is between `start` and `end`, and includes the
+   * mode flag `c`. For example the following span, with mode flag `i`:
+   * ```
+   * (?i)
+   * ```
+   */
+  private predicate flag_group(int start, int end, string c) {
+    exists(int inStart, int inEnd |
+      this.flag_group_start(start, inStart) and
+      this.groupContents(start, end, inStart, inEnd) and
+      this.getChar([inStart .. inEnd - 1]) = c
+    )
   }
 
   /**
    * Gets the mode of this regular expression string if
    * it is defined by a prefix.
    */
   string getModeFromPrefix() {
-    exists(string c | this.flag_group_start(_, _, c) |
+    exists(string c | this.flag_group(_, _, c) |
       c = "i" and result = "IGNORECASE"
       or
       c = "L" and result = "LOCALE"

python/ql/test/library-tests/regex/Characters.expected

@@ -36,6 +36,7 @@
 | (?:[^%]\|^)?%\\((\\w*)\\)[a-z] | 22 | 23 |
 | (?:[^%]\|^)?%\\((\\w*)\\)[a-z] | 24 | 25 |
 | (?P<name>[\\w]+)\| | 10 | 12 |
+| (?m)^(?!$) | 2 | 3 |
 | (?m)^(?!$) | 4 | 5 |
 | (?m)^(?!$) | 8 | 9 |
 | (\\033\|~{) | 1 | 5 |

python/ql/test/library-tests/regex/FirstLast.expected

@@ -22,8 +22,7 @@
 | (?P<name>[\\w]+)\| | first | 9 | 14 |
 | (?P<name>[\\w]+)\| | last | 9 | 13 |
 | (?P<name>[\\w]+)\| | last | 9 | 14 |
-| (?m)^(?!$) | first | 4 | 5 |
-| (?m)^(?!$) | first | 8 | 9 |
+| (?m)^(?!$) | first | 2 | 3 |
 | (?m)^(?!$) | last | 4 | 5 |
 | (?m)^(?!$) | last | 8 | 9 |
 | (\\033\|~{) | first | 1 | 5 |

python/ql/test/library-tests/regex/GroupContents.expected

@@ -8,6 +8,7 @@
 | (?:[^%]\|^)?%\\((\\w*)\\)[a-z] | 0 | 10 | (?:[^%]\|^) | 3 | 9 | [^%]\|^ |
 | (?:[^%]\|^)?%\\((\\w*)\\)[a-z] | 14 | 19 | (\\w*) | 15 | 18 | \\w* |
 | (?P<name>[\\w]+)\| | 0 | 15 | (?P<name>[\\w]+) | 9 | 14 | [\\w]+ |
+| (?m)^(?!$) | 0 | 4 | (?m) | 2 | 3 | m |
 | (?m)^(?!$) | 5 | 10 | (?!$) | 8 | 9 | $ |
 | (\\033\|~{) | 0 | 9 | (\\033\|~{) | 1 | 8 | \\033\|~{ |
 | \\[(?P<txt>[^[]*)\\]\\((?P<uri>[^)]*) | 2 | 16 | (?P<txt>[^[]*) | 10 | 15 | [^[]* |

python/ql/test/library-tests/regex/Regex.expected

@@ -77,9 +77,11 @@
 | (?P<name>[\\w]+)\| | sequence | 0 | 15 |
 | (?m)^(?!$) | $ | 8 | 9 |
 | (?m)^(?!$) | ^ | 4 | 5 |
-| (?m)^(?!$) | empty group | 0 | 4 |
+| (?m)^(?!$) | char | 2 | 3 |
 | (?m)^(?!$) | empty group | 5 | 10 |
+| (?m)^(?!$) | non-empty group | 0 | 4 |
 | (?m)^(?!$) | sequence | 0 | 10 |
+| (?m)^(?!$) | sequence | 2 | 3 |
 | (?m)^(?!$) | sequence | 8 | 9 |
 | (\\033\|~{) | char | 1 | 5 |
 | (\\033\|~{) | char | 6 | 7 |

python/ql/test/query-tests/Security/CWE-116-BadTagFilter/BadTagFilter.expected

@@ -1,16 +1,16 @@
 | tst.py:4:20:4:43 | <script.*?>.*?<\\/script> | This regular expression does not match script end tags like </script >. |
 | tst.py:5:20:5:43 | <script.*?>.*?<\\/script> | This regular expression does not match script end tags like </script >. |
 | tst.py:9:20:9:30 | <!--.*--!?> | This regular expression does not match comments containing newlines. |
-| tst.py:10:20:10:53 | <script.*?>(.\|\\s)*?<\\/script[^>]*> | This regular expression matches <script></script>, but not <script \\n></script> |
-| tst.py:11:20:11:51 | <script[^>]*?>.*?<\\/script[^>]*> | This regular expression matches <script>...</script>, but not <script >...\\n</script> |
-| tst.py:12:20:12:58 | <script(\\s\|\\w\|=\|")*?>.*?<\\/script[^>]*> | This regular expression does not match script tags where the attribute uses single-quotes. |
-| tst.py:13:20:13:58 | <script(\\s\|\\w\|=\|')*?>.*?<\\/script[^>]*> | This regular expression does not match script tags where the attribute uses double-quotes. |
-| tst.py:14:20:14:62 | <script( \|\\n\|\\w\|=\|'\|")*?>.*?<\\/script[^>]*> | This regular expression does not match script tags where tabs are used between attributes. |
-| tst.py:15:20:15:48 | <script.*?>.*?<\\/script[^>]*> | This regular expression does not match upper case <SCRIPT> tags. |
-| tst.py:16:20:16:66 | <(script\|SCRIPT).*?>.*?<\\/(script\|SCRIPT)[^>]*> | This regular expression does not match mixed case <sCrIpT> tags. |
-| tst.py:17:20:17:53 | <script[^>]*?>[\\s\\S]*?<\\/script.*> | This regular expression does not match script end tags like </script\\t\\n bar>. |
-| tst.py:19:20:19:54 | <script\\b[^>]*>([\\s\\S]*?)<\\/script> | This regular expression does not match script end tags like </script >. |
-| tst.py:20:20:20:62 | <(?:!--([\\S\|\\s]*?)-->)\|([^\\/\\s>]+)[\\S\\s]*?> | Comments ending with --> are matched differently from comments ending with --!>. The first is matched with capture group 1 and comments ending with --!> are matched with capture group 2. |
-| tst.py:21:20:21:161 | <(?:(?:\\/([^>]+)>)\|(?:!--([\\S\|\\s]*?)-->)\|(?:([^\\/\\s>]+)((?:\\s+[\\w\\-:.]+(?:\\s*=\\s*?(?:(?:"[^"]*")\|(?:'[^']*')\|[^\\s"'\\/>]+))?)*)[\\S\\s]*?(\\/?)>)) | Comments ending with --> are matched differently from comments ending with --!>. The first is matched with capture group 2 and comments ending with --!> are matched with capture group 3, 4. |
-| tst.py:22:17:22:71 | (<[a-z\\/!$]("[^"]*"\|'[^']*'\|[^'">])*>\|<!(--.*?--\\s*)+>) | Comments ending with --> are matched differently from comments ending with --!>. The first is matched with capture group 3 and comments ending with --!> are matched with capture group 1. |
-| tst.py:23:20:23:263 | <(?:(?:!--([\\w\\W]*?)-->)\|(?:!\\[CDATA\\[([\\w\\W]*?)\\]\\]>)\|(?:!DOCTYPE([\\w\\W]*?)>)\|(?:\\?([^\\s\\/<>]+) ?([\\w\\W]*?)[?/]>)\|(?:\\/([A-Za-z][A-Za-z0-9\\-_\\:\\.]*)>)\|(?:([A-Za-z][A-Za-z0-9\\-_\\:\\.]*)((?:\\s+[^"'>]+(?:(?:"[^"]*")\|(?:'[^']*')\|[^>]*))*\|\\/\|\\s+)>)) | This regular expression only parses --> (capture group 1) and not --!> as an HTML comment end tag. |
+| tst.py:12:20:12:53 | <script.*?>(.\|\\s)*?<\\/script[^>]*> | This regular expression matches <script></script>, but not <script \\n></script> |
+| tst.py:13:20:13:51 | <script[^>]*?>.*?<\\/script[^>]*> | This regular expression matches <script>...</script>, but not <script >...\\n</script> |
+| tst.py:14:20:14:58 | <script(\\s\|\\w\|=\|")*?>.*?<\\/script[^>]*> | This regular expression does not match script tags where the attribute uses single-quotes. |
+| tst.py:15:20:15:58 | <script(\\s\|\\w\|=\|')*?>.*?<\\/script[^>]*> | This regular expression does not match script tags where the attribute uses double-quotes. |
+| tst.py:16:20:16:62 | <script( \|\\n\|\\w\|=\|'\|")*?>.*?<\\/script[^>]*> | This regular expression does not match script tags where tabs are used between attributes. |
+| tst.py:17:20:17:48 | <script.*?>.*?<\\/script[^>]*> | This regular expression does not match upper case <SCRIPT> tags. |
+| tst.py:18:20:18:66 | <(script\|SCRIPT).*?>.*?<\\/(script\|SCRIPT)[^>]*> | This regular expression does not match mixed case <sCrIpT> tags. |
+| tst.py:19:20:19:53 | <script[^>]*?>[\\s\\S]*?<\\/script.*> | This regular expression does not match script end tags like </script\\t\\n bar>. |
+| tst.py:21:20:21:54 | <script\\b[^>]*>([\\s\\S]*?)<\\/script> | This regular expression does not match script end tags like </script >. |
+| tst.py:22:20:22:62 | <(?:!--([\\S\|\\s]*?)-->)\|([^\\/\\s>]+)[\\S\\s]*?> | Comments ending with --> are matched differently from comments ending with --!>. The first is matched with capture group 1 and comments ending with --!> are matched with capture group 2. |
+| tst.py:23:20:23:161 | <(?:(?:\\/([^>]+)>)\|(?:!--([\\S\|\\s]*?)-->)\|(?:([^\\/\\s>]+)((?:\\s+[\\w\\-:.]+(?:\\s*=\\s*?(?:(?:"[^"]*")\|(?:'[^']*')\|[^\\s"'\\/>]+))?)*)[\\S\\s]*?(\\/?)>)) | Comments ending with --> are matched differently from comments ending with --!>. The first is matched with capture group 2 and comments ending with --!> are matched with capture group 3, 4. |
+| tst.py:24:17:24:71 | (<[a-z\\/!$]("[^"]*"\|'[^']*'\|[^'">])*>\|<!(--.*?--\\s*)+>) | Comments ending with --> are matched differently from comments ending with --!>. The first is matched with capture group 3 and comments ending with --!> are matched with capture group 1. |
+| tst.py:25:20:25:263 | <(?:(?:!--([\\w\\W]*?)-->)\|(?:!\\[CDATA\\[([\\w\\W]*?)\\]\\]>)\|(?:!DOCTYPE([\\w\\W]*?)>)\|(?:\\?([^\\s\\/<>]+) ?([\\w\\W]*?)[?/]>)\|(?:\\/([A-Za-z][A-Za-z0-9\\-_\\:\\.]*)>)\|(?:([A-Za-z][A-Za-z0-9\\-_\\:\\.]*)((?:\\s+[^"'>]+(?:(?:"[^"]*")\|(?:'[^']*')\|[^>]*))*\|\\/\|\\s+)>)) | This regular expression only parses --> (capture group 1) and not --!> as an HTML comment end tag. |

python/ql/test/query-tests/Security/CWE-116-BadTagFilter/tst.py

@@ -7,6 +7,8 @@
     re.compile(r"""<!--.*-->""", re.IGNORECASE | re.DOTALL), # OK - we don't care regexps that only match comments
     re.compile(r"""<!--.*--!?>""", re.IGNORECASE | re.DOTALL), # OK
     re.compile(r"""<!--.*--!?>""", re.IGNORECASE), # NOT OK, does not match newlines
+    re.compile(r"""(?is)<!--.*--!?>"""), # OK
+    re.compile(r"""(?i)<!--.*--!?>"""), # NOT OK, does not match newlines [NOT DETECTED]
     re.compile(r"""<script.*?>(.|\s)*?<\/script[^>]*>""", re.IGNORECASE), # NOT OK - doesn't match inside the script tag
     re.compile(r"""<script[^>]*?>.*?<\/script[^>]*>""", re.IGNORECASE), # NOT OK - doesn't match newlines inside the content
     re.compile(r"""<script(\s|\w|=|")*?>.*?<\/script[^>]*>""", re.IGNORECASE | re.DOTALL), # NOT OK - does not match single quotes for attribute values
@@ -23,4 +25,4 @@
     re.compile(r"""<(?:(?:!--([\w\W]*?)-->)|(?:!\[CDATA\[([\w\W]*?)\]\]>)|(?:!DOCTYPE([\w\W]*?)>)|(?:\?([^\s\/<>]+) ?([\w\W]*?)[?/]>)|(?:\/([A-Za-z][A-Za-z0-9\-_\:\.]*)>)|(?:([A-Za-z][A-Za-z0-9\-_\:\.]*)((?:\s+[^"'>]+(?:(?:"[^"]*")|(?:'[^']*')|[^>]*))*|\/|\s+)>))"""), # NOT OK - capture groups
 ]
 
-doFilters(filters)
+doFilters(filters)

python/ql/test/query-tests/Security/CWE-730-ReDoS/ReDoS.expected

@@ -105,3 +105,5 @@
 | redos.py:391:15:391:25 | (\\u0061\|a)* | This part of the regular expression may cause exponential backtracking on strings starting with 'X' and containing many repetitions of 'a'. |
 | unittests.py:5:17:5:23 | (\u00c6\|\\\u00c6)+ | This part of the regular expression may cause exponential backtracking on strings starting with 'X' and containing many repetitions of '\u00c6'. |
 | unittests.py:9:16:9:24 | (?:.\|\\n)* | This part of the regular expression may cause exponential backtracking on strings containing many repetitions of '\\n'. |
+| unittests.py:11:20:11:28 | (?:.\|\\n)* | This part of the regular expression may cause exponential backtracking on strings starting with 's' and containing many repetitions of '\\n'. |
+| unittests.py:12:21:12:29 | (?:.\|\\n)* | This part of the regular expression may cause exponential backtracking on strings starting with 'is' and containing many repetitions of '\\n'. |

python/ql/test/query-tests/Security/CWE-730-ReDoS/unittests.py

@@ -7,3 +7,6 @@
 # Treatment of line breaks
 re.compile(r'(?:.|\n)*b') # No ReDoS.
 re.compile(r'(?:.|\n)*b', re.DOTALL) # Has ReDoS.
+re.compile(r'(?i)(?:.|\n)*b') # No ReDoS.
+re.compile(r'(?s)(?:.|\n)*b') # Has ReDoS.
+re.compile(r'(?is)(?:.|\n)*b') # Has ReDoS.