1berry/qtbase
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

QRegularExpression: allow users to skip the UTF-16 check of the subject string

Browse Source 
PCRE does not handle invalid UTF-16 sequences. For this reason we always
check a subject string's UTF-16 validity before attempting any match
over it (actually we let PCRE do that).

The only exception so far has been global matching -- once the first
match was done, we skipped re-doing the check over and over again the
same string (PCRE actually checks the /entire/ string, not only the part
it uses for matching).

Still, users had no way to skip this check if they were 100% sure the
string was a valid UTF-16 string. This commit introduces a way for them
to skip the check.

Change-Id: Iea352c06f531aa2153863b3a1681acaab7ac375c
Reviewed-by: Thiago Macieira <thiago.macieira@intel.com>
This commit is contained in:
Giuseppe D'Angelo 2014-05-11 11:48:47 +02:00 committed by The Qt Project
parent 4532669285
commit fd80cad07e
2 changed files with 14 additions and 3 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

14
src/corelib/tools/qregularexpression.cpp
View File

@ -763,6 +763,13 @@ QT_BEGIN_NAMESPACE
The match is constrained to start exactly at the offset passed to The match is constrained to start exactly at the offset passed to
match() in order to be successful, even if the pattern string does not match() in order to be successful, even if the pattern string does not
contain any metacharacter that anchors the match at that point. contain any metacharacter that anchors the match at that point.
\value DontCheckSubjectStringMatchOption
The subject string is not checked for UTF-16 validity before
attempting a match. Use this option with extreme caution, as
attempting to match an invalid string may crash the program and/or
constitute a security issue. This enum value has been introduced in
Qt 5.4.
*/ */
// after how many usages we optimize the regexp // after how many usages we optimize the regexp
@ -1221,7 +1228,8 @@ static int pcre16SafeExec(const pcre16 *code, const pcre16_extra *extra,
options \a matchOptions and returns the QRegularExpressionMatchPrivate of options \a matchOptions and returns the QRegularExpressionMatchPrivate of
the result. It also advances a match if a previous result is given as \a the result. It also advances a match if a previous result is given as \a
previous. The \a subject string goes a Unicode validity check if previous. The \a subject string goes a Unicode validity check if
\a checkSubjectString is CheckSubjectString (PCRE doesn't like illegal \a checkSubjectString is CheckSubjectString and the match options don't
include DontCheckSubjectStringMatchOption (PCRE doesn't like illegal
UTF-16 sequences). UTF-16 sequences).
Advancing a match is a tricky algorithm. If the previous match matched a Advancing a match is a tricky algorithm. If the previous match matched a
@ -1290,8 +1298,10 @@ QRegularExpressionMatchPrivate *QRegularExpressionPrivate::doMatch(const QString
else if (matchType == QRegularExpression::PartialPreferFirstMatch) else if (matchType == QRegularExpression::PartialPreferFirstMatch)
pcreOptions |= PCRE_PARTIAL_HARD; pcreOptions |= PCRE_PARTIAL_HARD;
if (checkSubjectStringOption == DontCheckSubjectString) if (checkSubjectStringOption == DontCheckSubjectString
|| matchOptions & QRegularExpression::DontCheckSubjectStringMatchOption) {
pcreOptions |= PCRE_NO_UTF16_CHECK; pcreOptions |= PCRE_NO_UTF16_CHECK;
}
bool previousMatchWasEmpty = false; bool previousMatchWasEmpty = false;
if (previous && previous->hasMatch && if (previous && previous->hasMatch &&

3
src/corelib/tools/qregularexpression.h
View File

@ -110,7 +110,8 @@ public:
enum MatchOption { enum MatchOption {
NoMatchOption = 0x0000, NoMatchOption = 0x0000,
AnchoredMatchOption = 0x0001 AnchoredMatchOption = 0x0001,
DontCheckSubjectStringMatchOption = 0x0002
}; };
Q_DECLARE_FLAGS(MatchOptions, MatchOption) Q_DECLARE_FLAGS(MatchOptions, MatchOption)