[Bug tree-optimization/49217] New: Wrong optimization of code

["personal at e-maxx dot ru" <gcc-bugzilla at gcc dot gnu dot org>]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=49217

           Summary: Wrong optimization of code
           Product: gcc
           Version: 4.6.0
            Status: UNCONFIRMED
          Severity: normal
          Priority: P3
         Component: tree-optimization
        AssignedTo: unassigned@gcc.gnu.org
        ReportedBy: personal@e-maxx.ru


Created attachment 24387
  --> http://gcc.gnu.org/bugzilla/attachment.cgi?id=24387
The program demostrating the bug (should return 0 if OK, but throws in case of
wrong optimization)

A bug in gcc 4.6.0 optimizer leads to incorrect program flow.

It looks like this exact bug is difficult to reproduce in small program, so the
attached program is a result of my minimization.

The bug requires quite complex environment: recursive function, std::vector,
std::vector::iterator, zeroing the global array, unreachable 'else'.

Though, I think, the main idea of the program is not so difficult: we call
DFS(0), it calls DFS(1), and as a result both 0 and 1 are marked as visited, so
the number of calls of DFS() from main() is equal to one. But in g++ 4.6.0 with
-O2 that's not true (it looks like DFS(0) never calls DFS(1), but that's my
hypothesis).

Inserting cerr or cout into some places of DFS makes the program work fine.