Re: [Boost-users] [xpressive] crash...

From: Ovanes Markarian (om_boost_at_[hidden])
Date: 2007-10-16 11:30:15

I now know that this is an invalid usage, after I know that 'smatch' is
based on iterators. Since 'smatch' does not encover in the docs what
internal structure it has, I did not conclude that it is based on iterators.
And a string const& as param does not mean for me, that this string should
be valid later on. May be I am wrong.

BTW: the trick with volatile does not work in MSVC 8.0 Express compiler. :(
Would be great if the upcoming standard would provide a temporary checking
facility ;)

Many thanks for the great lib and your time!
Ovanes

On 10/16/07, Eric Niebler <eric_at_[hidden]> wrote:
>
> Ovanes Markarian wrote:
> > Hello Eric,
> >
> > this code fragment will crash, since 'what' uses iterators into the
> > invalid string
> >
> >         x::smatch what;
> >         if(x::regex_match(std::string("      (some_type_ptr) xyz,"),
> > what, regex)) //string goes out of scope after this sentance
> >         {
> >             std::cout << what[0] << endl; // boom! string iterators are
> > no longer valid
> >             std::cout << what[1] << endl;
> >         }
> >
> > May be there is a way to probit this kind of behaviour during the
> > compilation or at least this can be stated in the docs...
>
>
> As I'm sure you know, this is an invalid use. It couldn't hurt to be
> explicit about it in the docs, though.
>
> FWIW, I once found a way to disallow code like this at compile time.
> This regex_match() overload takes a "std::string const &", which accepts
> temporaries. It could instead take a "std::string const volatile &"
> which would *not* bind to temporaries, and then I can cast away the
> volatile. But it seems like a dirty trick, so I haven't done it.
>
> --
> Eric Niebler
> Boost Consulting
> www.boost-consulting.com
>
> _______________________________________________
> Boost-users mailing list
> Boost-users_at_[hidden]
> http://lists.boost.org/mailman/listinfo.cgi/boost-users
>