Re: [Bug libstdc++/15523] New: Can't have vectors of vector::const_iterator

[llewelly at xmission dot com]

"bangerth at dealii dot org" <gcc-bugzilla@gcc.gnu.org> writes:

> I definitely like the debug mode in libstdc++, but happened to fall over this: 
> ----------------- 
> #include <vector> 
>  
> int main () 
> { 
>   std::vector<std::vector<int>::const_iterator> x(2); 
> } 
> ----------------- 
>  
> When compiled with -D_GLIBCXX_DEBUG -D_GLIBCXX_DEBUG_PEDANTIC, I got this: 
>  
> homes/csm2/bangerth/bin/gcc-3.4.0/bin/../lib/gcc/i686-pc-linux-gnu/3.4.0/../../../../include/c++/3.4.0/debug/safe_iterator.h:125: 
>     error: attempt to copy-construct an iterator from a singular iterator. 
>  
> Objects involved in the operation: 
> iterator "this" @ 0x0x8067ee0 { 
> type = 
> N11__gnu_debug14_Safe_iteratorIN9__gnu_cxx17__normal_iteratorIPKiN10__gnu_norm6vectorIiSaIiEEEEEN15__gnu_debug_def6vectorIiS7_EEEE 
> (constant iterator); 
>   state = singular; 
> } 
> iterator "other" @ 0x0xbfffdd40 { 
> type = 
> N11__gnu_debug14_Safe_iteratorIN9__gnu_cxx17__normal_iteratorIPKiN10__gnu_norm6vectorIiSaIiEEEEEN15__gnu_debug_def6vectorIiS7_EEEE 
> (constant iterator); 
>   state = singular; 
> } 
>  
>  
> For some reason, I fail to see why I shouldn't be allowed to copy-construct from an 
> invalid iterator. After all, the result would still be invalid and
> couldn't be dereferenced, no?

Copying an invalid iterator is undefined behavior. 24.1/5 .

> I think this should be allowed. 

Is there a gcc target where (a) loading an invalid address into an
    address register causes a hardware fault, and (b) copy
    constructing an invalid iterator might require such a load?