HEX
Server: LiteSpeed
System: Linux spg20.cloudpowerdns.com 5.14.0-611.35.1.el9_7.x86_64 #1 SMP PREEMPT_DYNAMIC Wed Feb 25 03:46:09 EST 2026 x86_64
User: carrerup (3153)
PHP: 8.1.33
Disabled: NONE
Upload Files
File: //opt/bitninja-waf3/coreruleset/util/crs-rules-check/rules-check.py
#!/usr/bin/env python3

import sys
import os
import glob
import msc_pyparser
import difflib
import argparse
import re

oformat = "native"

class Check(object):
    def __init__(self, data, txvars):

        # txvars is a global used hash table, but processing of rules is a sequential flow
        # all rules need this global table
        self.globtxvars  = txvars
        # list available operators, actions, transformations and ctl args
        self.operators   = "beginsWith|containsWord|contains|detectSQLi|detectXSS|endsWith|eq|fuzzyHash|geoLookup|ge|gsbLookup|gt|inspectFile|ipMatch|ipMatchF|ipMatchFromFile|le|lt|noMatch|pmFromFile|pmf|pm|rbl|rsub|rx|streq|strmatch|unconditionalMatch|validateByteRange|validateDTD|validateHash|validateSchema|validateUrlEncoding|validateUtf8Encoding|verifyCC|verifyCPF|verifySSN|within".split("|")
        self.operatorsl  = [o.lower() for o in self.operators]
        self.actions     = "accuracy|allow|append|auditlog|block|capture|chain|ctl|deny|deprecatevar|drop|exec|expirevar|id|initcol|logdata|log|maturity|msg|multiMatch|noauditlog|nolog|pass|pause|phase|prepend|proxy|redirect|rev|sanitiseArg|sanitiseMatched|sanitiseMatchedBytes|sanitiseRequestHeader|sanitiseResponseHeader|setenv|setrsc|setsid|setuid|setvar|severity|skipAfter|skip|status|tag|t|ver|xmlns".split("|")
        self.actionsl    = [a.lower() for a in self.actions]
        self.transforms  = "base64DecodeExt|base64Decode|base64Encode|cmdLine|compressWhitespace|cssDecode|escapeSeqDecode|hexDecode|hexEncode|htmlEntityDecode|jsDecode|length|lowercase|md5|none|normalisePathWin|normalisePath|normalizePathWin|normalizePath|parityEven7bit|parityOdd7bit|parityZero7bit|removeCommentsChar|removeComments|removeNulls|removeWhitespace|replaceComments|replaceNulls|sha1|sqlHexDecode|trimLeft|trimRight|trim|uppercase|urlDecodeUni|urlDecode|urlEncode|utf8toUnicode".split("|")
        self.transformsl = [t.lower() for t in self.transforms]
        self.ctls        = "auditEngine|auditLogParts|debugLogLevel|forceRequestBodyVariable|hashEnforcement|hashEngine|requestBodyAccess|requestBodyLimit|requestBodyProcessor|responseBodyAccess|responseBodyLimit|ruleEngine|ruleRemoveById|ruleRemoveByMsg|ruleRemoveByTag|ruleRemoveTargetById|ruleRemoveTargetByMsg|ruleRemoveTargetByTag".split("|")
        self.ctlsl       = [c.lower() for c in self.ctls]

        # list the actions in expected order
        # see wiki: https://github.com/SpiderLabs/owasp-modsecurity-crs/wiki/Order-of-ModSecurity-Actions-in-CRS-rules
        # note, that these tokens are with lovercase here, but used only for to check the order
        self.ordered_actions = [
            "id",                       # 0
            "phase",                    # 1
            "allow",
            "block",
            "deny",
            "drop",
            "pass",
            "proxy",
            "redirect",
            "status",
            "capture",                  # 10
            "t",
            "log",
            "nolog",
            "auditlog",
            "noauditlog",
            "msg",
            "logdata",
            "tag",
            "sanitisearg",
            "sanitiserequestheader",    # 20
            "sanitisematched",
            "sanitisematchedbytes",
            "ctl",
            "ver",
            "severity",
            "multimatch",
            "initcol",
            "setenv",
            "setvar",
            "expirevar",                # 30
            "chain",
            "skip",
            "skipafter",
        ]

        self.data           = data  # holds the parsed data
        self.current_ruleid = 0     # holds the rule id
        self.curr_lineno    = 0     # current line number
        self.chained        = False # holds the chained flag
        self.caseerror      = []    # list of case mismatch errors
        self.orderacts      = []    # list of ordered action errors
        self.auditlogparts  = []    # list of wrong ctl:auditLogParts
        self.undef_txvars   = []    # list of undefined TX variables
        self.pltags         = []    # list of incosistent PL tags
        self.plscores       = []    # list of incosistent PL scores
        self.dupes          = []    # list of duplicated id's
        self.ids            = {}    # list of rule id's
        self.newtags        = []    # list of new, unlisted tags
        self.ignorecase     = []    # list of combinations of t:lowercase and (?i)

        self.re_tx_var      = re.compile("%\{\}")

    def store_error(self, msg):
        # store the error msg in the list
        self.caseerror.append({
                                'ruleid' : 0,
                                'line'   : self.curr_lineno,
                                'endLine': self.curr_lineno,
                                'message': msg
                            })

    def check_ignore_case(self):
        # check the ignore cases at operators, actions,
        # transformations and ctl arguments
        for d in self.data:
            if "actions" in d:
                aidx = 0        # index of action in list
                if self.chained == False:
                    self.current_ruleid = 0
                else:
                    self.chained = False

                while aidx < len(d['actions']):
                    a = d['actions'][aidx]  # 'a' is the action from the list

                    self.curr_lineno = a['lineno']
                    if a['act_name'] == "id":
                        self.current_ruleid = int(a['act_arg'])

                    if a['act_name'] == "chain":
                        self.chained = True

                    # check the action is valid
                    if a['act_name'].lower() not in self.actionsl:
                        self.store_error("Invalid action", a['act_name'])
                    # check the action case sensitive format
                    if self.actions[self.actionsl.index(a['act_name'].lower())] != a['act_name']:
                        self.store_error("Action case mismatch: %s" % a['act_name'])

                    if a['act_name'] == 'ctl':
                        # check the ctl argument is valid
                        if a['act_arg'].lower() not in self.ctlsl:
                            self.store_error("Invalid ctl", a['act_arg'])
                        # check the ctl argument case sensitive format
                        if self.ctls[self.ctlsl.index(a['act_arg'].lower())] != a['act_arg']:
                            self.store_error("Ctl case mismatch: %s" % a['act_arg'])
                    if a['act_name'] == 't':
                        # check the transform is valid
                        if a['act_arg'].lower() not in self.transformsl:
                            self.store_error("Invalid transform: %s" % a['act_arg'])
                        # check the transform case sensitive format
                        if self.transforms[self.transformsl.index(a['act_arg'].lower())] != a['act_arg']:
                            self.store_error("Transform case mismatch : %s" % a['act_arg'])
                    aidx += 1
            if "operator" in d and d["operator"] != "":
                self.curr_lineno = d['oplineno']
                # strip the operator
                op = d['operator'].replace("!", "").replace("@", "")
                # check the operator is valid
                if op.lower() not in self.operatorsl:
                    self.store_error("Invalid operator: %s" % d['operator'])
                # check the operator case sensitive format
                if self.operators[self.operatorsl.index(op.lower())] != op:
                    self.store_error("Operator case mismatch: %s" % d['operator'])
            else:
                if d['type'].lower() == "secrule":
                    self.curr_lineno = d['lineno']
                    self.store_error("Empty operator isn't allowed")
            if self.current_ruleid > 0:
                for e in self.caseerror:
                    e['ruleid'] = self.current_ruleid
                    e['message'] += " (rule: %d)" % (self.current_ruleid)

    def check_action_order(self):
        for d in self.data:
            if "actions" in d:
                aidx = 0        # stores the index of current action
                max_order = 0   # maximum position of read actions
                if self.chained == False:
                    self.current_ruleid = 0
                else:
                    self.chained = False

                while aidx < len(d['actions']):
                    # read the action into 'a'
                    a = d['actions'][aidx]

                    # get the 'id' of rule
                    self.curr_lineno = a['lineno']
                    if a['act_name'] == "id":
                        self.current_ruleid = int(a['act_arg'])

                    # check if chained
                    if a['act_name'] == "chain":
                        self.chained = True

                    # get the index of action from the ordered list
                    # above from constructor
                    try:
                        act_idx = self.ordered_actions.index(a['act_name'].lower())
                    except ValueError:
                        print("ERROR: '%s' not in actions list!" % (a['act_name']))
                        sys.exit(-1)

                    # if the index of current action is @ge than the previous
                    # max value, load it into max_order
                    if act_idx >= max_order:
                        max_order = act_idx
                    else:
                        # prevact is the previous action's position in list
                        # act_idx is the current action's position in list
                        # if the prev is @gt actually, means it's at wrong position
                        if self.ordered_actions.index(prevact) > act_idx:
                            self.orderacts.append({
                                'ruleid' : 0,
                                'line'   : a['lineno'],
                                'endLine': a['lineno'],
                                'message': "action '%s' at pos %d is wrong place against '%s' at pos %d" % (prevact, pidx, a['act_name'], aidx,)
                            })
                    prevact = a['act_name'].lower()
                    pidx = aidx
                    aidx += 1
                for a in self.orderacts:
                    if a['ruleid'] == 0:
                        a['ruleid'] = self.current_ruleid
                        a['message'] += " (rule: %d)" % (self.current_ruleid)

    def check_ctl_audit_log(self):
        """check there is no ctl:auditLogParts action in any rules"""
        for d in self.data:
            if "actions" in d:
                aidx = 0        # stores the index of current action

                while aidx < len(d['actions']):
                    # read the action into 'a'
                    a = d['actions'][aidx]

                    # get the 'id' of rule
                    self.curr_lineno = a['lineno']
                    if a['act_name'] == "id":
                        self.current_ruleid = int(a['act_arg'])

                    # check if action is ctl:auditLogParts
                    if a['act_name'].lower() == "ctl" and a['act_arg'].lower() == "auditlogparts":
                        self.auditlogparts.append({
                                'ruleid' : self.current_ruleid,
                                'line'   : a['lineno'],
                                'endLine': a['lineno'],
                                'message': ""
                        })

                    aidx += 1

    def collect_tx_variable(self, fname):
        """collect TX variables in rules
        this function collects the TX variables at rules,
        if the variable is at a 'setvar' action's left side, eg
        setvar:tx.foo=bar

        Because this rule called before any other check,
        additionally it checks the duplicated rule ID
        """
        chained = False
        for d in self.data:
            if "actions" in d:
                aidx = 0        # stores the index of current action
                if chained == False:
                    ruleid = 0      # ruleid
                    phase = 2       # works only in Apache, libmodsecurity uses default phase 1
                else:
                    chained = False
                while aidx < len(d['actions']):
                    # read the action into 'a'
                    a = d['actions'][aidx]
                    if a['act_name'] == "id":
                        ruleid = int(a['act_arg'])
                        if ruleid in self.ids:
                            self.dupes.append({
                                'ruleid' : ruleid,
                                'line'   : a['lineno'],
                                'endLine': a['lineno'],
                                'message': "id %d is duplicated, previous place: %s:%d" % (ruleid, self.ids[ruleid]['fname'], self.ids[ruleid]['lineno'])
                            })
                        else:
                            self.ids[ruleid] = {'fname': fname, 'lineno': a['lineno']}
                    if a['act_name'] == "phase":
                        phase = int(a['act_arg'])
                    if a['act_name'] == "chain":
                        chained = True
                    if a['act_name'] == "setvar":
                        if a['act_arg'][0:2].lower() == "tx":
                            txv = a['act_arg'][3:].split("=")
                            txv[0] = txv[0].lower()
                            # set TX variable if there is no such key
                            # OR
                            # key exists but the existing struct's phase is higher
                            if (txv[0] not in self.globtxvars or self.globtxvars[txv[0]]['phase'] > phase) and \
                               not re.search("%\{[^%]+\}", txv[0]):
                                self.globtxvars[txv[0]] = {
                                    'phase'  : phase,
                                    'used'   : False,
                                    'file'   : fname,
                                    'ruleid' : ruleid,
                                    'message': "",
                                    'line'   : a['lineno'],
                                    'endLine': a['lineno']
                                }
                            else:
                                pass
                    aidx += 1

    def check_tx_variable(self, fname):
        """this function checks if a used TX variable has set

        a variable is used when:
          * it's an operator argument: "@rx %{TX.foo}"
          * it's a target: SecRule TX.foo "@..."
          * it's a right side value in a value giving: setvar:tx.bar=tx.foo

        this function collects the variables if it is used but not set previously
        """
        check_exists   = None   # set if rule checks the existence of varm eg `&TX:foo "@eq 1"`
        has_disruptive = False  # set if rule contains disruptive action
        chained = False
        for d in self.data:
            if d['type'].lower() in ["secrule", "secaction"]:
                aidx = 0        # stores the index of current action
                if chained == False:
                    phase = 2       # works only in Apache, libmodsecurity uses default phase 1
                    ruleid = 0
                else:
                    chained = False

                # iterate over actions and collect these values:
                # ruleid, phase, chained, rule has or not any disruptive action
                while aidx < len(d['actions']):
                    # read the action into 'a'
                    a = d['actions'][aidx]
                    if a['act_name'] == "id":
                        ruleid = int(a['act_arg'])
                    if a['act_name'] == "phase":
                        phase = int(a['act_arg'])
                    if a['act_name'] == "chain":
                        chained = True
                    if a['act_name'] in ['block', 'deny', 'drop', 'allow', 'proxy', 'redirect']:
                        has_disruptive = True

                    # check wheter tx.var is used at setvar's right side
                    val_act = []
                    val_act_arg = []
                    # example:
                    #    setvar:'tx.inbound_anomaly_score_threshold=5'
                    #
                    #  act_arg     <- tx.inbound_anomaly_score_threshold
                    #  act_atg_val <- 5
                    #
                    # example2 (same as above, but no single quotes!):
                    #    setvar:tx.inbound_anomaly_score_threshold=5
                    #  act_arg     <- tx.inbound_anomaly_score_threshold
                    #  act_atg_val <- 5
                    #
                    if "act_arg" in a and a['act_arg'] is not None:
                        val_act = re.findall("%\{(tx.[^%]*)\}", a['act_arg'], re.I)
                    if "act_arg_val" in a and a['act_arg_val'] is not None:
                        val_act_arg = re.findall("%\{(tx.[^%]*)\}", a['act_arg_val'], re.I)
                    for v in val_act + val_act_arg:
                        v = v.lower().replace("tx.", "")
                        # check whether the variable is a captured var, eg TX.1 - we do not care that case
                        if not re.match("^\d$", v, re.I):
                            # v holds the tx.ANY variable, but not the captured ones
                            # we should collect these variables
                            if (v not in self.globtxvars or phase < self.globtxvars[v]['phase']):
                                self.undef_txvars.append({
                                    'var'    : v,
                                    'ruleid' : ruleid,
                                    'line'   : a['lineno'],
                                    'endLine': a['lineno'],
                                    'message': "TX variable '%s' not set / later set (rvar) in rule %d" % (v, ruleid)
                                })
                            else:
                                self.globtxvars[v]['used'] = True
                        else:
                            if v in self.globtxvars:
                                self.globtxvars[v]['used'] = True
                    aidx += 1

                if "operator_argument" in d:
                    oparg = re.findall("%\{(tx.[^%]*)\}", d['operator_argument'], re.I)
                    if oparg:
                        for o in oparg:
                            o = o.lower()
                            o = re.sub("tx\.", "", o, re.I)
                            if (o not in self.globtxvars or phase < self.globtxvars[o]['phase']) and \
                              not re.match("^\d$", o) and \
                              not re.match("\/.*\/", o) and \
                              check_exists is None:
                                self.undef_txvars.append({
                                    'var'    : o,
                                    'ruleid' : ruleid,
                                    'line'   : d['lineno'],
                                    'endLine': d['lineno'],
                                    'message': "TX variable '%s' not set / later set (OPARG) in rule %d" % (o, ruleid)
                                })
                            elif o in self.globtxvars and phase >= self.globtxvars[o]['phase'] and \
                                not re.match("^\d$", o) and \
                                not re.match("\/.*\/", o):
                                    self.globtxvars[o]['used'] = True
                if "variables" in d:
                    for v in d['variables']:
                        # check if the variable is TX and has not a & prefix, which counts
                        # the variable length
                        if v['variable'].lower() == "tx":
                            if v['counter'] != True:
                                # * if the variable part (after '.' or ':') is not there in
                                #   the list of collected TX variables, and
                                # * not a numeric, eg TX:2, and
                                # * not a regular expression, between '/' chars, eg TX:/^foo/
                                # OR
                                # * rule's phase lower than declaration's phase
                                rvar = v['variable_part'].lower()
                                if (rvar not in self.globtxvars or (ruleid != self.globtxvars[rvar]['ruleid'] and phase < self.globtxvars[rvar]['phase'])) and \
                                  not re.match("^\d$", rvar) and \
                                  not re.match("\/.*\/", rvar):
                                    self.undef_txvars.append({
                                        'var'    : rvar,
                                        'ruleid' : ruleid,
                                        'line'   : d['lineno'],
                                        'endLine': d['lineno'],
                                        'message': "TX variable '%s' not set / later set (VAR)" % (v['variable_part'])
                                    })
                                elif rvar in self.globtxvars and phase >= self.globtxvars[rvar]['phase'] and \
                                    not re.match("^\d$", rvar) and \
                                    not re.match("\/.*\/", rvar):
                                        self.globtxvars[rvar]['used'] = True
                            else:
                                check_exists = True
                                self.globtxvars[v['variable_part'].lower()] = {
                                    'var'    : v['variable_part'].lower(),
                                    'phase'  : phase,
                                    'used'   : False,
                                    'file'   : fname,
                                    'ruleid' : ruleid,
                                    'message': "",
                                    'line'   : d['lineno'],
                                    'endLine': d['lineno']
                                }
                                if has_disruptive == True:
                                    self.globtxvars[v['variable_part'].lower()]['used'] = True
                                if len(self.undef_txvars) > 0 and self.undef_txvars[-1]['var'] == v['variable_part'].lower():
                                    del(self.undef_txvars[-1])
                if chained == False:
                    check_exists   = None
                    has_disruptive = False

    def check_pl_consistency(self):
        """this method checks the PL consistency

        the function iterates through the rules, and catches the set PL, eg:

        SecRule TX:DETECTION_PARANOIA_LEVEL "@lt 1" ...
        this means we are on PL1 currently

        all rules must consist with current PL at the used tags and variables

        eg:
            tag:'paranoia-level/1'
                                ^
            setvar:'tx.outbound_anomaly_score_pl1=+%{tx.error_anomaly_score}'"
                                              ^^^
        additional relations:
        * all rules must have the "tag:'paranoia-level/N'" if it does not have "nolog" action
        * if rule have "nolog" action it must not have "tag:'paranoia-level/N'" action
        * anomaly scoring value on current PL must increment by value corresponding to severity

        """
        curr_pl   = 0
        tags      = []       # collect tags
        _txvars   = {}       # collect setvars and values
        _txvlines = {}       # collect setvars and its lines
        severity  = None     # severity
        has_nolog = False    # nolog action exists

        for d in self.data:
            # find the current PL
            if d['type'].lower() in ["secrule"]:
                for v in d['variables']:
                    if v['variable'].lower() == "tx" and \
                       v['variable_part'].lower() == "detection_paranoia_level" and \
                       d['operator'] == "@lt" and re.match("^\d$", d['operator_argument']):
                            curr_pl = int(d['operator_argument'])

            if "actions" in d:
                aidx     = 0    # stores the index of current action
                chained  = False
                while aidx < len(d['actions']):
                    # read the action into 'a'
                    a = d['actions'][aidx]
                    if a['act_name'] == "id":
                        ruleid = int(a['act_arg'])
                    if a['act_name'] == "severity":
                        severity = a['act_arg'].replace("'", "").lower()
                    if a['act_name'] == "tag":
                        tags.append(a)
                    if a['act_name'] == "setvar":
                        if a['act_arg'][0:2].lower() == "tx":
                            # this hack necessary, because sometimes we use setvar argument
                            # between '', sometimes not
                            # eg
                            # setvar:crs_setup_version=334
                            # setvar:'tx.inbound_anomaly_score_threshold=5'
                            txv = a['act_arg'][3:].split("=")
                            txv[0] = txv[0].lower()                     # variable name
                            if len(txv) > 1:
                                txv[1] = txv[1].lower().strip("+\{\}")  # variable value
                            else:
                                txv.append(a['act_arg_val'].strip("+\{\}"))
                            _txvars[txv[0]] = txv[1]
                            _txvlines[txv[0]] = a['lineno']
                    if a['act_name'] == "nolog":
                        has_nolog = True
                    if a['act_name'] == "chain":
                        chained = True
                    aidx += 1

                has_pl_tag = False
                for a in tags:
                    if a['act_arg'][0:14] == "paranoia-level":
                        has_pl_tag = True
                        pltag = int(a['act_arg'].split("/")[1])
                        if has_nolog:
                            self.pltags.append({
                                'ruleid' : ruleid,
                                'line'   : a['lineno'],
                                'endLine': a['lineno'],
                                'message': "tag '%s' with 'nolog' action, rule id: %d" % (a['act_arg'], ruleid)
                            })
                        elif pltag != curr_pl and curr_pl > 0:
                            self.pltags.append({
                                'ruleid' : ruleid,
                                'line'   : a['lineno'],
                                'endLine': a['lineno'],
                                'message': "tag '%s' on PL %d, rule id: %d" % (a['act_arg'], curr_pl, ruleid)
                            })

                if has_pl_tag != True and has_nolog == False and curr_pl >= 1:
                    self.pltags.append({
                        'ruleid' : ruleid,
                        'line'   : a['lineno'],
                        'endLine': a['lineno'],
                        'message': "rule does not have `paranoia-level/%d` action, rule id: %d" % (curr_pl, ruleid)
                    })

                for t in _txvars:
                    subst_val = re.search("%{tx.[a-z]+_anomaly_score}", _txvars[t], re.I)
                    val = re.sub("[\+\%\{\}]", "", _txvars[t]).lower()
                    scorepl = re.search("anomaly_score_pl\d$", t)   # check if last char is a numeric, eg ...anomaly_score_pl1
                    if scorepl:
                        if curr_pl > 0 and int(t[-1]) != curr_pl:
                            self.plscores.append({
                                'ruleid' : ruleid,
                                'line'   : _txvlines[t],
                                'endLine': _txvlines[t],
                                'message': "variable %s on PL %d, rule id: %d" % (t, curr_pl, ruleid)
                            })
                        if severity is None and subst_val: # - do we need this?
                            self.plscores.append({
                                'ruleid' : ruleid,
                                'line'   : _txvlines[t],
                                'endLine': _txvlines[t],
                                'message': "missing severity action, rule id: %d" % (ruleid)
                            })
                        else:
                            if val != 'tx.%s_anomaly_score' % (severity) and val != "0":
                                self.plscores.append({
                                    'ruleid' : ruleid,
                                    'line'   : _txvlines[t],
                                    'endLine': _txvlines[t],
                                    'message': "invalid value for anomaly_score_pl%d: %s with severity %s, rule id: %d" % (int(t[-1]), val, severity, ruleid)
                                })
                        # variable has found so we need to mark it as used
                        self.globtxvars[t]['used'] = True

                # reset local variables if we are done with a rule <==> no more 'chain' action
                if chained == False:
                    tags      = []       # collect tags
                    _txvars   = {}       # collect setvars and values
                    _txvlines = {}       # collect setvars and its lines
                    severity  = None     # severity
                    has_nolog = False    # rule has nolog action

    def check_tags(self, fname, tagslist):
        """
        check that only tags from the util/APPROVED_TAGS file are used
        """
        chained = False
        ruleid = 0
        for d in self.data:
            if "actions" in d:
                aidx = 0        # stores the index of current action
                if chained == False:
                    ruleid  = 0
                else:
                    chained = False
                while aidx < len(d['actions']):
                    # read the action into 'a'
                    a = d['actions'][aidx]
                    if a['act_name'] == "id":
                        ruleid = int(a['act_arg'])
                    if a['act_name'] == "chain":
                        chained = True
                    if a['act_name'] == "tag":
                        # check wheter tag is in tagslist
                        if tagslist.count(a['act_arg']) == 0:
                            self.newtags.append({
                                'ruleid' : ruleid,
                                'line'   : a['lineno'],
                                'endLine': a['lineno'],
                                'message': "rule uses unknown tag: '%s'; only tags registered in the util/APPROVED_TAGS file may be used; rule id: %d" % (a['act_arg'], ruleid)
                            })
                    aidx += 1

    def check_lowercase_ignorecase(self):
        ruleid = 0 
        for d in self.data:
            if d['type'].lower() == "secrule":
                if d['operator'] == "@rx":
                    regex = d['operator_argument']
                    if regex.startswith("(?i)"):
                        if "actions" in d:
                            aidx = 0        # stores the index of current action
                            while aidx < len(d['actions']):
                                # read the action into 'a'
                                a = d['actions'][aidx]               
                                if a['act_name'] == "id":
                                    ruleid = int(a['act_arg'])
                                if a['act_name'] == 't':
                                    # check the transform is valid
                                    if a['act_arg'].lower() == "lowercase":
                                        self.ignorecase.append({
                                            'ruleid' : ruleid,
                                            'line'   : a['lineno'],
                                            'endLine': a['lineno'],
                                            'message': "rule uses (?i) in combination with t:lowercase: '%s'; rule id: %d" % (a['act_arg'], ruleid)
                            })
                                aidx += 1

def remove_comments(data):
    """
    In some special cases, remove the comments from the beginning of the lines.

    A special case starts when the line has a "SecRule" or "SecAction" token at
    the beginning and ends when the line - with or without a comment - is empty.

    Eg.:
    175	# Uncomment this rule to change the default:
    176	#
    177	#SecAction \
    178	#    "id:900000,\
    179	#    phase:1,\
    180	#    pass,\
    181	#    t:none,\
    182	#    nolog,\
    183	#    setvar:tx.blocking_paranoia_level=1"
    184
    185
    186	# It is possible to execute rules from a higher paranoia level but not include

    In this case, the comments from the beginning of lines 177 and 183 are deleted and
    evaluated as follows:

    175	# Uncomment this rule to change the default:
    176	#
    177	SecAction \
    178	    "id:900000,\
    179	    phase:1,\
    180	    pass,\
    181	    t:none,\
    182	    nolog,\
    183	    setvar:tx.blocking_paranoia_level=1"
    184
    185
    186	# It is possible to execute rules from a higher paranoia level but not include

    """
    _data = []  # new structure by lines
    lines = data.split("\n")
    marks = re.compile("^#(| *)(SecRule|SecAction)", re.I) # regex what catches the rules
    state = 0   # hold the state of the parser
    for l in lines:
        # if the line starts with #SecRule, #SecAction, # SecRule, # SecAction, set the marker
        if marks.match(l):
            state = 1
        # if the marker is set and the line is empty or contains only a comment, unset it
        if state == 1 and l.strip() in ["", "#"]:
            state = 0

        # if marker is set, remove the comment
        if state == 1:
            _data.append(re.sub("^#", "", l))
        else:
            _data.append(l)

    data = "\n".join(_data)

    return data

def errmsg(msg):
    if oformat == "github":
        print("::error %s" % (msg))
    else:
        print(msg)

def errmsgf(msg):
    if oformat == "github":
        if 'message' in msg and msg['message'].strip() != "":
            print("::error%sfile={file},line={line},endLine={endLine},title={title}: {message}".format(**msg) % (msg['indent']*" "))
        else:
            print("::error%sfile={file},line={line},endLine={endLine},title={title}".format(**msg) % (msg['indent']*" "))
    else:
        if 'message' in msg and msg['message'].strip() != "":
            print("%sfile={file}, line={line}, endLine={endLine}, title={title}: {message}".format(**msg) % (msg['indent']*" "))
        else:
            print("%sfile={file}, line={line}, endLine={endLine}, title={title}".format(**msg) % (msg['indent']*" "))

def msg(msg):
    if oformat == "github":
        print("::debug %s" % (msg))
    else:
        print(msg)

if __name__ == "__main__":
    parser = argparse.ArgumentParser(description="CRS Rules Check tool")
    parser.add_argument("-o", "--output", dest="output", help="Output format native[default]|github", required=False)
    parser.add_argument("-r", "--rules", metavar='/path/to/coreruleset/*.conf', type=str,
                            nargs='*', help='Directory path to CRS rules', required=True,
                            action="append")
    parser.add_argument("-t", "--tags-list", dest="tagslist", help="Path to file with permitted tags", required=True)
    args = parser.parse_args()

    crspath = []
    for l in args.rules:
        crspath += l

    if args.output is not None:
        if args.output not in ["native", "github"]:
            print("--output can be one of the 'native' or 'github'. Default value is 'native'")
            sys.exit(1)
    oformat = args.output

    tags = []
    try:
        with open(args.tagslist, "r") as fp:
            tags = [l.strip() for l in fp.readlines()]
            # remove empty items, if any
            tags = list(filter(lambda x: len(x) > 0, tags))
    except:
        errmsg("Can't open tags list: %s" % args.tagslist)
        sys.exit(1)

    retval = 0
    try:
        flist = crspath
        flist.sort()
    except:
        errmsg("Can't open files in given path!")
        sys.exit(1)

    if len(flist) == 0:
        errmsg("List of files is empty!")
        sys.exit(1)

    parsed_structs = {}
    txvars         = {}

    for f in flist:
        try:
            with open(f, 'r') as inputfile:
                data = inputfile.read()
                # modify the content of the file, if it is the "crs-setup.conf.example"
                if f.startswith("crs-setup.conf.example"):
                    data = remove_comments(data)
        except:
            errmsg("Can't open file: %s" % f)
            sys.exit(1)

        ### check file syntax
        msg("Config file: %s" % (f))
        try:
            mparser = msc_pyparser.MSCParser()
            mparser.parser.parse(data)
            msg(" Parsing ok.")
            parsed_structs[f] = mparser.configlines
        except Exception as e:
            err = e.args[1]
            if err['cause'] == "lexer":
                cause = "Lexer"
            else:
                cause = "Parser"
            errmsg("Can't parse config file: %s" % (f))
            errmsgf({
                'indent' : 2,
                'file'   : f,
                'title'  : "%s error" % (cause),
                'line'   : err['line'],
                'endLine': err['line'],
                'message': "can't parse file"})
            retval = 1
            continue

    msg("Checking parsed rules...")
    for f in parsed_structs.keys():

        msg(f)
        c = Check(parsed_structs[f], txvars)

        ### check case usings
        c.check_ignore_case()
        if len(c.caseerror) == 0:
            msg(" Ignore case check ok.")
        else:
            errmsg(" Ignore case check found error(s)")
            for a in c.caseerror:
                a['indent'] = 2
                a['file']   = f
                a['title']  = "Case check"
                errmsgf(a)
                retval = 1

        ### check action's order
        c.check_action_order()
        if len(c.orderacts) == 0:
            msg(" Action order check ok.")
        else:
            errmsg(" Action order check found error(s)")
            for a in c.orderacts:
                a['indent'] = 2
                a['file']   = f
                a['title']  = 'Action order check'
                errmsgf(a)
                retval = 1

        ### make a diff to check the indentations
        try:
            with open(f, 'r') as fp:
                fromlines = fp.readlines()
                if f.startswith("crs-setup.conf.example"):
                    fromlines = remove_comments("".join(fromlines)).split("\n")
                    fromlines = [l + "\n" for l in fromlines]
        except:
            errmsg("  Can't open file for indent check: %s" % (f))
            retval = 1
        # virtual output
        mwriter = msc_pyparser.MSCWriter(parsed_structs[f])
        mwriter.generate()
        #mwriter.output.append("")
        output = []
        for l in mwriter.output:
            if l == "\n":
                output.append("\n")
            else:
                output += [l + "\n" for l in l.split("\n")]

        if len(fromlines) < len(output):
            fromlines.append("\n")
        elif len(fromlines) > len(output):
            output.append("\n")

        diff = difflib.unified_diff(fromlines, output)
        if fromlines == output:
            msg(" Indentation check ok.")
        else:
            errmsg(" Indentation check found error(s)")
            retval = 1
        for d in diff:
            d = d.strip("\n")
            r = re.match("^@@ -(\d+),(\d+) \+\d+,\d+ @@$", d)
            if r:
                line1, line2 = [int(i) for i in r.groups()]
                e = {
                    'indent' : 2,
                    'file'   : f,
                    'title'  : "Indentation error",
                    'line'   : line1,
                    'endLine': line1+line2,
                    'message': "an indentation error has found"
                }
                errmsgf(e)
            errmsg(d.strip("\n"))

        ### check `ctl:auditLogParts=+E` right place in chained rules
        c.check_ctl_audit_log()
        if len(c.auditlogparts) == 0:
            msg(" no 'ctl:auditLogParts' action found.")
        else:
            errmsg(" Found 'ctl:auditLogParts' action")
            for a in c.auditlogparts:
                a['indent'] = 2
                a['file']   = f
                a['title']  = "'ctl:auditLogParts' isn't allowed in CRS"
                errmsgf(a)
                retval = 1

        ### collect TX variables
        #   this method collects the TX variables, which set via a
        #   `setvar` action anywhere
        #   this method does not check any mandatory clause
        c.collect_tx_variable(f)

        ### check duplicate ID's
        #   c.dupes filled during the tx variable collected
        if len(c.dupes) == 0:
            msg(" no duplicate id's")
        else:
            errmsg(" Found duplicated id('s)")
            for a in c.dupes:
                a['indent'] = 2
                a['file']   = f
                a['title']  = "'id' is duplicated"
                errmsgf(a)
                retval = 1

        ### check PL consistency
        c.check_pl_consistency()
        if len(c.pltags) == 0:
            msg(" paranoia-level tags are correct.")
        else:
            errmsg(" Found incorrect paranoia-level/N tag(s)")
            for a in c.pltags:
                a['indent'] = 2
                a['file']   = f
                a['title']  = "wrong or missing paranoia-level/N tag"
                errmsgf(a)
                retval = 1
        if len(c.plscores) == 0:
            msg(" PL anomaly_scores are correct.")
        else:
            errmsg(" Found incorrect (inbound|outbout)_anomaly_score value(s)")
            for a in c.plscores:
                a['indent'] = 2
                a['file']   = f
                a['title']  = "wrong (inbound|outbout)_anomaly_score variable or value"
                errmsgf(a)
                retval = 1

        ### check existence of used TX variables
        c.check_tx_variable(f)
        if len(c.undef_txvars) == 0:
            msg(" All TX variables are set.")
        else:
            errmsg(" There are one or more unset TX variables.")
            for a in c.undef_txvars:
                a['indent'] = 2
                a['file']   = f
                a['title']  = "unset TX variable"
                errmsgf(a)
                retval = 1
        ### check new unlisted tags
        c.check_tags(f, tags)
        if len(c.newtags) == 0:
            msg(" No new tags added.")
        else:
            errmsg(" There are one or more new tag(s).")
            for a in c.newtags:
                a['indent'] = 2
                a['file']   = f
                a['title']  = "new unlisted tag"
                errmsgf(a)
                retval = 1
        ### check for t:lowercase in combination with (?i) in regex       
        c.check_lowercase_ignorecase()
        if len(c.ignorecase) == 0:
            msg(" No t:lowercase and (?i) flag used.")
        else:
            errmsg(" There are one or more combinations of t:lowercase and (?i) flag.")
            for a in c.ignorecase:
                a['indent'] = 2
                a['file']   = f
                a['title']  = "t:lowercase and (?i)"
                errmsgf(a)
                retval = 1

    msg("End of checking parsed rules")
    msg("Cumulated report about unused TX variables")
    has_unused = False
    for tk in txvars:
        if txvars[tk]['used'] == False:
            if has_unused == False:
                msg(" Unused TX variable(s):")
            a = txvars[tk]
            a['indent'] = 2
            a['title']  = "unused TX variable"
            a['message'] = "unused variable: %s" % (tk)
            errmsgf(a)
            retval = 1
            has_unused = True

    if has_unused == False:
        msg(" No unused TX variable")

    sys.exit(retval)