id	summary	reporter	owner	description	type	status	milestone	component	version	severity	resolution	keywords	cc
12471	gzip compressor decompressor broken	anonymous	Jonathan Turkanis	"Hi guys, when I use the iostreams with the gzip tools it fails to give correct answers on final dezip. my data is all small integers separated by commas, and the recovered data interchanges the order of commas and numbers as if it was a bad multithreaded program. The result is worse than useless. This is a big file phenomena (up to gb size).

Looking at the compressor code (gzip.hpp) I see many problems between 32 and 64 bit. Currently I am looking at an assignment:
One that is problematic at the 
{{{
    template<typename Sink>
    std::streamsize write(Sink& snk, const char_type* s, std::streamsize n)
    {
        if (!(flags_ & f_header_done)) {
            std::streamsize amt = 
                static_cast<std::streamsize>(header_.size() - offset_);
            offset_ += boost::iostreams::write(snk, header_.data() + offset_, amt);
            if (offset_ == header_.size())
                flags_ |= f_header_done;
            else
                return 0;
        }
        return base_type::write(snk, s, n);
    }
}}}
offset_ is size_t while boost::iostreams::write(returns streamsize a signed 64 bit type (long long). So in 32 bit code there is a problem. Streamsize occurs many places, as does size_t; but one seems to be consistently 64 bit across platforms. offset seems to be 32 bit in 32 bit code.
I do not think this this is the cause of the issues with gzip but short of rewriting it I don't know where to start.

I would like ot use and have confidence in this compression code but so far it only gives grief...
"	Bugs	new	To Be Determined	iostreams	Boost 1.61.0	Problem