Boost C++ Libraries: Ticket #3299: boost regex regex_search crash

component changed; owner set

Steven Watanabe — Fri, 31 Jul 2009 19:29:09 GMT

owner set to John Maddock
component None → regex

Can you provide minimal code and input that reproduces the problem? What exception is thrown exactly?

ufwt@… — Tue, 04 Aug 2009 13:17:20 GMT

crash example:

boost::regex re("[a-z].*(d_notexiststring)",boost::regex::perl|boost::regex::no_except);

std::string s="f"; s.append(1024*100,'a'); s.append("dd"); boost::match_results<std::string::const_iterator> what; boost::match_flag_type flags = boost::match_default;

std::string::const_iterator start, end; start = s.begin(); end = s.end(); boost::regex_search(start,end,what,re,flags);

except:

uncaught exception of type boost::exception_detail::clone_impl<boost::exception_detail::error_info_injector<std::runtime_error> >

Memory exhausted

raise at state_count > max_state_count

status changed

John Maddock — Thu, 06 Aug 2009 16:22:32 GMT

status new → assigned

The code doesn't crash: it throws an exception that it's documented to throw.

In this case it's because the number of states visited in the FSM has grown too large and it bails out to prevent an "eternal" match attempt.

Perl manages to optimise this particular expression, but if you make it trivially more complex, say '[a-z].*(xxxxx)|x', then it takes a very long time indeed to return from a match attempt on your string.

I'll leave this open for now though, as it looks like this special case can be optimised a little more.

John.

ufwt@… — Thu, 06 Aug 2009 16:26:26 GMT

so i use try catch to avoid program crash!!

ufwt@… — Thu, 06 Aug 2009 16:31:02 GMT

i also change re_repeat's max let re_pepeat max - min < 1024 i can change this ?

anonymous — Fri, 07 Aug 2009 11:24:38 GMT

"i also change re_repeat's max let re_pepeat max - min < 1024 i can change this ? "

Sorry I don't understand what you are asking. Change what code precisely?

ufwt@… — Sat, 08 Aug 2009 13:20:33 GMT

i chanage re_repeat'max , code:

void changeRegexRepeatNum(boost::regex &regex,size_t max) {

boost::re_detail::re_syntax_base* state=regex.get_data().m_first_state; while(state){

switch(state->type){

case boost::re_detail::syntax_element_rep: case boost::re_detail::syntax_element_dot_rep: case boost::re_detail::syntax_element_char_rep: case boost::re_detail::syntax_element_short_set_rep: case boost::re_detail::syntax_element_long_set_rep: {

boost::re_detail::re_repeat *repeat=static_cast<boost::re_detail::re_repeat*>(state); if(repeat->max - repeat->min >max){

repeat->max = repeat->min + max;

}

} break; default:

break;

} state = state->next.p;

}

anonymous — Sat, 08 Aug 2009 16:52:12 GMT

Sure you can do that, but I'm not sure why you would want to?

John.

ufwt@… — Sun, 09 Aug 2009 08:15:48 GMT

because my txt is very long ,i do this can regex_search speed quick!

status changed; resolution set

John Maddock — Tue, 02 Mar 2010 17:02:11 GMT

status assigned → closed
resolution → wontfix