boost::filesystem::path operator>> drops leading slash.

[norman.wilson@…]

The problem manifests itself most obviously using boost::lexical_cast. See the attached test case.

[norman.wilson@…]

Unit test for lexical_cast<path>

[luke@…]

I ran into this same issue. However, I think its actually worse than dropping the leading slash. It drops the leading character whatever it is unless it is a double quote.

boost\filesystem\v3\path.hpp

template <class Char, class Traits> inline std::basic_istream<Char, Traits>& operator>>(std::basic_istream<Char, Traits>& is, path& p) {

std::basic_string<Char> str; is >> boost::io::quoted(str, static_cast<Char>('&')); p = str; return is;

}

The problem is coming from the boost::io::quoted function.

boost\io\detail\quoted_manip.hpp

extractor for non-const std::basic_string& proxies template <class Char, class Traits, class Alloc> std::basic_istream<Char, Traits>& operator>>(std::basic_istream<Char, Traits>& is,

const quoted_proxy<std::basic_string<Char, Traits, Alloc>&, Char>& proxy)

{

Char c; is >> c; if (c != proxy.delim) {

proxy.string = c; is >> proxy.string; return is;

} proxy.string.clear(); {

boost::io::ios_flags_saver ifs(is); is >> std::noskipws; for (;;) {

is >> c; if (!is.good()) cope with I/O errors or end-of-file

break;

if (c == proxy.escape) {

is >> c; if (!is.good()) cope with I/O errors or end-of-file

break;

} else if (c == proxy.delim)

break;

proxy.string += c;

}

} return is;

}

The function is checking if the first character in the stream is a double quote, if it is all works fine. However if its not it assigns the value of the first character to the proxy.string and then uses the insertion operator to get the rest of the characters. However the insertion is not an append it is an assignment so the effect is the first character is lost. I'm not exactly sure if this is expected behavior of this function of not. A workaround is to change the proxy.string = c; to is.unget(); This will preserve the first character.

[Richard Hazlewood <boost@…>]

This appears to have been addressed on the trunk, please can we ensure this makes it to the next (1.45) release?

​http://lists.boost.org/boost-commit/2010/08/27576.php

I know this is not specifically a filesystem issue, however, AFAIK fs is the only 'client' at the moment.

[anonymous]

Merged to release, revision 68159

Let me know if this doesn't solve the problems.

Thanks,

--Beman