Lars Gullik Bjønnes wrote:
> PreviewLoader.C:594: instantiated from here
>
/opt/gcc34/lib/gcc/i686-pc-linux-gnu/3.4/../../../../include/c++/3.4/bits/stl_list.h:84:
> error: `__gnu_norm::_List_node<_Tp>::_M_data' has incomplete type
> ../../src/ParagraphList_fwd.h:18: error: forward declaration of
> `struct Paragraph'
Thanks, Lars.
The code it doesn't like:
593 // Loop over the insets in the buffer and dump
all the math-macros.
594 Buffer::inset_iterator it =
buffer_.inset_const_iterator_begin();
595 Buffer::inset_iterator end =
buffer_.inset_const_iterator_end();
597 for (; it != end; ++it)
598 if (it->lyxCode() == InsetOld::MATHMACRO_CODE)
599 it->latex(buffer_, os, runparams);
In turn, buffer.h has this:
#include "ParagraphList_fwd.h"
class Buffer {
public:
class inset_iterator {
private:
ParagraphList::iterator pit;
ParagraphList::iterator pend;
};
};
So, my question becomes: do you need to know 'paragraph' to
instantiate std::list<paragraph>::iterator? I guess the answer to
that is 'yes' (or the compiler would not be complaining). That being
the case, then I also guess that the CORRECT place to #include
"paragraph" (and I hate to say this) is buffer.h itself.
Thoughts?
Why do non-g++3.4 compilers seem quite happy with this though?
--
Angus
