[llvm-bugs] [Bug 26175] New: regex does not throw an exception for an ECMAScript pattern ending in a trailing backslash

[via llvm-bugs]

https://llvm.org/bugs/show_bug.cgi?id=26175

            Bug ID: 26175
           Summary: regex does not throw an exception for an ECMAScript
                    pattern ending in a trailing backslash
           Product: libc++
           Version: unspecified
          Hardware: Macintosh
                OS: MacOS X
            Status: NEW
          Severity: normal
          Priority: P
         Component: All Bugs
          Assignee: unassignedclangbugs at nondot.org
          Reporter: josh.petrie at gmail.com
                CC: llvm-bugs at lists.llvm.org, mclow.lists at gmail.com
    Classification: Unclassified

Created attachment 15643
  --> https://llvm.org/bugs/attachment.cgi?id=15643&action=edit
Changes I made to <regex> to produce the behavior I "think" is correct.

When constructing a std::regex object with the pattern "\\" or R"(\)" (that is,
a single backslash), no exception is thrown. I would expect a std::regex_error
to be thrown with the "error_escape" error code, because as far as I have been
able to determine a trailing backslash is not a valid pattern.

I first noticed this in Xcode 7.2 (7C68), compiling with C++14 enabled and the
system libc++ (on OS X 10.11.2). However, I was able to reproduce the issue
using a libc++ built from source (SVN revision 257997). 

The following program demonstrates the issue:

#include <iostream>
#include <regex>

int main () {
  try {
    std::regex example("\\");
  } catch (const std::regex_error & error) {
    // Never executes, but probably should:
    std::cout << "code: " << error.code() << "\n";
  }
}

It appears that __parse_atom_escape is the problematic function; it *seems*
that __parse_atom_escape should look something like:

template <class _CharT, class _Traits>
template <class _ForwardIterator>
_ForwardIterator
basic_regex<_CharT, _Traits>::__parse_atom_escape(_ForwardIterator __first,
                                                  _ForwardIterator __last)
{
    if (__first != __last && *__first == '\\')
    {
        _ForwardIterator __t1 = _VSTD::next(__first);
        if (__t1 == __last)                                   //!
            throw regex_error(regex_constants::error_escape); //!

        _ForwardIterator __t2 = __parse_decimal_escape(__t1, __last);
        ...

Where the additional two lines, marked with //!, check if advancing __first has
run in to the end of the expression, and if so throws the appropriate error.

(However my experience in hacking on the standard library is limited to
precisely the above, so I might be very, very off-base with this suggestion.)

-- 
You are receiving this mail because:
You are on the CC list for the bug.
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.llvm.org/pipermail/llvm-bugs/attachments/20160116/db0d38f1/attachment-0001.html>