Re: [Boost-users] [newbie] Boost 1.39 in Mac XCode get "error: attempt to copy-construct an iterator from a singular iterator."

Subject: Re: [Boost-users] [newbie] Boost 1.39 in Mac XCode get "error: attempt to copy-construct an iterator from a singular iterator."
From: Victor Tsang (victor.tsang_at_[hidden])
Date: 2009-07-29 03:23:20

I wonder how can you identify the bug in the source files as I was totally
no idea when I saw the error message.
Will you submit the fix to developer list?  :)

Regards,

Victor Tsang
victor.tsang_at_[hidden]

From: Christopher Jefferson <chris_at_[hidden]>
>
> On 28 Jul 2009, at 09:10, Victor Tsang wrote:
>
> > Hi,
> > ...
> >
> > The Debugger Debugger is attaching to process
> >
> > It generates a runtime error "error: attempt to copy-construct an
> > iterator
> > from a singular iterator.".
> >
> >
> > However, if I compile the code in terminal:
> > g++ -o test -g -Wall main.cpp -lboost_system-xgcc40-mt
> >
> > It runs okay without any runtime error.
> >
>
> To duplicate the error from the command line, add ' -D_GLIBCXX_DEBUG '
> to the command line, which turns on g++'s iterator debugging. Doing
> this, I can duplicate your problem.
>
> The problem can be boiled down to the following, illegal, code.
>
> #include <vector>
> #include <list>
>
> int main(void)
> {
>     std::vector<std::list<int>::iterator> v;
>     v.resize(1);
> }
>
> The code can be fixed (although there may be other occurrences of
> similar problems) by changing around line 220 on asio/detail/
> hash_map.hpp from:
>
>     // Update number of buckets and initialise all buckets to empty.
>     buckets_.resize(num_buckets);
>     for (std::size_t i = 0; i < buckets_.size(); ++i)
>       buckets_[i].first = buckets_[i].last = end;
>
> to:
>
>     // Update number of buckets and initialise all buckets to empty.
>     bucket_type bucket;
>     bucket.first = bucket.last = end;
>     buckets_.resize(num_buckets, bucket);
>
> I'm surprised boost isn't tested with iterator checking. Any
> particular reason, or has simply no-one tried?
>